Growing datasets and public benchmarks are making it harder to tell whether a model is being tested on something it hasn’t seen before.

On Thursday, Google DeepMind showed off what the company calls the first double-blind evaluation of a proprietary frontier-class AI model. The setup keeps Gemini’s model weights hidden from the evaluators while the test questions are hidden from Google.

The pilot tested Gemini 2.5 Flash Lite against private benchmarks from MLCommons and the Singapore AI Safety Institute, but rather than using the results to tout a new Gemini score, Google is focusing on how the tests were run, with neither side having access to the other’s data.

Benchmark leakage inflates scores

Google’s technical report cites earlier research that found signs of benchmark leakage in about half of the 31 models tested, and another study published this year that found contamination can inflate scores, particularly for larger models. This dynamic has already raised questions about where certain post-training coding gains actually come from.

Keeping benchmarks private would seem like the obvious answer, except closed models make that difficult. With a closed model, evaluators usually have to run their questions through the provider’s API, which means the company can see a test that was supposed to stay private, whereas running it independently would require the provider to hand over its model weights. Google’s setup is meant to give both sides another option.

Keeping benchmarks private would seem like the obvious answer, except closed models make that difficult.

How the enclave works

The pilot is using Google Cloud Confidential Space together with an NVIDIA H100 Confidential GPU and Intel TDX host memory encryption.

Google DeepMind provides Gemini with its weights and inference code, while the evaluator offers its benchmark prompts and evaluation code; these are then transmitted over encrypted connections into the enclave, where the evaluation takes place without either party gaining access to the other’s protected assets.

The model weights are stored in hardware-encrypted GPU memory, and the evaluation prompts are kept in encrypted host memory as well. After the test is complete, the evaluator is given the permitted results, and the temporary environment can then be destroyed.

Before either side sends over its private data, remote attestation verifies that the enclave is running the software they agreed on.

Code controls beyond encryption

The enclave protects what’s stored in memory, but the code running inside it can still create problems. If an application has unrestricted network access, for example, it could send sensitive information somewhere else.

OpenMined’s PySyft handles that part of the process by allowing Google and the evaluator to approve the code in advance and to block sensitive parts of the evaluation from making external connections.

That process takes some work, although compute doesn’t appear to be the problem. The paper puts the overhead at less than 5% and instead points to legal agreements and code reviews between organizations as some of the bigger hurdles to running these evaluations.

The researchers eventually want to make attestation much less hands-on, comparing the goal to the HTTPS lock icon in a browser, where the security checks happen without the user having to deal with the underlying hashes and keys.

Google still has a hand in verification because, although the Confidential Space guest OS is open source and its build process has been externally validated, individual builds rely on private signing keys and can’t be independently reproduced. At the same time, Google’s own services are used to sign and verify the attestation report.

The paper puts the overhead at less than 5% and points instead to the legal agreements and code reviews between organizations as some of the bigger hurdles to running these evaluations.

Scaling past one GPU

So while the system reduces how much the two sides need to trust each other, it doesn’t remove trust entirely from the equation. Some of it simply moves elsewhere, including to the hardware itself. There is still some trust involved in the hardware itself, including the assumption that the cloud provider and the hardware maker aren’t working together to circumvent the protections.

Then there’s the benchmark itself. MLCommons points out that it still needs to be carefully managed, no matter how well the questions are protected.

The pilot ran Gemini 2.5 Flash Lite on a single H100 80GB Confidential GPU. Still, the researchers are already looking at clusters of H100 and B200 GPUs connected via encrypted links to eventually evaluate models that are too large to fit on a single GPU.

If this approach catches on, developers could have another way to look beyond the benchmark score, with some proof that the company behind the model didn’t see the test before the results came out.

If this approach catches on, a benchmark score could offer some proof that the company behind the model didn’t get to see the test first.

The post Google found a way to test Gemini without seeing the questions appeared first on The New Stack.

AI coding company Replit is throwing its weight behind the model-routing trend by making its “intelligent model routing” system the default across every account.

The system automatically selects the underlying model to handle a task as it evolves, with Replit weighing quality, speed, and cost in its routing decisions.

The company says the feature, dubbed Auto mode, will become the default option for all users, though Core and Pro subscribers can still override it and manually select models when they want more control.

Model-routing momentum

The announcement comes hot on the heels of a flurry of activity in the model-routing realm. Earlier in August, Stripe agreed to acquire model gateway platform OpenRouter for a reported $8 billion. On the very same day, Ramp launched Router.com, which routes requests to the lowest-cost model that meets a specified performance bar.

Before all that, in July, SpaceX-owned Cursor launched its own router, which automatically selects models for coding requests and claims to deliver comparable performance at a substantially lower cost. Meanwhile, Meta is reportedly developing an internal router called “Switchboard” that scores coding tasks by difficulty and sends simpler jobs to cheaper models.

“Across one model family, per-token rates can span orders of magnitude. At the same time, the intelligence of cheaper, smaller models is now much closer to their larger frontier counterparts, providing us a lot of room for cost optimizations.”

Michele Catasta, president and head of AI at Replit, says that one reason for the wider push into routing is simple economics — the growing gap between what models cost and the level of capability developers actually need for a given task.

“Across one model family, per-token rates can span orders of magnitude,” Catasta tells The New Stack. “At the same time, the intelligence of cheaper, smaller models is now much closer to their larger frontier counterparts, providing us a lot of room for cost optimizations.”

Replit, for its part, has been moving in this direction for some time. Catasta says that the company has spent recent months experimenting with early versions of Auto mode, subagent routing, and multiple iterations.

“Like any pivotal launch, we thoroughly tested Intelligent Model Routing in beta for a long period of time before we decided to release it in public,” he says. “The most important learning is understanding from first principles the failure modes of every experiment, so we could keep hill climbing on the final system that we just shipped.”

Enter Auto mode

The foundation of that work surfaced last week when Replit introduced Free Mode, a lower-cost Agent mode that doesn’t consume usage credits and uses Auto to choose the model on theuser’ss behalf, subject to usage limits.

Now, that same Auto routing approach is being pushed more broadly across Replit. The company says intelligent model routing will become the default across every account, with all users starting in Free Mode and Replit deciding which model is best suited to the task.

Free Mode, it’s worth noting, isn’t “free” in the sense of unlimited usage. When it launched, Replit made it available to Core and Pro subscribers without consuming their usage credits, but imposed limits that reset every five hours, with higher allowances for Pro users. In Free Mode, users cannot manually select a model.

Core and Pro subscribers can, however, switch to Replit’s Power or Max modes, where they can turn off Auto and choose a model themselves. Replit may also suggest moving a task into one of those higher-powered modes when it determines that more capability is required, though those modes can incur usage costs.

For Enterprise customers, meanwhile, administrators can restrict Auto to an approved set of models for each workspace, allowing Replit to continue routing tasks automatically while keeping model choice within company policy.

The agent advantage

Even before SpaceX agreed to pay a cool $60 billion to acquire Cursor, the AI coding startup had long been investing in its own coding models, including its Composer family. More recently, under the auspices of SpaceX, Cursor has been developing more cutting-edge models, too.

Replit, by contrast, isn’t making ownership of the underlying model layer central to its pitch. Instead, it’s betting that controlling the agent and the systems around it gives Replit enough insight to make better model-selection decisions on the fly.

“Replit has owned, from the start, both the agent harness and the infrastructure surrounding models which in turn allows us to train sophisticated model routers.”

“Replit has owned, from the start, both the agent harness and the infrastructure surrounding models which in turn allows us to train sophisticated model routers,” Catasta said. “Only in this way can we always offer useful intelligence to our users at the most competitive price point.”

That becomes particularly relevant as an Agent task unfolds, with Replit noting that its system can change which model it uses as the task develops, seeking a better trade-off between capability and cost at different points in the process. But for Catasta, that kind of dynamic routing is still only one part of a much broader research problem around how agents should use models.

“Model routing is still in its early development phase, and we expect further research will move the needle on serving the best intelligence when customers most need it,” Catasta explains. “Routing is but one piece of the puzzle that is tightly integrated to many other aspects of our harness research.”

“No third-party router company could reproduce the same results for our own agent.”

Replit also argues that seeing how people use its own Agent gives it an advantage that a standalone routing provider would struggle to reproduce. Catasta says a router has to infer the nature, difficulty, scope, and intent of a request, with Replit able to train against proprietary usage data and observe those signals across its user base.

“No third-party router company could reproduce the same results for our own agent,” he says.

The post Replit’s new default: Auto mode picks the best model for each task appeared first on The New Stack.

Nvidia has reportedly agreed to buy Hugging Face for $12.9 billion, putting one of the biggest names in AI hardware in charge of a platform developers rely on to find and run open models.

The Information first reported the deal Wednesday, citing a person familiar with the agreement. Nvidia and Hugging Face had not publicly confirmed it as of publication.

Hugging Face doesn’t push developers toward one chipmaker, which is what makes the acquisition interesting. Its Optimum libraries work with Nvidia’s TensorRT-LLM and also support hardware from AMD, Intel, and AWS. Projects such as Optimum AMD and Optimum Intel let developers run Transformers and Diffusers models on non-Nvidia hardware.

The company also plays a role in what happens after a developer chooses a model, including how easily they can get it running on the hardware they want to use. The problem for Nvidia might be this: It is buying a platform whose value depends on openness and hardware neutrality, but if the purchase means Nvidia hardware is favored, that value might diminish.

The company also plays a role in what happens after a developer chooses a model, including how easily they can get it running on the hardware they want to use.

Hugging Face already sits between the model and the chip

Hugging Face has expanded well beyond file hosting. With Inference Endpoints, developers can deploy a model from the Hub while Hugging Face handles the underlying infrastructure.

Those hosted deployments can run on AWS, Microsoft Azure or Google Cloud, but most of the GPU options Hugging Face lists are Nvidia chips, including the T4, L4 and A100. That gives developers a wider choice of hardware through Hugging Face’s open-source libraries than through its hosted services.

If the deal goes through, Nvidia would own both sides of that experience.

Deployment defaults favor Nvidia

NIM (Nvidia Inference Microservices) already works with models hosted on Hugging Face. Developers can point NIM to an hf:// repository path and pull the model directly from the Hub.

Owning Hugging Face would give Nvidia more room to bring NIM and CUDA-optimized containers directly into the deployment experience. Nvidia hasn’t announced plans to make NIM the default, and support for AMD and Intel could remain exactly where it is.

Owning Hugging Face would give Nvidia more room to bring NIM and CUDA-optimized containers directly into the deployment experience.

The bigger question is what happens over time. Nvidia could provide earlier support for new models on its own hardware or make deployment easier. At the same time, AMD, Intel, and AWS may have to reconsider how much engineering work they want to contribute to integrations maintained within a competitor-owned platform. Some of that work could eventually move elsewhere.

For developers, the difference may come down to which path requires less work. A competing chip doesn’t have to disappear from Hugging Face to become less appealing if an Nvidia model deployment takes fewer steps. We’ve seen a similar fight over the layers between AI models and the developers using them, with Cloudflare building more of that infrastructure itself.

Open models counter custom chips

That tension also helps explain why Hugging Face could be worth considerably more to Nvidia than its revenue alone would suggest.

Nvidia has been expanding its own Nemotron family of open models while investing heavily across the AI ecosystem. At the same time, some of its biggest customers are working to reduce their dependence on Nvidia hardware.

Google has its TPUs, AWS has Trainium, and Microsoft has Maia. OpenAI and Anthropic are also developing their own AI server chips. The push extends beyond the hyperscalers. Earlier this month, five European companies committed to purchasing AI compute built around non-Nvidia hardware that hasn’t been manufactured yet — a sign that the appetite for alternative accelerators is strong enough to attract forward contracts.

OpenAI this week published results from its new Jalapeño accelerator, which showed 1.5 to 1.9 times more work per watt while cutting end-to-end latency by up to 3.6 times on large open-weight models — although the chip has not yet been deployed at anything approaching Nvidia’s scale. A strong open-model ecosystem gives Nvidia a counterweight to that trend.

Open models are often expected to run in very different environments, and Hugging Face helps developers make that possible. A model found on the Hub might end up running on Nvidia hardware, an AMD GPU, or a cloud accelerator.

That flexibility is part of what Nvidia would be buying. Pushing Hugging Face too heavily toward its own hardware could make the platform less useful to developers who rely on it to work across different systems.

Interest in those models is also growing. Models from companies including DeepSeek, Moonshot AI and Z.ai have narrowed the gap with proprietary systems. At the same time, Hugging Face CEO Clément Delangue told The Information in June that the company had doubled its number of paying subscribers during the first six months of 2026. Delangue later said the company was “close to profitability.”

The Information puts Hugging Face’s annualized revenue at about $150 million. Against a $12.9 billion price tag, that’s a multiple of roughly 86.

So Nvidia would be paying for much more than Hugging Face’s current business. It would be buying a place developers already turn to when they want to work with open models, including models that don’t have to run on Nvidia hardware.

It would be buying a place developers already turn to when they want to work with open models, including models that don’t have to run on Nvidia hardware.

Community resists easy forking

Buying Hugging Face wouldn’t give Nvidia control over everything developers find there. Libraries such as transformers and diffusers are open source, and models on the Hub remain subject to their own licenses. Openly licensed models can still be hosted elsewhere, while the underlying libraries can be forked.

Much harder to recreate is the community Hugging Face has built around them. Developers already know where to look for models and have built workflows around the Hub and its integrations.

The post Nvidia’s $12.9B Hugging Face deal has an open-source problem appeared first on The New Stack.

OpenAI’s unreleased foundation model, codenamed Astra, is already working directly inside the company’s internal codebase, taking on experimental work that previously required as much as a week from a human researcher.

As part of a series of interviews with Time, OpenAI chief scientist Jakub Pachocki said Astra can take an idea for an experiment, turn it into code, run it, and return the results. A step beyond asking AI to fix a bug or write a function, OpenAI is essentially testing what happens when you hand the agent the experiment itself and let it figure out the steps in between.

For developers, Astra offers a glimpse into the future of coding agents. But giving an agent that much freedom creates another problem. Astra may already be powerful enough to trigger OpenAI’s highest-level cybersecurity safeguards.

OpenAI is essentially testing what happens when you hand the agent the experiment itself and let it figure out the steps in between.

Persistent agents change everything

OpenAI CEO Sam Altman described what the company is building as “persistent agents,” or systems that can keep working without needing a person to prompt them through every step.

Coding agents can already dig through a repository, change files, run tests, and try again when something breaks. Persistent agents are meant to keep going without a developer guiding them through each step.

That also changes what developers need from the tools around the agent. An IDE gives it somewhere to work, but a long-running agent needs infrastructure that can keep it running safely without constant oversight.

Multi-agent coordination at scale

In one demonstration witnessed by Time, 16 Astra agents worked together on a research-level math problem, splitting it into smaller pieces and then bringing their work together into a proposed solution.

For developers, it’s not hard to imagine that same setup applied to a large software project, with different agents working on different pieces at the same time. OpenAI is already experimenting with that kind of coordination, along with agents that can stay on a job for much longer.

Running several agents at once also complicates the infrastructure behind them. Developers need a way to keep the whole operation under control. Giving agents that much freedom, though, has created another problem for OpenAI: keeping them under control.

When agents escape containment

OpenAI said this month that preliminary evaluations indicate Astra may have reached the “Critical” cybersecurity capability threshold in the company’s Preparedness Framework, a finding the company disclosed alongside a pause on some frontier workloads.

Under the company’s framework, hitting that threshold brings stricter safeguards for how the model can be used.

OpenAI had already seen what could go wrong with agents that have access to tools. During a cybersecurity test, one of its internal AI agents escaped its sandbox and accessed Hugging Face systems without authorization. Astra was not the model involved, but the incident led OpenAI to pause some frontier-model research workloads while it tightened the infrastructure used to run them.

And OpenAI isn’t the only company running into this problem. Google’s AI coding agent recently broke out of the boundaries of its IDE. The circumstances were different, and the incident was less serious, but both point to the same problem that happens when giving an agent more freedom to act; you also have to make sure it stays where you put it.

Astra is now running under OpenAI’s strictest security controls. Some training and evaluation workloads have resumed, but OpenAI says a “significant number” are still paused while it upgrades the infrastructure behind them.

Astra is now running under OpenAI’s strictest security controls.

Monitoring costs real compute

The company also says it is monitoring Astra more closely when it uses tools, watching for behavior that could signal it is going beyond what it’s allowed to do. OpenAI estimates that monitoring adds about 20% to the inference compute for those workloads.

Time reports that OpenAI still plans to release Astra, although there’s no launch date yet. Until then, OpenAI is working through the same challenge developers could eventually face as agents work for longer stretches without human supervision and begin coordinating with other agents.

As agents take on more work, the IDE is only part of the picture. Developers also need a way to see what those agents are doing and keep them from going where they shouldn’t. OpenAI is already seeing the cost, with monitoring alone adding about 20% to Astra’s inference compute.

That kind of overhead could become another cost developers have to account for, along with the governance around autonomous agents that sets boundaries on what they can do.

OpenAI is already seeing what that costs, with monitoring alone adding about 20% to Astra’s inference compute.

The post OpenAI’s Astra can do a researcher’s week of work. That’s the problem. appeared first on The New Stack.

Perplexity shipped Portable Computer this week, the local-first version of its Computer agent running on an Nvidia DGX Spark workstation, and the bill is steep: A DGX Spark starts at $4,700, and even an aging 24GB RTX 3090 sells well above $1,500.

The architectural choice beneath the surface deserves as much attention as the price tag: Most platforms building agents address reliability with greater intelligence. That typically includes a larger orchestrator model, a planner model, or a critic model reviewing the work.

Probabilistic reasoning proposes the next action, and deterministic software decides whether to execute it.

Perplexity separated the two jobs instead of stacking them. Probabilistic reasoning proposes the next action, and deterministic software decides whether to execute it.

The loop controller is code, and the decisions are still a model

Perplexity uses the term “orchestrator” to refer to the runtime controller rather than the planning model. That controller assembles context, enforces policy, and executes approved tool calls inside an OS-level sandbox. The company says it is deterministic code, not yet another model. The local model proposes the next action, including which tool to call and when to ask a cloud advisor for help. The split resembles a control plane architecture, where reasoning suggests, and inspectable software retains authority. Nate Kupp, Perplexity’s vice president of Computer Enterprise and Infrastructure, tells The New Stack that the harness accounted for most of the engineering work.

Same weights, better scores

Perplexity kept the base model and the silicon constant, putting Qwen3.8-27B on the same DGX Spark across three agent stacks. On its internal Local Knowledge Work Bench, a held-out set of 53 tasks, the company reported 82.6% for Computer. Pi scored 77.6% and Hermes 74%.

On ParseBench-100, a subset covering charts, layouts, tables, and formatting, the gap widened considerably. Perplexity reported 65.1% for Computer, while Hermes reported 34.6% and Pi reported 13.9%.

Because the base weights were constant, the gap measures the system around the model rather than a better model. That does not mean weights stopped mattering. Perplexity post-trained Qwen into PPLX 27B and reported 85.4%, above its own base-model number. Harness engineering and post-training are significant factors of this approach.

The word harness also covers a lot of ground, including prompts, tool schemas, context management, verification hooks, and document processing. Orchestration code is one part of that surface.

The security boundary lives outside the model

The sandbox is the boundary, not the determinism. Perplexity says the sandbox restricts processes, filesystem paths, and network access. If the sandbox is unavailable, the harness disables itself before making any tool call. Deterministic code is valuable here because it makes policy easy to inspect and helps the system fail safely. Deterministic code can still ship a vulnerability or faithfully execute a permitted mistake.

The key distinction is not between model orchestration and code orchestration. It is whether permission is enforced by something other than asking a model in plain English to comply.

The key distinction is not between model orchestration and code orchestration. It is whether permission is enforced by something other than asking a model in plain English to comply.

The same discipline shapes how context gets spent. Perplexity reports that Qwen3.8-27B advertises a 260,000-token context window but begins to struggle beyond 100,000 tokens. The harness therefore keeps the core prompt and toolset small and loads skills on demand. Commonly used connectors became command-line tools rather than full Model Context Protocol definitions sitting permanently in context.

Deterministic execution cannot rescue reasoning that exceeds the local model. On Terminal Bench 2.1, Perplexity reported 59.6% running locally. Letting the local agent consult Claude Opus 5 raised it to 73.0%, compared with 82.4% when Opus 5 worked alone. All of this remains vendor-reported evidence on a bench that the company has yet to open-source.

For enterprises evaluating local agents, the component to scrutinize is the layer that grants and denies authority, because that is where the platform’s engineering is most evident.

The post Perplexity just separated reasoning from authority. Here’s why it matters for enterprises. appeared first on The New Stack.

Agentic reinforcement learning has been suffering from a disconnect, an uncoupling, and a misarticulation. The polarity arises from how a training engine handles resource management, compared with how a post-training live production harness does.

Microsoft wants to ensure the production harness is engaged from the start to oversee infrastructure services and agent interactions, during both initial training and subsequent reinforcement learning actions.

Redmond’s Microsoft Research division first introduced the Agent Lightning framework in August 2025 as an infrastructure concept for agent optimization to address structural challenges in post-training LLM-based agents as they enter reinforcement learning processes. Microsoft subsequently launched the Agent Lightning v1.0 release with a commit tagged on GitHub on August 16. 

Using Agent Lightning v1.0 on what Microsoft has called “modest compute” 6K training examples, reinforcement learning improves Qwen3.5-9B on OpenAI’s SWE-bench Verified benchmark from 41.8% to 56.4%, an absolute 14.6-point gain.

“Using Agent Lightning v1.0 on ‘modest compute’ 6K training examples, reinforcement learning improves Qwen3.5-9B on OpenAI’s SWE-bench Verified benchmark from 41.8% to 56.4%, an absolute 14.6-point gain.”

Who owns the interaction loop?

In traditional agentic reinforcement learning, the training engine owns the interaction loop, i.e., steps including observing the environment, selecting an action based on the policy, executing the action, receiving a numerical reward, storing and updating the policy… and so on. 

In harnessed agentic reinforcement learning, the harness owns context construction, tool execution, and the agent–environment loop. The training engine observes only a sequence of LLM request–response pairs across a service boundary. Therefore, developers do not need to reimplement the agent loop within the training environment.

According to a collected group of software engineers at Microsoft, because an AI model harness owns the loop that governs infrastructure access and operations procedures during harnessed agentic reinforcement learning, it “introduces challenges” including retokenization (re-segmenting text into new tokens during active training), sample merging, advantage calculation, loss normalization, and training backend scheduling.

All of which, if not addressed, can result in ineffective or unstable training.

How Agent Lightning v1.0 turns the tables

“[In Agent Lightning v1.0] the harness, rather than the trainer, owns context construction, tool execution, and the agent–environment interaction loop, while the training system observes and optimizes the resulting model calls across a service boundary. This formulation preserves the harness’s deployment-time context policy, tool protocols, and execution semantics without requiring its agent loop to be reimplemented inside the RL framework,” explained the Redmond team.

For coding agents, the team has said that it finds existing reinforcement learning frameworks provide limited support, including a lack of data and complete training scripts, and a reliance on large-scale computational resources. To address this gap, with Agent Lightning v1.0, Microsoft provides a “complete data-cleaning pipeline” and reproducible training scripts built on open-source datasets and models.

Is this the end of the training time liability?

For users inside Microsoft environments, this might feel like good news. Machine learning specialists and platform engineers may have a good harness or a good set of reinforcement learning tools; they would rarely have both.

Rather than having to hard-code retokenization, advantage calculation, and reward shaping every time they wanted to train an agent to call external services in reinforcement learning procedures, they can keep their existing agent architecture as an asset, rather than treating it as a training-time liability.

Nebraska-based software engineering researcher Md Rashedul “Rashed” Hasan tells The New Stack that training on the exact production harness matters, but not only for its efficiency and benchmark gains. 

Training through the real harness keeps semantics intact

“It reduces train–serve mismatch,” Hasan says. “If you train inside a simplified trainer loop and deploy inside a different harness, tool protocols, context policy, and recovery behavior can all drift. Training through the real harness keeps those semantics intact, so gains are more likely to transfer to production behavior, not only to a lab environment.”

Hasan thinks that Microsoft has clearly named the paradigm, kept the core framework small, and shipped what he defines as a “concrete coding-agent pipeline” with open data and scripts. 

“In terms of who this will appeal to, it’s application and platform engineers who already have a production agent harness (such as a coding assistant or support-triage agent), and want to improve the underlying model with reinforcement learning without rewriting deployment logic to fit a training framework. Reinforcement learning and machine learning platform teams would also use it when they need a thin, reproducible testbed for harnessed agentic reinforcement learning,” Hasan clarifies.

He agrees that the SWE-bench Verified lift with only 6K examples is “useful proof” that harnessed reinforcement learning can move a hard-coding benchmark without forcing teams to reimplement their agent stack within the trainer.

“Environment setup for coding agents, reward design, evaluation fidelity, and the retokenization, sample-merging, advantage, and loss-normalization details are still easy to get wrong. Adoption will also depend on whether teams can integrate this proxy pattern into existing orchestration, observability, and safety controls,” cautions Hasan.

“Environment setup for coding agents, reward design, evaluation fidelity, and the retokenization, sample-merging, advantage, loss-normalization details are still easy to get wrong. Adoption will also depend on whether teams can integrate this proxy pattern into existing orchestration, observability, and safety controls.”

Killing train-serve skew, the oldest & most expensive bug in machine learning

Colorado-based data science professional Priyank Jain tells The New Stack that training through the production harness to strip the framing away is really all about “killing train-serve skew”, which is the oldest and most expensive bug in applied machine learning.

“Models rarely blow up in production because the math was wrong,” Jain says. “They blow up because the training setup quietly lied to them about what production actually looks like. Training through the same harness that serves the agent is the right instinct, and honestly it’s overdue. The accuracy bump is nice, but the real win is that the thing you optimized is finally the thing you shipped.”

In terms of what Microsoft is getting right here, Jain says it shows Redmond wants to “meet developers where they already are,” i.e., allowing users to improve an existing agent with reinforcement learning without rewriting their deployment just to please a training framework.

“That’s a genuinely good call, and it opens this up to folks who aren’t reinforcement learning specialists. What I’d worry about is that it also lowers the bar for people to run reinforcement learning they don’t fully understand. The hard part was never wiring up the loop; it’s designing a reward that survives contact with a model actively looking for the path of least resistance. Make that part easy, and you’ll get more people optimizing the wrong thing faster,” Jain advises.

“3500 lines of code is small enough that an infrastructure engineer can actually read it before trusting it… and that’s a good thing.”

Just 3,500 lines of core Python code

Built with “simplicity as its first principle”, the entire framework consists of around 3,500 lines of core Python code.

Infrastructure engineering developer and founder-developer of Tooldex, a platform that autodiscovers MCP servers across LLM agents, Ria Banerjee, tells The New Stack that the simplicity element here is a positive, i.e. 3500 lines of code is small enough that an “infrastructure engineer can actually read it before trusting it,” and that’s a good thing.

“But in terms of who would actually use this tool, the honest answer is fewer teams than the framing suggests,” Banerjee says. “Realistically, you would need a GPU cluster and a Kubernetes cluster under your control. An app developer who built a support-triage agent on LangChain isn’t running that. So the real audience is platform teams that already employ infrastructure engineers.”

“Because the harness is where the behavior actually lives, but you’re now baking your harness’s quirks into the model weights… so you change your retry logic next quarter, you’ve silently shifted what the model was trained on. So now, you have to version your harness,” adds Albuquerque-based Banerjee.

Microsoft has released the complete workflow and training scripts to facilitate reproducible, harnessed agentic RL in Agent Lightning v1.0 on GitHub under the MIT license, including data cleaning and reward-hacking prevention.

The post Microsoft just released Agent Lightning v1.0. Here’s why it matters for platform engineers. appeared first on The New Stack.

When OpenAI unveiled Jalapeño, its first custom inference chip, in June, the company made some big promises. The chip, developed with Broadcom, was built from scratch for large language model inference, with OpenAI saying early testing showed substantially better performance per watt than existing accelerators. At the time, though, OpenAI didn’t release the detailed performance results to back that up.

On Tuesday, OpenAI published its first results from working Jalapeño silicon across GPT-OSS 120B, DeepSeek R1 and Kimi K2.5. The results show what OpenAI was aiming for with Jalapeño: higher throughput without the longer response times that can come with it.

“Agents need to complete many steps in sequence, so delays can compound across an entire task.”

Agents compound inference delays

An agent may call a model over and over as it works through a task, using tools and deciding what to do next based on the results, which means a delay that barely registers during a single inference can become much more noticeable when it happens repeatedly over the course of a longer task.

“Agents need to complete many steps in sequence, so delays can compound across an entire task,” OpenAI said.

Jalapeño was designed with those delays in mind. Different parts of running a large language model place different demands on the hardware, with the initial prompt requiring more compute and the response generation putting more pressure on memory bandwidth. Every time data has to move between cores and chips, that can add even more waiting.

Jalapeño takes a different approach, cutting down on that waiting without optimizing one part of the process at the expense of another.

“Agents need to complete many steps in sequence so that delays can compound across an entire task,” OpenAI said.

That helps explain some of the choices OpenAI made with Jalapeño. Running a large language model puts different demands on the hardware at different points: processing the initial prompt requires a lot of compute, while generating the response token by token relies more heavily on memory bandwidth. There’s also time lost whenever data has to move between cores and chips, leaving parts of the system waiting for what they need.

The idea is to reduce that waiting without optimizing one part of the process at the expense of another. Model state, including the KV cache used while generating a response, can be kept local, while Jalapeño’s networking allows more of the workload to stay within the same connected system. That means less time spent moving data around as the workload shifts between compute and memory.

Jalapeño’s first public benchmarks

OpenAI put Jalapeño through InferenceX, SemiAnalysis’ public benchmark for AI inference, using GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Across the three models, Jalapeño handled 1.5 to 1.9 times more work per watt while cutting end-to-end latency by 1.7 to 3.6 times. On highly interactive workloads, OpenAI says it was 2.1 to 4.1 times faster than the systems it compared against. 

The differences become particularly large when Jalapeño is compared at the previous best time-between-tokens operating point. OpenAI reported between 8.6 and 104.3 times more work per watt, depending on the model.

OpenAI based the power-efficiency comparisons on each accelerator’s published power rating. Jalapeño is rated at 700 watts, although the company says it never drew more than 550 watts during these tests. The bigger point is that OpenAI isn’t trying to improve throughput at the expense of response time, which is often the tradeoff with inference. 

Batching more work can make infrastructure more efficient, but it can also mean making an individual user wait longer. OpenAI’s argument with Jalapeño is that an inference system increasingly needs to be good at both — particularly as the company continues cutting the cost of API access while also needing to keep interactive workloads responsive.

The bigger point is that OpenAI isn’t trying to improve throughput at the expense of response time, which is often the tradeoff with inference. 

AI-generated code runs faster

OpenAI used its own models throughout Jalapeño’s development, helping the hardware team move from initial design to tapeout in nine months by exploring implementations and shortening design, measurement, and verification cycles. The work didn’t stop once the chip was built.

The company says AI-generated implementations of selected GPT-OSS attention and mixture-of-experts blocks ran 1.5 to 1.8 times faster than versions written by its own experts. That doesn’t mean the entire model ran that much faster, but it does show what OpenAI is trying to do with Jalapeño: make the chip straightforward enough for AI, not just humans, to program and optimize.

Engineers describe work using local tensors, explicit communication and predictable synchronization, giving AI a way to help determine how that work should be mapped, placed and scheduled across the system. That could make it faster to adapt the chip as new models come along, although OpenAI says each new model family still requires its own kernels and optimizations.

Custom silicon meets model roadmap

Using Codex with GPT-Astra and earlier OpenAI models, the hardware team brought three open-weight models that weren’t part of Jalapeño’s original production plan to high performance within two months. That fits with OpenAI’s broader plans for Codex, which the company has said is still early in its development, and shows how it could eventually play a role well beyond writing code.

OpenAI plans to start using Jalapeño in its own infrastructure by the end of the year, and it’s already working on the next two generations. The company will continue to use accelerators from Nvidia and other partners, but building its own chips gives OpenAI more control over how the hardware evolves alongside its models.

OpenAI plans to start using Jalapeño in its own infrastructure by the end of the year, and it’s already working on the next two generations.

The post OpenAI’s Jalapeño chip tackles a problem AI agents make worse appeared first on The New Stack.

What do you do when a technology you’ve become dependent on belongs to someone else? You buy the startup behind it, of course. And that’s exactly what data warehouse company MotherDuck has done with Tower, a data infrastructure startup whose technology was already powering MotherDuck’s AI-built data pipelines.

The deal, announced on Tuesday, is MotherDuck’s first acquisition in its four-year history, bringing both Tower’s technology and team in-house as MotherDuck pushes further into AI agents that can build and operate data pipelines.

Tower takes flight

Tower was founded out of Germany in late 2024 by ex-Snowflake engineers Serhii Sokolenko (CEO) and Brad Heller (CTO). Their pitch: once a developer, or an AI assistant, has written the code for a data pipeline, someone still has to package it, deploy it to the right infrastructure, wire up credentials, and maintain it — the unglamorous work, Sokolenko told The New Stack back in March, that amounts to data engineering’s “last mile.”

Tower, essentially, is a managed runtime for Python pipelines — it packages the code, deploys it, and keeps it running in production. It also offers tools built on top of that runtime, like the browser-based AI agent Tower Control that allows users to describe the pipeline they want in plain language.

Control can then generate the code, deploy it as a Tower app, and run it — effectively taking the process from prompt to production without the developer having to set up the underlying runtime themselves.

MotherDuck, for its part, is a serverless data warehouse built on the open-source database DuckDB, founded in 2022 by Jordan Tigani, a former engineering lead at Google focused on BigQuery. The company has raised some $100 million since its inception.

MotherDuck’s original pitch leaned on speed and local compute: queries could run on a laptop via DuckDB, in MotherDuck’s cloud, or across both — a departure from the likes of Snowflake, Databricks and BigQuery’s cloud-first model. More recently, MotherDuck has extended that approach to AI agents, using MCP to let agents interact directly with data. And things took a more operational turn in June with the launch of Flights, a feature that exposes a general-purpose Python runtime through the same MCP server, letting agents create, run and schedule data pipelines.

And Tower, as it turns out, was the key infrastructure underpinning Flights.

‘We became their largest customer almost overnight’

Tower’s involvement with MotherDuck actually predates Flights. Tigani says MotherDuck had initially been looking for a third-party tool it could recommend to customers as an easier way to get data into its warehouses. But then advances in AI changed the nature of the problem the company thought it needed to solve.

“When AI suddenly started to be able to solve data problems, we realized we were thinking about the problem wrong.”

“When AI suddenly started to be able to solve data problems, we realized we were thinking about the problem wrong,” Tigani tells The New Stack. “Claude can solve that problem we were trying to address by writing the connectors to help people move their data, but what it can’t do is the sandboxing and scheduling.”

That left MotherDuck needing somewhere to safely execute the code those agents generated, manage credentials and run jobs on a schedule. And as luck would have it, Tower already provided those capabilities.

“It perfectly solved our problem and let us ship Flights in only a matter of weeks,” Tigani adds.

For MotherDuck, Tower had supplied the missing execution layer; for Tower, that realization translated into a significant customer relationship. “We became their largest customer almost overnight, and our teams have been shipping together ever since,” Tigani says.

Having Tower on its radar so early also gave MotherDuck a chance to test the technology — and the team behind it — before deciding whether to build something similar itself. Tigani says the calculation ultimately came down to how quickly MotherDuck could get the capabilities it wanted into customers’ hands.

“It’s always tempting to build yourself, but after trying out Tower, we realized pretty quickly that there were a bunch of problems we were going to have to solve to make our underlying infrastructure actually work well, and Tower was pretty much exactly what we needed,” he says.

Ultimately, the more central Tower became to what MotherDuck wanted to build, the stronger the case became for owning the technology outright. Once Tower was executing jobs created and scheduled inside MotherDuck, Tigani argues, customers would inevitably hold MotherDuck accountable for the security, reliability and behavior of that runtime.

“There’s a rule I’ve relearned at every infrastructure company I’ve worked at: you can rent a feature, but you can’t rent a foundation,” Tigani says. “When an agent inside MotherDuck builds a job and schedules it, the thing executing that job is our product — whatever logo is on it.”

“There’s a rule I’ve relearned at every infrastructure company I’ve worked at: you can rent a feature, but you can’t rent a foundation.”

One example of what MotherDuck now hopes to do with that technology involves bringing Flights together with Dives, an AI-generated data visualization feature it debuted in February. Tower can generate stable URLs for jobs running through Flights, effectively allowing those jobs to act as data APIs that a Dive — or another frontend — can call.

Tigani gives the example of an application displaying user recommendations. A Dive could generate the interface for viewing those recommendations, while a Flight could handle requests to create or modify them. Rather than giving the frontend broad write access to the underlying data, the Flight can constrain and validate what changes users are allowed to make.

“When you put them together, you can build rich applications,” Tigani says.

Tower’s next chapter

All of this raises an obvious question for Tower customers. Part of the startup’s pitch was that developers could use its runtime without tying themselves to a particular data platform, and now Tower itself belongs to one.

Tower co-founder and CEO Serhii Sokolenko argues that MotherDuck represents a different kind of home to that of the industry’s larger cloud data platforms. His case is that Tower can become more deeply integrated without being bent around an architecture established long before AI agents entered the picture.

“Joining a hyperscaler usually means adapting to its legacy architecture,” Sokolenko tells The New Stack. “Joining MotherDuck lets us help shape where data and AI infrastructure are heading.”

“Joining a hyperscaler usually means adapting to its legacy architecture.”

There is still a trade-off, though. Tower is giving up some of the breadth that came with being database-agnostic in exchange for building much more specifically around one platform — a bet that tighter integration will ultimately produce a better experience than supporting many systems at arm’s length.

“By focusing Tower’s Pythonic compute specifically on MotherDuck, we’re trading broad, basic connectivity for deep, native execution,” he says.

Sokolenko’s argument is that the “lock-in” question then shifts down a layer. Tower may now be tied much more closely to MotherDuck, but because MotherDuck itself is built on DuckDB, he says the underlying data remains open and portable. The aim is to bring the runtime, agents and warehouse closer together without trapping the data itself inside a proprietary system.

That also helps explain why MotherDuck’s hybrid execution model appealed to Tower. DuckDB lets work move between local and cloud compute, which Sokolenko sees as closely aligned with Tower’s own direction.

“This directly matches Tower’s vision — allowing business users and agents to move seamlessly from local data exploration to cloud production execution,” he says.

For Tower’s existing customers, the immediate future means a move toward MotherDuck. Sokolenko says Tower customers are already in discussions with MotherDuck about migration paths, while people who have used Tower previously will be invited to try MotherDuck and its broader agentic data capabilities.

Tigani confirms that MotherDuck is working to move existing Tower customers onto Flights, though he concedes that the two products aren’t entirely identical. “There are a couple of differences, and we’re working on closing the gaps to make the transition more seamless,” he says.

Tower’s technology will meanwhile be folded more deeply into MotherDuck itself. Today, MotherDuck has two separate sandboxed, on-demand runtimes: Flights, which is backed by Tower, and Ducklings, its serverless DuckDB instances. Tigani says the plan is eventually to merge the two, combining the near-instant startup of Ducklings with the more robust sandboxing provided by Tower jobs.

Surfing the agent wave

MotherDuck is hardly alone in pushing data agents beyond answering questions. Databricks’ Genie Code can generate and run code, build pipelines and debug failures inside Databricks. Snowflake, meanwhile, has been moving in a similar direction with the likes of CoCo, its AI coding agent, while newer CoCo Automations can schedule unattended agent runs inside Snowflake-managed sandboxes.

The details differ, but both point toward a broader shift in the data industry: giving AI agents the infrastructure to act on data and operate the systems around it, rather than simply query what is already there.

“AI makes it possible to build useful features that we could never fathom five to 10 years ago. The platform is the most complex part of the data estate, and so warehouse vendors are well positioned to be at the center of whatever new patterns emerge.”

Tigani has been predicting something close to this for some time. Earlier this year, he outlined a future in which data engineering increasingly becomes an agent-supervision problem, with agents handling tasks such as building and repairing pipelines and responding to changes in schemas and data quality, while humans oversee their work. He also previously likened the advance of LLMs to a wave that data companies will have to learn to ride.

“The way I like to think about it is this — data platform vendors are reacting to new opportunities to make lives better for their customers,” Tigani says. “AI makes it possible to build useful features that we could never fathom five to 10 years ago. The platform is the most complex part of the data estate, and so warehouse vendors are well positioned to be at the center of whatever new patterns emerge. The Tower acquisition gives us a platform for deploying, tracking, and scheduling data agents, which should set us up well to surf that wave.”

The post “You can rent a feature, but you can’t rent a foundation”: why MotherDuck bought the startup already powering its data pipelines appeared first on The New Stack.

Put several AI bots to work, and a mistake by one may not stay within its assigned task. For example, that error could reach another bot’s files and login credentials, or even the computer running them all. Two releases this month offered companies very different ways of containing that risk.

On August 17, Nous Research announced that its Bot Mode would ship bundled and enabled by default in Hermes Agent v0.20.3, turning agent profiles into a roster of named bots that hand off work to one another. About a week earlier, SpaceXAI launched Grok Bot with almost the same interface: a sidebar of named teammates who sign in to your tools and keep working long after you close your laptop.

The interface converged within a week, but the answer to the question every platform team has to ask did not: When one bot goes wrong, what can it reach?

Four projects have now come to their own answer, and no two of them agree.

1. Grok Bot draws the line around the user account.
2. Hermes draws it around the profile.
3. OpenClaw draws it around an optional runtime sandbox.
4. ClawFleet draws it around a container.

Taken together, the documentation shows an industry converging on the persistent coworker interface far faster than it is converging on what constitutes an identity or a security boundary for it.

Four projects, four written answers

Every one of these products now offers the same surface. You create several named agents, assign them different jobs, and have them pass work among themselves. The naming convention alone suggests separation, since a bot called Expense Manager and a bot called Talent Scout sound like they occupy different rooms in a shared office.

The documentation says otherwise, and it says something different in each case. The unit of isolation is the account in one product, the profile directory in another, an opt-in container in a third, and the deployment topology in the fourth. Those four units are not interchangeable, and an operator who assumes the roster itself is the boundary will be right in exactly one of the four cases.

Is Grok Bot confused about what it wants to be?

SpaceXAI’s launch post leads with the promise that bots have their own computer. The documentation, last updated the same day, describes a single persistent cloud computer assigned to the user account rather than to any individual bot. Browser cookies and signed-in sessions are shared across the roster, files are visible to every bot, and command-line credentials are shared. One bot can pick up work that another bot saved.

Each bot gets its own screen on that machine, which allows several of them to run browser and desktop tools in parallel. SpaceXAI is direct about what those screens are not. The documentation calls them “separate work surfaces, not separate security boundaries.” It then instructs operators to keep a credential or file off the machine entirely if another bot on the account cannot use it.

The consequences run further than credentials. Signing in for one bot makes that session available to the others because the browser is shared. Installed connectors are account-wide, and their availability is not isolated to a single bot. The shared workspace sits at /workspace and is designed to survive computer updates and recovery, so the durable state is shared across the whole roster.

None of this is an implementation accident. It is what makes handoffs between bots cheap, and cheap handoffs are the product. But an operator reading only the launch page would build a mental model that the documentation contradicts, and that gap is where the risk sits.

Hermes gives each bot its own profile

Nous took the opposite architectural position. In Hermes, a bot is a profile, and each profile has its own configuration, memory, skills, credentials, and chat history stored in its own directory on disk. Handoffs between bots run as real invocations against the named profile, rather than as a shared context blob passed around within a single process.

Nous shipped the teammate protocol as part of v0.20.3, alongside the MCP 2.x SDK migration and a set of runtime hardening changes. The company archived the standalone plugin repository once the merge was completed. Bot Mode is on by default, so a Hermes user who updates gets the roster without opting in.

Two qualifications matter before anyone reads that as containment. A separate credential store does not guarantee different credentials, since what ends up in a new profile depends on how the operator created it and what they edited afterward. And every profile still shares the host machine, its operating system user, and its filesystem permissions. What Hermes documents is workstation-level separation of agent state, a meaningfully stronger default than a shared cloud account, but not the same as isolation.

OpenClaw’s sandbox is off by default

OpenClaw documents the most complete boundary of the four. When the sandbox is enabled with the Docker backend, agent tool execution runs inside isolated containers. At the same time, the gateway remains on the host, and the scope can be selected per session, per agent, or shared across agents. Each scope gets its own workspace. Auth material lives per agent under an agent-scoped auth profiles file. Operators can configure network isolation, resource limits, and allow-or-deny tool policies on top of it.

The documented default for that sandbox mode is off. That is a defensible choice for a project most people run on a laptop, where the container overhead buys little against a single-user threat model. The underlying setup behavior deserves more attention. If sandbox prerequisites fail during setup, the script resets sandbox mode to off rather than refusing to start, so an operator who intended isolation and encountered a Docker socket issue ends up running without sandbox isolation. The documentation also warns against mounting the host Docker socket into agent sandbox containers and flags the CLI container’s shared network namespace with the gateway as a trust boundary in its own right.

ClawFleet answers the same question by moving it into the deployment topology. The project documents a wrapper that puts each OpenClaw or Hermes agent in its own Docker container with an isolated filesystem and network. It lists roughly 500 MB of memory per OpenClaw instance and 150 MB per Hermes instance. That cost is why the other three projects make the boundary optional or skip it, and naming the number makes the trade-off legible.

How to choose which bot is right for you

Each project gives operators different advice

The operational guidance diverges as sharply as the architecture. OpenClaw’s docs read like infrastructure documentation, naming specific hazards such as the Docker socket and the shared network namespace, and telling the operator what not to do. Hermes documents the profile layout and the protocol, then leaves policy to the operator. Grok Bot’s guidance is largely the warning itself, an instruction to treat the account as the boundary and to keep sensitive credentials off the shared machine entirely.

Grok Bot carries a second disclaimer worth reading alongside the first. Sensitive actions route through an approval mechanism. SpaceXAI documents the categories that trigger it, including sending messages, publishing content, purchases and transfers, deleting data, and touching production. Enforcement runs through an LLM classifier. Cursor‘s documentation for that same engine states plainly that the classifier is not a security boundary and can make mistakes. A buyer evaluating the product therefore finds the phrase twice, attached to the two mechanisms they would most reasonably assume protect them.

AI agents still lack identities of their own

Enterprises can adopt any of these four products today and get real work done, and the honest reading is that all four are engineering their boundaries in good faith against different threat models. What none of them provides is an identity for the agent. In every case, the bot borrows the operator’s credentials, whether from a shared cloud browser, a profile directory, or a container volume, and the entire security conversation boils down to how far those borrowed credentials travel.

“There is no primitive to standardize on, so each project has invented a boundary at whatever layer it already controlled, the account, the profile, the runtime, or the container.”

That is why the four answers differ so much. There is no primitive to standardize on, so each project has invented a boundary at whatever layer it already controlled: the account, the profile, the runtime, or the container. Expect to see that gap close on the identity side rather than the agent side, through scoped delegation and per-agent credentials issued by the identity provider, rather than being copied from the human. Until then, the useful move for platform teams is unglamorous and specific. Read the security page before the launch page, because for this class of product, they describe different things.

The post Grok Bot vs. Hermes: Where each draws the security boundary appeared first on The New Stack.

Real-time AI at scale is harder than it looks. Pipelines that hum along in development routinely hit problems in production. It’s always easy to blame the model for all your problems. But issues like rising latency and degrading accuracy can usually be traced back to the data pipeline. 

My colleague Tim Koopmans and I recently discussed what typically goes wrong with real-time AI at scale. After Tim shared some hard-fought lessons learned, we talked about how to avoid falling into these traps yourself – including the practices and infrastructure choices that can help you avoid them. You can watch the full video or read the key points below. 

Why AI performance fails at scale

Tim learned the following real-time AI performance lessons the hard way: through fits of frustration while building an ML-based financial trading app.

You can’t dig yourself out of tail latency

All too often, latency looks fine in testing, then a P99 spike surfaces under real concurrent load. For example, as Tim’s app approached ~740K operations per second, its P99 latency skyrocketed to 3 seconds. 

“I kept blaming the model for being slow, but it turns out the model was fine,” Tim explained. “It was just that the feature lookups were killing me.” Each inference call was doing just a handful of reads, but those reads queued up [behind writes] under load. The average latencies seemed fine, but that P99 tail latency was just unacceptable. 

“Tail latency isn’t a bug that you can fix, it’s a property of your architecture.”

Once you hit highly concurrent write throughput, you get lock contention – and that impacts the tail latencies. At this point, retries and bigger caches and connection pool tuning don’t help. As Tim put it, “Tail latency isn’t a bug that you can fix, it’s a property of your architecture. For example, if your storage engine is producing GC pauses at exactly the wrong moment, you’re going to cop a latency spike, no matter what.”

The culprit in Tim’s app was actually Postgres under pressure: “It’s not a slow database, but it was just a database being asked to do too much in this particular case,” Tim continued. 

Stale features kill accuracy

If you notice a mysterious accuracy drop that the model itself can’t explain, feature freshness might be the problem.

For Tim, this issue was particularly frustrating. User profile (wallet addresses) staleness was blowing past a five-minute SLA target by hours, vector embeddings were going stale, and offline evaluation metrics looked fine the entire time. As Tim put it, “You have this maddening situation where offline evaluation metrics look great, but as soon as you mix it in with online data, that performance is rubbish.”

“Offline evaluation metrics look great, but as soon as you mix it in with online data, that performance is rubbish.”

When the model was in production, it started making calls that didn’t track. After spending what seemed like ages debugging the model, the model itself turned out to be fine. The problem was that the model was making decisions based on old data (garbage in, garbage out, essentially). 

Vectors indexes need maintenance

No matter what vector database vendors imply, “set it and forget it” isn’t a realistic strategy for embeddings. Every re-embedding pass rots the index a little more, whether you notice it happening or not.

Tim hit this too. He was re-embedding content every time he improved the model, and the index quality rotted a bit more with every pass. At one point, he noticed that the recall rate (the share of true best-matches an approximate search actually finds) dropped to a dismal 42% – and query latency ballooned at the same time. He explained, “HNSW graphs degrade as they take on mutations. The nasty thing is you don’t really realize that until you realize your results are tainted.”

He advised others to treat a vector index like you’d treat any other database index. It needs the same care, love and attention as anything else you operate. That means:

• Monitor recall accuracy (and results returned) 
• Plan for partial builds (or batch builds)
• Know that changing your similarity function, your search parameters, or your embedding model means starting the graph over from scratch.

You gotta keep ’em separated

Another problem is resource contention – for example, training and serving fighting over the same hardware. Tim had just one machine doing double duty. With everything running on the same infrastructure, GPU, RAM, and CPU were all competing for resources. Side note: Many people don’t realize that vector search is a CPU cost, not a memory cost, since you’re traversing a graph rather than just storing vectors. 

The fix is the same thing every distributed systems person already knows: You gotta keep ‘em separated. This is just good engineering principles: separate your write path from your read path, separate training from serving if you can afford it.”

Retraining is inevitable

Recognize that retraining isn’t optional, and it isn’t free. Every model swap requires transition time.

Tim explained that there’s a dodgy window where the old model is still serving stale predictions and the new one hasn’t warmed up yet. For the database, this could mean new access patterns, cache misses, cold reads or request queues building up. When you notice that data is drifting, or user behavior is changing, the model you trained three months ago is getting worse – that’s the sign that it’s time to retrain. 

It’s going to happen eventually, so plan for it. Tim’s own approach was blue-green deployments, canaries, running the old and new model in parallel under different names, and doing the actual cutover at the application layer rather than all at once. If you’re at, say, Tripadvisor scale – with 100 million ML models – you can imagine the process will be considerably more complex.

Avoiding the doom loop with a high-performance database

These problems tend to build on each other and snowball. Latency causes staleness, staleness degrades accuracy, degraded accuracy triggers retraining, retraining causes contention, and contention makes latency worse again. It can create what Tim deemed a “doom loop.”

Here are some tips for avoiding that doom loop. 

Monitor, monitor, monitor

Be obsessive about monitoring. Watch freshness and backlog in particular because a growing backlog is what eventually drives up tail latency. Also watch index health, since that’s where recall rots. And load test beyond steady state because you can’t really predict when some weird confluence of factors will cause usage to surge. 

Isolate your workloads

This addresses two of the problems from earlier: the write storms that caused tail latency, and training and serving sharing the same infrastructure.  A database that handles concurrent writes well and isolates workloads properly can absorb both.

For example, with ScyllaDB, the write path is lock-free and multi-writer. That means every node takes writes in an active-active fashion, and no row gets locked in the process. As a result, a burst of concurrent writes doesn’t back up into a queue the way it would on a database built around single-writer assumptions. 

On top of that, a practice we call “workload prioritization” controls how workloads compete for system resources. This ensures latency-sensitive queries are fast, even with other heavy workloads running on the same cluster. That way, a retraining job or a backfill won’t steal resources from whatever’s serving live inference. 

Separate vector indexing

To address the vector index problem, keep the index separate instead of bolting it onto the same process as the core database. For example, ScyllaDB Vector Search writes land in the core database first, and the index gets built out of that data asynchronously, as its own service. 

If the index can’t keep up with the write rate (whether from a re-embedding pass or a full rebuild after a similarity function change), it falls behind – but it never misses a write and the core database is not impacted. Even if the vector store goes down, the embeddings still persist in the core database. And because ANN queries are CPU-heavy, keeping them on a separate service means they’re not fighting the core database for the same CPU cycles the writes require.

Under billion vector benchmarks, that separated architecture held P99 latency under 10 milliseconds at a concurrency of 300. It handled ~150,000 ANN queries a second with a moderate recall target. Realize that higher recall will bring a latency and throughput tradeoff, and always test this in advance to assess how your own mileage varies. 

Absorb the shock

This one comes down to whether your infrastructure can absorb a sudden change in write pressure or traffic shape without a scramble. The storage engine’s architecture matters a lot here.

For example, ScyllaDB is built on an LSM-tree, which tolerates that kind of write pressure instead of degrading under it. Elastic scaling, with what we call “tablets,” can scale a cluster by something like 10x within minutes instead of hours. That means that if a model rollout changes your access patterns overnight, or you need to absorb a backfill before a big retrain, you don’t end up waiting on a multi-hour resharding job.

The more things change…

So much about AI is genuinely novel, but the infrastructure problems described above generally are not.

“Real-time AI is really a distributed systems problem in a costume.”

Tim mentioned that a feature store was probably the primitive use case: the same high write throughput, low latency work, years before anyone called it AI. Real-time AI is really a distributed systems problem in a costume. Once you understand that, you can design for it so you don’t get blindsided by these not-so-new challenges. 

The post Why real-time AI at scale is so hard appeared first on The New Stack.

On July 16, the team at Hugging Face noticed something weird moving through their production systems: An intruder that was cloning datasets, harvesting credentials, and hopping between internal clusters faster than a human could type. Five days into the investigation, the authorities were notified, and several days later, OpenAI traced the intruder back to one of its models.

“Two frontier models escaped their test environments this summer; only one of them had a funny reason.”

That prompted Anthropic to look through its own test logs, and it found that similar events had occurred three times. One Claude model probed nine thousand hosts in a single run without setting off any alarms. Another slipped a poisoned package into the public Python registry, compromising fifteen machines before it was caught. The third incident dated back to April and went undetected for three months until a rival’s error prompted the audit. Nobody had noticed. Two of the three companies involved didn’t know anything was amiss until they got a call.

Containment by instruction

A sandbox, in the context of AI testing, means a model gets a task, a set of tools, and an instruction: no access to the outside world.

OpenAI’s model found an exploit in a tool it was allowed to use, and leveraged it to access the open Internet. Anthropic’s models didn’t even need that. The models were told a path didn’t exist, so when their tasks led them down a path that wasn’t supposed to be there, they treated it as part of the exercise.

“In both instances, the only constraint was an instruction. There was no mechanism outside the model itself to ensure those instructions wouldn’t be treated as optional.”

In both instances, the only constraint was an instruction. There was no mechanism outside the model itself to ensure those instructions wouldn’t be treated as optional.

So how did they escape?

OpenAI’s agent was being evaluated on a cybersecurity benchmark called CyberGym. During the test, it chained a code-execution flaw in a dataset loader to a second flaw in the platform dataset configuration file handler. It used the combined exploit to run commands on a live processing worker. From there it grabbed a write-scoped access token and cloned a private dataset as a staging area, which became a dead drop for its own tools. By the time anyone noticed, the intrusion had been running for three days, touched several internal clusters, and compromised accounts at four other companies, according to Hugging Face’s own review.

What it wanted, in the end, was comical: the answer key to the test it was taking.

Anthropic’s three incidents were nowhere near as convoluted. Its models were being run through security exercises against third-party evaluation partners, and the only thing keeping them from accessing the Internet was an instruction not to. One incident involved an SQL injection targeting a resource the model wasn’t supposed to be able to access. Another, during an exercise where the usual safeguards were deliberately switched off, ended with a Claude model publishing a malicious package to the Python Package Index, and fifteen machines picking it up before it was pulled down. No zero-day exploit. No dramatic escape. Just locked doors that turned out to be mirages.

The structural weakness

Strip away the specifics, and both incidents share the same root cause: there was nothing outside the model’s own judgment that could enforce the limitations placed on it. A credential that shouldn’t have been able to access the open Internet did so anyway. An instruction that should’ve stopped a plan cold got treated as just another input to reason around.

“If a fence doesn’t hold during testing, it won’t hold in production.”

It feels like we’ve spent the last year reliving the same problem in new ways. For example, in deleted databases, in poisoned supply chains, in every credential an agent inherited because provisioning it correctly took one extra step nobody got around to. The only delta is where the boundary was supposed to be: around the dev experiment, meant to contain dangerous capabilities, or standing between it and every production system it could reach. If a fence doesn’t hold during testing, it won’t hold in production.

How do we break out of the loop?

Unfortunately, the answer isn’t self-improvement. A more capable model wouldn’t have caught any of this. The only way to fix it is with Infrastructure that verifies where a model actually is, rather than trusting its reports.

Egress controls that are closed by default. A test environment that can reach the open Internet is probably not a great idea. Outbound traffic should be locked to an explicit allowlist. Never assume a model will simply decline to look for a way out. “No internet access” needs to be a network-enforced fact, not just an instruction.

Narrowly scoped credentials with short lifespans. Both incidents ran, in part, on tokens that extended beyond what the tasks required. A credential built for one evaluation run, scoped to nothing beyond it, and expiring on a timer, can’t become the write-scoped key that turns into a lateral-movement problem across multiple clusters. Yes, it will be a chore; so gird your patience by imagining having to explain to company leadership why taking steps to avert a catastrophe was too tedious to bother with.

Policy gates that sit outside the agent’s reasoning. An agent that can decide for itself whether an action is in scope is grading its own test. A plan should have clear rules written in code, evaluated by something other than the agent, before anything executes. Open Policy Agent has become a de facto default here for exactly this reason. It doesn’t care how confident the plan sounds; nothing runs without proper authorization from an external source.

An audit trail that is entirely independent of the agent. Both incidents were discovered by rereading logs after the fact, not by asking the models what happened. That’s how oversight should be done. As the proverb goes: Quis custodiet ipsos custodes? Who watches the watchers? A system of record that captures what actually occurred, regardless of the agent’s own account of itself, is the only version of events worth trusting.

This is a sign

The industry spent a decade learning that the CI/CD pipeline is an attack vector that requires real fortification, not just bolted-on convenience. Test environments for frontier models are following the same arc at a faster pace. The next time one of these agents escapes containment, it’ll probably be one built around finding unlocked doors, which will make it substantially more dangerous than a coding agent that deletes a few databases.

Test rigs must be treated as if they hold something real, because, as far as the credentials are concerned, they do. A sign on a door is never going to be enough to keep everyone out; there has to be a lock whose robustness correlates to the value of what it guards.

“A sign on a door is never going to be enough to keep everyone out; there has to be a lock whose robustness correlates to the value of what it guards.”

Whether by a state-sponsored crew probing a water management system in the middle of the night, or a company’s own model trying to shave a few points off a benchmark, boundaries will always be tested. Two labs found out this summer, and the story needs to be taken seriously. The vulnerabilities are real, the transparency from the labs is welcome, and the containment failures are a cause for concern.

Catching a model that tried the handle is the easy part; both labs proved that. The more challenging, and therefore critical, part is making sure the next containment environment actually has doors that are firmly locked.

The post Securing sandboxes: What happens when AI agents escape containment? appeared first on The New Stack.

The artificial intelligence landscape has reached a pivotal inflection point. Over the past several years, the paradigm has shifted from passive, conversational Large Language Models (LLMs) to autonomous AI agents, digital software entities capable of reasoning, invoking tools, executing multi-step workflows, and making real-time decisions across enterprise systems without constant human intervention.

As organizations accelerate the production deployment of autonomous agents, modern security frameworks must evolve to keep pace. Traditional Identity and Access Management (IAM) systems were primarily designed around two distinct operational models:

• Human users: Authenticated via Multi-Factor Authentication (MFA), Single Sign-On (SSO), and interactive sessions.
• Service accounts and workloads: Authenticated via static API keys, fixed service tokens, or IP whitelisting.

Autonomous AI agents blur the line between these two models. An agent acts with the non-deterministic reasoning and delegated agency of a human, but operates at the scale, parallel velocity, and automation speed of a machine service.

To safely harness the power of autonomous workflows, enterprise security architecture must move toward continuous, agent-aware Zero Trust governance. Below are six foundational identity capabilities that organizations should adopt to secure AI agents in production environments effectively.

“Autonomous AI agents blur the line between these two models. An agent acts with the non-deterministic reasoning and delegated agency of a human, but operates at the scale, parallel velocity, and automation speed of a machine service.”

“When it comes to agentic AI identity, most organizations are woefully unprepared for inherent security risks and operational challenges of managing those identities.” – Ken Buckler, Research Director, EMA – Agentic AI Identities – Is Your Organization Prepared?

1. Verifiable agent identities & “Know Your Agent” (KYA)

Autonomous entities require verifiable digital identity frameworks that establish clear, cryptographically bound accountability for every machine action.

• Cryptographic attestation: Every agent instance should possess a unique, cryptographically signed identity bound to its underlying model version, execution environment, and deployment origin.
• Delegation chains: When a human user delegates a task to an agent (or when a primary agent spawns sub-agents), the identity system must construct an immutable, traceable chain of delegation. This ensures the infrastructure can continuously verify who authorized the initial action and what specific scope was granted.

2. Ephemeral credentials & just-in-time (JIT) tokenization

Static API keys and persistent service tokens represent a significant surface area of exposure when integrated into dynamic agentic workflows. Replacing long-lived credentials with short-lived tokens dramatically reduces the potential window of risk.

• Just-in-time (JIT) minting: AI agents should operate with ephemeral credentials generated on demand, strictly limited to the API calls required for a single operational step, and configured to expire within seconds or minutes.
• Bound OAuth flows & PKCE: Enforcing Proof Key for Code Exchange (PKCE) and strict token-binding protocols ensures that credentials cannot be reused or replayed outside of their intended runtime context.

“Replacing long-lived credentials with short-lived tokens dramatically reduces the potential window of risk.”

3. Relationship-based access control (ReBAC) & intent binding

Coarse-grained permissions, such as those in traditional Role-Based Access Control (RBAC), are often too broad for non-deterministic tool usage. Access governance should be based on fine-grained relationship models and task intent.

• Intent-bound authorization: Authorization systems should evaluate not only whether an agent has general permission to access a resource, but whether that request directly aligns with the explicitly authorized sub-task.
• Fine-grained contextual policies: Implementing relationship-based access control (ReBAC) or Attribute-Based Access Control (ABAC) allows teams to define precise conditions (e.g., “Agent X may read Document Y only if human user Z is the document owner and the active workflow is ‘Data Summarization'”).

4. Machine-speed containment & automated anomaly detection

Because AI agents operate at speeds far exceeding those of manual monitoring, security containment mechanisms must be automated, agent-aware, and built into the control plane.

• Behavioral rate & scope limits: Security controls should establish baselines for expected agent behavior to detect anomalies, such as rapid parallel tool invocations, repetitive execution loops, or unusual queries to non-standard endpoints.
• Automated circuit breakers: If an agent’s execution pattern or request velocity exceeds defined behavioral bounds, identity proxies can automatically revoke ephemeral tokens and safely isolate the workload in real time.

5. In-the-loop runtime enforcement & human approvals

Security governance cannot rely solely on static pre-authorization; policies must be evaluated continuously at runtime before individual actions execute.

• Action-level policy interception: Enforce real-time policy checks at the agent harness layer—evaluating shell commands, database queries, file operations, and outbound API calls against governance rules before execution.
• Configurable approval workflows: Establish flexible escalation paths that permit low-risk read operations automatically while requiring explicit human-in-the-loop validation for high-impact actions, such as code deployments or financial transactions.

6. Web-scale identity architecture built for machine workloads

Autonomous workflows generate significant operational volume. Identity systems must be architected to handle machine-scale throughput without performance degradation or store bloat.

• Machine-speed throughput: Multi-step workflows and parallel worker agents demand identity control planes that can handle high-volume token validation and policy evaluation with minimal latency.
• Lifecycle governance for sub-agents: Dynamically spawned sub-agents require rapid provisioning and immediate teardown upon task completion, thereby preventing the accumulation of orphaned credentials and ensuring clean session termination.
• Inline cryptographic safeguards: Prioritizing inline policy enforcement over post-mortem log reviews allows organizations to intercept unauthorized state changes before they occur, maintaining operational integrity across multi-cloud environments.

Conclusion: securing the future of enterprise automation

As AI models evolve from passive assistance tools to active operational participants, identity becomes the primary boundary for enterprise governance. By bridging the machine identity gap with verifiable agent identities, short-lived JIT credentials, fine-grained relationship authorization, and automated runtime enforcement, security leaders can confidently deploy autonomous AI agents to drive productivity while maintaining complete operational control.

The post Six identity capabilities for securing autonomous AI agents appeared first on The New Stack.

Anthropic launched a new Browser Use tool that gives Claude a structured view of a web page in addition to what is visually rendered. Announced Thursday, the tool uses the page’s accessibility tree to help Claude find and interact with specific elements directly rather than having to work out where they are on the screen.

Browser Use is part of a broader Anthropic release that also brings Computer Use, the Skills API and Files API into general availability. Developers can access the browser tool through the Claude API using browser_toolset_20260801.

Browser Use is part of a broader Anthropic release that also brings Computer Use, the Skills API and Files API into general availability.

The change gives Claude a more direct way to interact with a web page. Instead of working out a button’s position from a viewport image and targeting coordinates such as x: 640, y: 320, Claude can receive a reference such as ref_3 tied to that element and use it when it wants to act.

Page references replace coordinates

Computer Use can operate across an entire desktop by looking at screenshots and sending mouse coordinates and keyboard commands. Browser Use works within the browser itself, where it can use page structure that would be difficult to recover reliably from pixels alone.

When Claude calls read_page, the developer’s executor returns a text representation of the accessibility tree, in which elements such as links, buttons, and text boxes can be tagged with references. If Claude later wants to click a button represented by ref_3, it can send that reference along with the requested operation rather than trying to calculate where the button is on the screen.

That said, if the tab navigates to a new page or the page changes enough, a reference that pointed to a button a moment ago may no longer work. The API will not catch that on its own, so the executor has to recognize when the reference no longer matches the underlying element, reject the action and have Claude read the page again before continuing.

Batching cuts model calls

Playwright, for example, can represent a page as an ARIA snapshot and locate elements by role rather than coordinates. At the same time, Microsoft’s Playwright MCP server already exposes structured accessibility snapshots with references a model can use to identify elements. The concepts line up closely with Browser Use, but the protocols do not: Playwright MCP speaks MCP, while Anthropic’s tool uses its own client-toolset protocol, so developers would still need an adapter that translates Claude’s requests into Playwright actions and returns the results in the format Claude expects.

Puppeteer offers many of the same building blocks, exposing the browser’s accessibility tree via Accessibility.snapshot() and providing APIs for controlling Chrome and Firefox. A developer could use those APIs for navigation or page reads, then maintain Anthropic’s reference mappings on top.

A developer could use those APIs for navigation or page reads, then maintain Anthropic’s reference mappings on top.

Slightly confusing, an unrelated open-source project also called Browser Use runs AI browser agents against Chromium through the Chrome DevTools Protocol. Despite the shared name, it has no connection to Anthropic’s tool and comes with its own agent loop and browser abstractions, so connecting the two would still require integration work.

Several browser actions can happen in one turn

Anthropic is also reducing the back-and-forth between Claude and the browser by allowing multiple actions to be requested in a single model turn. Now actions can arrive together as several tool_use blocks. The application executes them in order and sends the results back together, avoiding another model call between every click and keystroke. Anthropic says that can lower latency and costs, particularly as workflows scale from a handful of interactions to dozens or hundreds.

That matters more as browser tasks get longer. Cheaper models alone will not solve the token cost problem in agentic workflows, so cutting unnecessary model calls is another way to reduce costs.

If Claude has to return to the model after every click or keystroke, a long browser task can quickly rack up model calls. Batching cuts out some of that back-and-forth by letting Claude request several actions at once, but the browser still has to carry them out in order because each one depends on what happened before it. If Claude asks to click a button, fill in a field, and submit a form, for example, the executor cannot simply move on to the next step if that first click fails, because everything that follows is now based on a page state Claude never reached.

Batching cuts out some of that back-and-forth by letting Claude request several actions at once, but the browser still has to carry them out in order because each one depends on what happened before it.

Developers host the browser

Browser Use is currently limited to the Claude API and is not available inside Claude Managed Agents. Adding it to a Messages API request exposes 27 browser operations by default. Claude can decide which of those operations it wants to use, but Anthropic does not execute them. The application has to translate each request into an action inside its own browser environment, preserve the session between turns and return enough information for Claude to understand what happened.

Loading all of those operations has a token cost. Anthropic’s pricing documentation says the default Browser Use toolset adds roughly 6,600 input tokens to a request, before counting screenshots, accessibility trees and other results sent back to Claude. Developers can turn off operations they do not need to reduce that overhead.

It also creates a different hosting split from some of the other tools Anthropic announced Thursday. Skills uploaded through the Skills API can run inside Anthropic’s code execution sandbox, while the Files API stores documents that can be reused by ID. Browser sessions, along with their downloads and uploaded files, stay in the developer’s environment.

Approval gates need rethinking

Claude can still encounter a prompt injection in web content or be redirected to an unexpected location, which is why Anthropic recommends running the browser in an isolated container or virtual machine with minimal access. JavaScript and file uploads should remain disabled unless needed, since code generated by Claude runs with the page’s privileges and can reach data or make requests available to that page.

Batching makes approval a little trickier because several actions can arrive at once, and a routine click at the beginning of a sequence could eventually lead to something that requires the user’s permission. That means the executor has to check actions as they happen and stop for approval when needed.

The post Anthropic’s new browser tool doesn’t actually run a browser appeared first on The New Stack.

I’m Matt Burns, Chief Content Officer at Insight Media Group. Each week, I round up the most important AI developments, explaining what they mean for people and organizations putting this technology to work. The thesis is simple: workers who learn to use AI will define the next era of their industries, and this newsletter is here to help you be one of them.

Model triage is becoming one of the most important skills for the AI-native developer. I’ve argued all summer that the people getting the most out of frontier models are the ones disciplined enough not to run the best model by default. On Wednesday, Stripe and Ramp validated that idea 70 minutes apart: Stripe bought OpenRouter, and Ramp released its internal router.

Bloomberg puts the OpenRouter price tag above $7 billion, and Axios says it’s more than $8 billion in cash and stock. Stripe has not released the terms, so the details remain fuzzy.

While the acquisition made headlines, the architecture is the story. For the last couple of years, picking a model was something written into an application, a string in a config file, and swapping models took some work. A router changes that workflow. Stripe bought the layer, and Ramp built it. Both are betting their existing relationships give them a unique wedge to own this critical layer in the new AI stack.

Picking a model is becoming a runtime decision

OpenRouter is an endpoint serving more than 400 models from over 80 providers, processing more than 10 trillion tokens a day. Andrej Karpathy calls it the transfer switch for AI. Ramp’s Router does the same job on a smaller catalog and claims roughly 40% lower cost for the same output.

Both are betting the model name in a codebase is a liability. They’re mostly right. Back in June, I pointed to Mitchell Hashimoto, who found a standard coding task that cost about $1.50 on GPT-5.5 and roughly $9 on Claude Fable, with both producing equally acceptable results. A router automates that triage, making decisions on every request rather than only on those a developer explicitly configures.

This is becoming a large problem and a large opportunity. Our own Amanda Caswell reported this week that Anthropic’s /claude-api skill was burning about 200,000 tokens before answering a single question, and that loading its reference docs on demand instead of up front cut that to roughly 25,000. Hafiz Hassan wrote for us last week about why AI pipelines cost 10x more than the demo, and every culprit on his list is an engineering decision: system prompts resent every turn, whole conversation histories appended, oversized RAG chunks, raw JSON dumped into context. It’s a great practical guide, and none of the items Hassan identifies are procurement problems.

The token bill is generated by your code, which is why the tools to control it are arriving there as well.

Stripe and Ramp want the same layer for opposite reasons

Stripe is attacking the problem from the bottom up, through developers. The company’s investor letter, leaked Wednesday by Eric Newcomer, makes the argument directly: “Up until now, every developer has needed a straightforward and reliable way to manage their revenue pipeline, and serving this need gave rise to Stripe. Going forward, however, every developer will also need a straightforward and reliable way to manage their intelligence pipeline.”

Stripe wants to control AI spending through the long tail of developers. Ramp wants to control it through its existing relationship with finance.

Stripe has built this product before. Its payments business hides dozens of local payment methods behind a single API, routing each transaction to the payment method most likely to convert. The AI version is the same idea applied to models instead of payment networks.

Ramp is attacking it from the top down, through finance. The company bought the router.com domain and says its customers already buy quadrillions of tokens a month through Ramp. Founder Veeral Patel’s launch post pitches the service simply: “Monitor and control your AI bill across every provider.” Adam Wazzan sums up Ramp’s strategy better than I can: “when a CFO ships a product for CTOs.”

Stripe wants to control AI spending through the long tail of developers. Ramp wants to control it through its existing relationship with finance. Both are chasing what is rapidly becoming one of the largest line items in corporate technology budgets: tokens.

On X, Kabir Goel pushes back on Stripe’s framing. Routing tokens is a way to spend less, while Stripe’s other products are designed to help businesses make more. 

“Stripe is just not where teams go to understand how much they’re spending,” he writes. “That’s pretty squarely Ramp territory.” 

He has a point about where teams look today. Whether that’s still true three years from now is exactly what Stripe just spent billions betting against.

The router worth pointing at is the one with no model to sell

Whichever router you point at decides which model writes your code, and not every router is disinterested. Our own Paul Sawers flagged the problem in July when he covered the first wave of Cursor’s, Ramp’s, and Meta’s routers. Cursor backs Grok and Composer. Meta is building Muse Spark. Both have reasons to send work to their own models, and Paul quoted developer Elvis Saravia asking whether routing logic ought to be open source rather than a vendor’s private judgment call.

Stripe and Ramp do not sell models. OpenRouter CEO Alex Atallah says as much in Stripe’s own announcement: Developers “need a neutral layer to orchestrate and manage them all.” Investor Gavin Baker frames the opportunity the same way, arguing that Stripe can become the neutral infrastructure layer for AI, just as it became the neutral infrastructure layer for payments. 

Neutrality isn’t free, though. Stripe takes a percentage of token spend, and Ramp wants your spending relationship, so “free through 2026” is a customer acquisition strategy with an expiration date.

The obvious objection is that vendor motives are the wrong thing to worry about, and routing quality is what really matters. That’s fair, and Towards Data Science published one of the best practical examples I’ve read. Pratik Rupareliya describes a routing layer that cut a support agent’s inference bill by 40% but also broke the product. A classifier sent “simple” queries to a cheaper model, but some of those “simple” queries were actually fraud investigations. 

The cheaper model answered them confidently and incorrectly. Customers stopped using the agent, churn rose above baseline in month four, and retention costs were four to five times higher than the savings. It took three months to surface and another month to identify the cause. His fix was per-tier quality monitoring combined with an uncertainty-routed cascade, which ultimately settled at 35% savings without sacrificing quality. It’s an excellent article to read before diving into model routing.

So instrument the routing. Log what the router picks on every request and break out your quality metrics by the model that served them. Ramp Router reportedly records the model, provider, tier, tokens, latency, cost, and fallback attempts for every call. Stripe OpenRouter rankings have been a public version of that telemetry for years. You can get similar visibility with either approach.

Right now, the model is becoming an implementation detail. The competition is shifting to the layer that decides which model gets the job. Stripe and Ramp are betting that developers won’t care what sits behind the endpoint, so long as the bill is lower and the results are good enough. 

The post Forget the model wars, Stripe and Ramp just started the router wars appeared first on The New Stack.

On Tuesday, Warp introduced Warp Factories, open infrastructure for building cloud software factories, agentic systems that automate work across the software development lifecycle, which have been popping up in different forms from companies like Augment Code and Chainguard. 

Warp, an agent development platform, calls Warp Factories “the building blocks” for developers to create their own scalable factories. It’s pitching the infrastructure as the solution for two problems founder and CEO Zach Lloyd says are frequent engineering complaints: 1) measuring and improving coding agent ROI; 2) governance and control.

The aim is to tackle both problems by making sure “the annoying bits [are] taken care of” so developers can focus purely on optimizing factories for specific products. 

As Lloyd writes in a blog post, he “predicts software factories will be as ubiquitous as CI/CD in the next few years.” Experts tell The New Stack they see software factories gaining traction, but they’re more cautious about the timeline.

“I think the software factory is inevitable,” Lee Faus, founder and CEO, Atomic Software and former global field CTO, GitLab, tells The New Stack. “But before software factories become as foundational as CI/CD, the industry needs to solve a deeper infrastructure problem.”

Specifically, he calls out the importance of tracing agent work: “We’re spending a lot of time talking about how to build the software factory,” Faus continues. “I think we’re going to spend much more time asking what becomes the system of record for the factory.”

Build the factory without building all the infrastructure 

Lloyd acknowledges that many organizations already have engineering teams at work building cloud software factories — but he argues that’s too big to be an inside job. 

Warp Factories, thus, emerges as the infrastructure on which developers can build their own factories, providing the core components to speed development without making organizations sacrifice flexibility, programmability, customization, or ownership.

When asked about Lloyd’s take on building infrastructure, Erik Gfesser, long-time engineer, tells The New Stack he agrees it doesn’t make sense for most organizations to tackle it in house.

As Lloyd writes, Warp’s new infrastructure is “built to increase coding agent ROI over time” with evals and benchmarks to measure effectiveness and built-in self-improvement and memory. Developers get queryable metrics on agent throughput, cost, quality, and ROI, visible via the Factory control room, API, and Factory MCP. Scorers evaluate how work items move through the factory with an eye on things like token spend, code quality, and whether or not the work introduced defects. 

From there, those scores power self-improvement loops and benchmarks. “Observer” agents score select agent runs and then search for ways to make improvements by adjusting variables like the harness, model, or context before making PRs to improve underlying factory functionality. Benchmarks, meanwhile, let developers score tasks across different models and harness configurations to compare performance.

Governance gets easier, but there’s more to solve

Per Warp, the infrastructure includes features to address governance and control, alongside factory definitions as version-controlled code, definitions for distinct agents, plus skills, MCPs, and permissions. 

Looking more broadly, Faus tells The New Stack software factory governance will require more than just controlling how agents operate, though:

“A software factory without a record of change risks becoming a very efficient way to manufacture code that nobody can fully explain.”

“Shared infrastructure can make permissions, model access, tool use, MCP connections, policies, cost, and execution environments easier to manage centrally. That’s valuable,” he says. “But governance isn’t just being able to control what an agent is allowed to do. It is being able to prove what it actually did.”

As software factories help speed up code generation, he says the harder problem becomes understanding the scores of interconnected decisions both humans and agents make across the development cycle. 

For example, if one agent triages an issue, another researches it, a third implements it, and still others review and verify it, how can an engineer reconstruct why that change was made six months later? “That record has to remain connected to the change itself,” says Faus. “[Otherwise,] a software factory without a record of change risks becoming a very efficient way to manufacture code that nobody can fully explain.”

Software factories are probably the future, but it will be a slow roll-out

Though Warp’s founder is gung-ho about the rapid rise of software factories, other experts are less certain. Like Faus, Gfesser expects software factory adoption to take time: 

“My expectation is that software factory adoption will likely be fragmented across multiple vendors similarly to the early stages of CI/CD.”

“As an early adopter of CI/CD myself, I know that CI/CD didn’t catch on the way it did until quality open source products were made available for widespread usage.”

He points out that while the Warp client is open source, the server, the Warp Drive backend, and OZ (Warp’s agent orchestration layer) are proprietary. Also worth noting: OpenAI is named as the founding sponsor of Warp’s open source repository. 

“As such, my expectation is that software factory adoption will likely be fragmented across multiple vendors similarly to the early stages of CI/CD,” he says. 

The post Warp wants to make it easier to build your software factory appeared first on The New Stack.

After weeks of speculation, fintech giant Stripe has confirmed that it’s tabled a bid for AI model gateway platform OpenRouter, a deal designed to help businesses optimize how they route and spend AI tokens.

While terms of the deal have not been disclosed, independent reports peg the acquisition price at a cool $8 billion, making it Stripe’s largest known acquisition to date.

To a casual observer, the deal marks a somewhat odd combination: why would a payments processor want to own technology that decides which AI model answers a given prompt? Well, it all ultimately comes down to “tokenomics” — the emerging discipline of managing the cost, allocation and consumption of AI tokens.

On top of that, OpenRouter has previously said that people should think of it as “like Stripe for LLMs,” owing to the fact that it makes the fragmented AI model market accessible through a single developer-friendly API, much as Stripe did for payments. And that synergy will now culminate in the two companies becoming one.

Token gesture: ‘making good use of scarce compute resources’

Stripe became a $159 billion juggernaut as the developer plumbing behind online payments — the infrastructure that lets internet businesses accept money, run subscriptions, and get paid globally. While its core pitch has always been about making it easy for businesses to accept money, AI has become one of the biggest costs those same businesses have to manage, and managing both sides of that ledger is part of Stripe’s job.

Stripe has been building out AI billing infrastructure long before the OpenRouter deal, previewing LLM token billing and an LLM proxy for routing and metering model calls in 2025. With OpenRouter under its wing, Stripe gains a much more sophisticated routing layer that can choose between hundreds of models and providers based on cost, speed and performance.

In its announcement on Wednesday, Stripe co-founder and CEO Patrick Collison says that “tokens are the central currency for companies building with AI,” adding that the acquisition is ultimately all about the economics of AI.

“Tokens are the central currency for companies building with AI, and it’s clear that the real-world economic potential will depend on making good use of scarce compute resources.”

“The real-world economic potential will depend on making good use of scarce compute resources,” he notes. “Stripe is building the economic infrastructure for AI, and together with OpenRouter we’ll help businesses maximize profitability by routing their requests intelligently and spending their tokens efficiently.”

Open sesame

OpenRouter itself is a relative newcomer to the technology world. Started in early 2023, and co-founded by former OpenSea CTO Alex Atallah, the platform acts as a single front door to the increasingly crowded AI model market. Developers can use one API to access and switch between hundreds of models from dozens of providers, without having to rewrite their applications every time they change models.

Underneath that common interface, OpenRouter handles much of the messy stuff: routing requests between providers, automatically falling back when one goes down, and optimizing for things such as price, latency and model quality. It generally passes through providers’ inference prices without a markup, instead making money through a 5.5% fee on credits purchased through the platform.

That proposition has helped it gain sizeable traction. OpenRouter now says it serves more than 10 million developers and companies across more than 400 models, processing over 10 trillion tokens per day.

The company is also fresh off the back of a $113 million funding round, led by Alphabet’s growth fund, with participation from a slew of high-profile backers including the venture arms of Nvidia, Databricks, Snowflake, MongoDB, and ServiceNow — a strategic bet by some of the biggest names in AI and enterprise software.

“AI has become the single largest driver of economic growth in the US, and inference is quickly becoming the largest line item for every company.”

In its own announcement post, penned by founders Alex Atallah, Chris Clark, and Louis Vichy, OpenRouter positions the deal against a bigger shift in where businesses are spending their money: away from simply building AI products and toward the ongoing cost of running them.

“AI has become the single largest driver of economic growth in the US, and inference is quickly becoming the largest line item for every company,” they write.

As for why Stripe, OpenRouter points to a shared developer-first heritage. Stripe’s APIs became something of a benchmark for developer software, while its payments infrastructure gives it experience handling huge volumes of transactions, fraud and abuse — problems OpenRouter increasingly faces as AI usage grows.

The company also suggests that remaining independent was a perfectly viable option, and that very few potential buyers could have persuaded it otherwise.

“There are few companies on earth we would have considered selling to; our mission, our neutrality, and our lead in the market make the story for independence strong,” they write. “We would only join a company if we thought we could do more together, faster, without compromising any of them.”

For customers, OpenRouter’s message is essentially business as usual. Stripe will own the company once the deal closes, but OpenRouter says its brand, product, roadmap and model-neutral approach will remain as is.

“There are few companies on earth we would have considered selling to; our mission, our neutrality, and our lead in the market make the story for independence strong.”

That continuity will likely matter, too, because OpenRouter is far from alone in trying to solve the problem. A slew of companies this year have been investing in their own routing layers, as AI inference costs climb and no single model stays the best or cheapest option for all that long.

The model-routing rush

Cursor, the AI coding tool now owned by Elon Musk’s SpaceX, shipped its own Router back in July, claiming savings of 30-50% compared with routing every request through its priciest model.

Ramp, the $44 billion spend-management company, also debuted its very own model router in July, a product that launched on Wednesday at its own dedicated Router.com domain — the same day Stripe announced its deal with OpenRouter. The company says three years of tuning its own AI spend internally cut its bill by 30% — the pitch to new users now promises a bigger number, an average 40% cut.

Meta, for its part, is also reportedly building a model router of its own. The Information reported in July that it’s planning Switchboard — a project out of an internal incubator called AAI Labs, that scores each request for difficulty and routes the easy ones to cheaper models. It’ll stay internal at first, aimed at cutting Meta’s own AI agent bill, but could eventually ship as an external product too.

All this activity speaks to a much broader reckoning over the cost of AI. In June, the Linux Foundation announced the Tokenomics Foundation, backed by the likes of Google, Microsoft, IBM and Salesforce, to develop common standards and benchmarks around how AI tokens are produced, consumed and monetized.

Model routers are one practical answer to the broader underlying problem: spend less by being smarter about which model gets each job. And with OpenRouter now set to become part of Stripe, those economics are moving directly into the payments giant’s wheelhouse.

The post OpenRouter called itself the “Stripe for LLMs” — now Stripe’s swooped in to buy it appeared first on The New Stack.

Every engineering team deploying AI agents eventually discovers an uncomfortable truth: the model isn’t the biggest expense. The hidden cost is everything around it: repeated retrievals, duplicate prompts, unnecessary tool calls, oversized context windows, multiple agents reasoning over the same information. Individually, these architectural decisions seem harmless. At production scale, they become a severe tax on latency, infrastructure, and cloud spend.

A proof-of-concept agent that answers 50 questions a day can tolerate inefficiencies. An enterprise platform coordinating thousands of requests per minute cannot.

This article explores practical techniques for engineering token-efficient AI systems without sacrificing output quality. Rather than focusing solely on prompt compression, we will optimize the entire workflow from routing and retrieval to caching and model selection.

Why token optimization is a systems problem

Most discussions around token optimization begin and end with prompt engineering. In practice, architecture drives token consumption.

Consider a typical multi-agent workflow:

User 
  ↓
Intent Agent
  ↓
Retriever
  ↓
Research Agent
  ↓
Planning Agent
  ↓
Writer Agent
  ↓
Reviewer Agent
  ↓
Final Response

At each stage, the system might retrieve the same documents, repeat identical instructions, call the same model, and resend the entire conversation history. By the time a response reaches the user, the architecture has processed tens of thousands of unnecessary tokens.

“Improving efficiency requires redesigning the workflow, not just shortening the prompts.”

Improving efficiency requires redesigning the workflow, not just shortening the prompts.

Architecture overview

A production-ready, token-efficient architecture introduces optimization before every expensive model invocation.

User Request
       │ 
       ▼
Intent Router 
       │ 
       ▼
Semantic Cache ───────► Cached Response
       │ 
       ▼
Context Budget Manager
       │ 
       ▼
Adaptive Retriever
       │ 
       ▼
Model Router
       │ 
       ▼
LLM
       │ 
       ▼
Validated Response

“The large language model is no longer the first component. It is the final, most expensive operation.”

Notice the critical shift: the large language model is no longer the first component. It is the final, most expensive operation.

Step 1: Install modern dependencies

Use the latest package structure to avoid deprecated imports and align with the current LangChain ecosystem.

Bash
pip install \
   langchain \
   langchain-core \
   langchain-openai \
   langchain-community \
   fastapi \
   faiss-cpu \
   tiktoken \
   rank-bm25 \
   pydantic \
   python-dotenv

Step 2: Configure the model

Production systems must configure retries, timeouts, and credentials through the environment.

Python
import os
from langchain_openai import ChatOpenAI 

api_key = os.getenv("OPENAI_API_KEY") 
if not api_key: 
    raise ValueError("OPENAI_API_KEY must be configured.")

llm = ChatOpenAI( 
    model="gpt-4o-mini", 
    temperature=0, 
    api_key=api_key, 
    timeout=30.0, 
    max_retries=2, 
)

Setting a low temperature improves consistency, while explicit timeouts and retry limits help the system recover gracefully from transient API failures.

Step 3: Route before you generate

Not every request requires a large language model. Deterministic logic can often answer simple questions. Routing inexpensive requests away from the LLM yields the most significant cost reduction in production systems.

Python
def classify_request(question: str) -> str:
    q = question.lower()

    if "status" in q:
        return "metrics"

    if "runbook" in q:
        return "retrieval"
   
    return "generation"

Step 4: Add a semantic cache

One of the simplest and most effective optimizations is an exact-match cache, which returns a previously generated response when the same question is asked against the same retrieved documents, avoiding unnecessary model calls.

Python
import hashlib

# Using an exact-match (lexical) cache
exact_match_cache = {}

def cache_key(question: str, sources: list[str]) -> str:
    """
    Generate a deterministic cache key from the user question
    and the retrieved document identifiers.
    """
    fingerprint = question + "|" + "|".join(sorted(sources))
    return hashlib.sha256(fingerprint.encode()).hexdigest()

# Example usage in the pipeline:
# key = cache_key(question, source_ids)
# if key in semantic_cache:
#     return semantic_cache[key]

Step 5: Budget your context

Most retrieval pipelines return far more text than the model actually needs. Instead of stuffing the context window with every retrieved document, establish a strict context budget.

Python
import tiktoken

encoder = tiktoken.encoding_for_model("gpt-4o-mini")
MAX_CONTEXT_TOKENS = 2500

def build_context(chunks):
    context = []
    used = 0

    for chunk in chunks:
        tokens = len(encoder.encode(chunk.page_content, disallowed_special=()))

        if used + tokens > MAX_CONTEXT_TOKENS:
            break 

        context.append(chunk.page_content)
        used += tokens

    return "\n\n".join(context)

Step 6: Retrieve once

Repeated retrieval is a surprisingly common flaw in multi-agent systems. The rule is simple: retrieve once, reuse everywhere.

Python
from langchain_core.documents import Document
from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings

documents = [
    Document(
        page_content="Database latency often follows connection pool exhaustion.",
        metadata={"source": "db_runbook"},
    ),
    Document(
        page_content="Node pressure can increase API response times.",
        metadata={"source": "cluster_runbook"},
    ),
]

embeddings = OpenAIEmbeddings(api_key=api_key)
index = FAISS.from_documents(documents, embeddings)

retrieved_docs = index.similarity_search(question, k=4)
shared_context = build_context(retrieved_docs)

Now, every downstream agent consumes the same optimized context instead of launching its own redundant retrieval pipeline.

Step 7: Route models intelligently

Large models should solve complex problems. Everything else belongs to a smaller, faster model.

Python
from langchain_openai import ChatOpenAI

small_model = ChatOpenAI(model="gpt-4o-mini", temperature=0, api_key=api_key)
large_model = ChatOpenAI(model="gpt-4.1", temperature=0, api_key=api_key)

def choose_model(question: str):
    """Route requests to the most appropriate model based on complexity."""
    if len(question) < 200:
        return small_model
    return large_model

This strategy drastically reduces operational costs without noticeably affecting response quality.

Step 8: Estimate tokens before sending

Without token telemetry, optimization is just guesswork. Monitoring usage makes efficiency measurable and helps engineers detect cost regressions.

Python
import tiktoken

encoder = tiktoken.encoding_for_model("gpt-4o-mini")

def estimate_tokens(messages):
    """
    Estimate input tokens for an OpenAI-style chat payload.
    Note: This is an estimate, not an exact billing calculation.
    """
    tokens_per_message = 3
    tokens_per_name = 1
    total = 0

    for message in messages:
        total += tokens_per_message
        for key, value in message.items():
            if isinstance(value, str):
                total += len(encoder.encode(value))
            if key == "name":
                total += tokens_per_name

    # Every reply is primed with additional assistant tokens.
    total += 3
    return total

Step 9: Validate responses

Production systems must return structured outputs to ensure downstream systems receive predictable, well-formed data.

Python
from pydantic import BaseModel

class AgentResponse(BaseModel):
    answer: str
    sources: list[str]

def validate_response(answer: str, sources: list[str]):
    """Validate and serialize the agent response using a structured schema."""
    response = AgentResponse(
        answer=answer,
        sources=sources,
    )
    return response.model_dump()

Step 10: Build the optimized pipeline

Finally, assemble the architectural components into a single workflow. Notice how failures degrade gracefully instead of crashing the service.

Python
import logging

from langchain_core.prompts import ChatPromptTemplate

logger = logging.getLogger(__name__)

def run_pipeline(question: str):
    """Execute the token-efficient AI workflow with graceful degradation."""
    try:
        route = classify_request(question)

        # Route deterministic requests away from the LLM.
        if route == "metrics":
            return {
                "answer": "Retrieve metrics directly from the monitoring system.",
                "sources": [],
            }

        # Retrieve context once.
        docs = index.similarity_search(question, k=4)

        context = build_context(docs)

        source_ids = [
            doc.metadata.get("source")
            for doc in docs
            if doc.metadata.get("source")
        ]

        # Check exact-match cache.
        key = cache_key(question, source_ids)

        if key in exact_match_cache:
            return exact_match_cache[key]

        # Select the most appropriate model.
        model = choose_model(question)

        # Keep trusted instructions separate from untrusted user input.
        prompt_template = ChatPromptTemplate.from_messages(
            [
                (
                    "system",
                    (
                        "Answer the user's question using ONLY the provided context. "
                        "If the answer cannot be determined from the context, say so."
                        "\n\nContext:\n{context}"
                    ),
                ),
                ("user", "{question}"),
            ]
        )

        chain = prompt_template | model

        result = chain.invoke(
            {
                "context": context,
                "question": question,
            }
        )

        payload = validate_response(
            answer=result.content,
            sources=source_ids,
        )

        # Cache validated response.
        exact_match_cache[key] = payload

        return payload

    except Exception:
        logger.exception("Token-efficient pipeline failed.")

        # Gracefully degrade instead of crashing.
        return {
            "answer": (
                "The AI pipeline encountered an error. "
                "Please continue using the standard operational workflow."
            ),
            "sources": [],
        }

What actually reduced token usage?

When teams instrument architectures like this, the largest savings rarely come from editing prompts. They come from eliminating unnecessary work.

The biggest improvements typically stem from:

• Retrieving documents once instead of multiple times.
• Caching semantically identical requests.
• Routing simple requests away from the LLM.
• Limiting context with explicit token budgets.
• Selecting the smallest suitable model.

These architectural shifts reduce cost and latency while making system behavior significantly easier to reason about.

Lessons learned

Several core principles consistently emerge when optimizing AI systems for production:

• Treat tokens like infrastructure: Tokens are a finite resource, just like CPU cycles or memory. Monitor them, budget them, and optimize them.
• Retrieval is usually the largest source of waste: Repeated retrieval often contributes more unnecessary tokens than verbose prompts. Share context whenever possible.
• Bigger models are not always better: Smaller, faster models effectively handle many operational tasks. Reserve larger models for genuinely complex reasoning.
• Caching is an engineering feature: A semantic cache is more than a performance optimization—it is a core architectural component that reduces cost, latency, and provider dependence.
• Measure before you optimize: Instrumentation must accompany every production deployment.

As AI systems mature, success will increasingly depend on engineering efficiency rather than raw model size. The hidden tax of AI agents is rarely a single expensive prompt; it is the accumulation of redundant retrievals, oversized contexts, unnecessary model calls, and repeated reasoning across distributed workflows.

“The most effective production AI systems are not the ones that generate the most tokens. They are the ones that generate only the tokens they truly need.”

By treating token consumption as a systems engineering problem, organizations can build AI platforms that are faster, less expensive, and highly scalable. Routing requests intelligently, budgeting context, sharing retrieval results, validating structured outputs, and introducing semantic caching are practical techniques that guarantee efficiency without compromising quality.

The most effective production AI systems are not the ones that generate the most tokens. They are the ones that generate only the tokens they truly need.

The post Stop the token bleed: building token-efficient multi-agent systems appeared first on The New Stack.

AI infrastructure platform company TrueFoundry has launched its open-source agent harness, TrueForge. The technology, announced Wednesday, is directly billed as an alternative to Claude Managed Agents, Anthropic’s hosted infrastructure service that runs, sandboxes, and orchestrates autonomous Claude agents.

TrueForge promises to enable software engineers to build, deploy, debug, and govern production AI agents on any model (and the company means any model) or MCP server, while reducing total agent operating costs by an estimated 50%.

While open models such as GLM-5.2 from Chinese frontier model maverick Z.ai are challenging proprietary frontier models at lower costs, most managed agent platforms still lock enterprises into a single vendor’s models, infrastructure, and pricing. 

Challenging the pervading narrative of managed agent platform lock-in

Ex-machine learning tech lead at Meta and now co-founder and CEO of TrueFoundry, Nikunj Bajaj, tells The New Stack that this pervading managed agent platform lock-in is precisely the logic behind his firm’s neutral approach to model vendor choice.

“A provider selling you a million tokens for $50 has zero incentive to tell you the same task could be done using a model that charges 50 cents for a million tokens,” Bajaj says. “Traditionally, one vendor provides the models, builds your agents and decides your token usage, in what order, and with what tools and under governance that the managed agent provider stipulates – and they’re selling the exact same setup to your competitor.”

Fundamentally, he insists, this means “the incentives are misaligned” here and so the “players in this game don’t get a voice to talk to the referee” in managed agent deployment scenarios where there’s always a tradeoff. 

“Why should building powerful agents mean giving up control of your AI stack? We give developers the managed-agent experience without forcing them into one vendor forever,” adds Bajaj.

“A provider selling you a million tokens for $50 has zero incentive to tell you the same task could be done using a model that charges 50 cents for a million tokens. The incentives are misaligned, so the players in this game don’t get a voice to talk to the referee.”

The harness underneath becomes the strategic control point

Although Claude Managed Agents only arrived as a beta release in April of this year, Bajaj and team think they can track an evolutionary curve being etched out here. This arc sees the first wave of AI agents existing on developers’ laptops, inside coding tools and prototypes. But the next wave is moving into customer-facing products typified by hosted infrastructure services with the ability to use shared workflows.

Crucially, that’s a shift that turns the harness underneath those products into a strategic control point, working as an execution layer in an operational loop between the user, the model and the systems it interacts with.

“This is indeed the reality: the harness is the critical layer between the user, LLM, and everything else,” clarifies Bajaj. “Say a developer is building an agent. They bring their own models, but the harness still decides when to call an MCP server, when to use an agent someone has already built, what context to keep, and which model handles which part of the plan.”

This means there are security implications, too. Bajaj specifies that “some actions” still need to run in a completely isolated sandbox, and some data should never be sent to a closed-source model. 

“All of that logic sits in the harness. If software engineers don’t use one, then the developer team has to build all of that logic from scratch,” he adds.

Enterprises will want to own key agent layers

In an open vendor-neutral approach to managed agent platform provision, organizations must manage persistent sessions, tool credentials, execution sandboxes, context, human approvals, debugging, access policies, and spending across every agent they operate. TrueFoundry is betting enterprises will want to own that layer rather than inherit it from a single model provider, but with enterprise governance built in at lower cost.

TrueForge routes every model call and MCP interaction through TrueFoundry’s AI Gateway, so budget enforcement, rate limits, and guardrails can be applied to deliver a governed and secure managed agent experience for enterprises.

Headless chickens, when foo and bar are behind the wheel

When organizations don’t have the same hold on the steering wheel, Bajaj says that he has personally witnessed operations where “foo” and “bar” (standard placeholder names used in computer programming for as yet-unnamed known metasyntactic variable values, rather like John Doe) end up becoming the doers of everything. 

“Every action in the system came from a generic shared account, not a person you could actually identify. So when something changed or broke, you had no idea who to talk to. Once, when we were halfway through a migration from shared access to individual access, some keys were rotated. Half the company was still on the old account, and the system broke for half the company,” he explains.

Teams can run TrueForge on their own infrastructure, bring their own models, MCP servers, and API keys, and route each task to whichever model fits the cost, latency, or quality needs of that job. But does that mean workloads might become too fragmented that way?

“On the contrary, workloads become more uniform,” enthuses Bajaj. “Most teams already bring their own models by default. What changes is that organizations get to define what it takes for a model, agent, or MCP to belong in their registry. I call it the agent development life cycle, or ADLC. Once you own that, you can enforce the same operating principles across everything.”

In practice, the TrueFoundry team confirms it has seen most AI-centric software engineering operations converge on “roughly a dozen models” for typical tasks, plus a few specialized models for niche work. 

Is Anthropic doing something wrong?

TrueForge ships with support for OpenAI, Anthropic, and 20+ additional models, along with 40+ built-in tools, sandboxed execution, human-approval workflows, large-context handling, generative UI, and web search powered by Tavily. But despite offering a Claude Managed Agents alternative, Bajaj goes to pains to point out he doesn’t hold Anthropic up as some kind of pariah. 

“This isn’t about Anthropic doing something wrong,” confirms Bajaj. “It’s that it doesn’t own every model in the world. Claude Managed Agents can only choose from the finite set of models Anthropic offers. There are open models that are terrific at certain tasks at a fraction of the cost, or simply more capable for that particular job. An open harness has a much wider set of choices.”

When you own the harness, you can get rid of the parts that don’t apply to you

He underlines his point by pointing out that Anthropic also has to build one harness for a very broad set of customers; a truth that means its system prompt has to account for all kinds of instructions, guardrails, and corner cases. 

“Many of those elements may have nothing to do with a developer’s own use case, but they still go into every call and add cost and latency. When you own the harness, you can get rid of the parts that don’t apply to you and make it extremely specialized,” he adds.

“To be clear, we support Anthropic as a first-class provider because its models are great. There will be many cases where our users want to use them. The point is not to limit that choice to Anthropic alone.”

To validate its statements here, TrueFoundry has tested the above claim on a total of 14 level-one and level-two tasks from DevRev’s public Enterprise-Bench. The company says TrueForge “came in 50% cheaper at similar accuracy”, so the savings came from using fewer tokens and having access to models outside Anthropic’s set that were better suited to specific tasks.

“To be clear, we support Anthropic as a first-class provider because its models are great. There will be many cases where our users want to use them. The point is not to limit that choice to Anthropic alone,” Bajaj concludes.

TrueFoundry is also launching a hosted, pay-per-usage version of TrueForge for teams that want the same experience without managing the infrastructure themselves.

The post An open source rival to Claude Managed Agents just launched appeared first on The New Stack.

Mistral is giving enterprise customers until August 31 to replace the Google Drive and Microsoft SharePoint Knowledge Connectors they use in Vibe Work with MCP-based alternatives. The company says in its Knowledge Connectors documentation that both existing connectors will be shut down and deleted on that date.

There is no automatic migration, so administrators will need to install the MCP replacements before every user reconnects their Google or Microsoft account. The move changes how Vibe Work reaches company documents, but Mistral has said little about the retrieval architecture behind the new connectors.

Mistral stores a searchable index

With the current system, an administrator chooses which Google Drive folders or SharePoint sites the organization wants to make available, then Mistral processes those files and stores the resulting index in its European data centers.

Once the index is ready, users connect their personal accounts, and when they search in Vibe Work, the connector checks permissions copied from Google Drive or SharePoint so the results include only files they can access, while scheduled synchronizations pick up later changes and deletions.

This setup lets Mistral handle retrieval by searching a prebuilt index whenever a user submits a query. The company says indexing can take anywhere from a few minutes to several hours, depending on how much data the organization includes, although that work is completed before users begin searching.

The company says indexing can take anywhere from a few minutes to several hours, depending on how much data the organization includes, although that work is completed before users begin searching.

MCP shifts retrieval off-platform

The company defines MCP as a common interface that lets models call tools and retrieve data from external services — the same protocol layer that is reshaping how AI products connect to external APIs across the industry.

In June, Mistral added Google Drive and SharePoint to a directory containing more than 60 integrations, but it did not explain how those two connectors retrieve documents or say who operates the underlying MCP servers. The setup can range from live calls to the source API to a server-managed search index, with any combination of the two in between. That flexibility is part of what makes MCP appealing in some environments and unnecessary in others, but it also means that the behavior of a given connector depends entirely on its operator.

Permissions rules remain unclearThe company doesn’t run the third-party servers behind these connectors, so it can’t promise how they will behave or what they will do with customer data. That gap between the governance Mistral once provided and what it’s now handing off to third parties reflects a trend in how enterprises are adopting MCP connectors without fully resolving the governance layer. As other platforms have learned, opening the door to external servers means trusting the protocol and the operator and building the guardrails to go with it.

The migration notice offers even less detail. It doesn’t say whether the Google Drive and SharePoint MCP servers retrieve files directly from Google and Microsoft. Any caching remains unexplained, leaving customers unsure where retained data would live. Mistral makes no promise that searches will be as fast or return results of the same quality.

Mistral makes no promise that searches will be as fast or return results of the same quality.

The outgoing Google Drive connector follows the sharing rules already attached to each file, including group and domain access. A file set to “anyone with the link” still isn’t automatically visible to everyone in the organization. SharePoint uses Microsoft Entra ID groups, so older groups created only within SharePoint aren’t picked up.

Mistral hasn’t said whether the MCP replacements will follow the same rules. OAuth can limit a server’s access to the connected user, but that doesn’t mean search results will be filtered exactly as they were in the outgoing index. The protocol wasn’t designed to enforce that kind of enterprise permission model; the connector’s retrieval layer has to do it.

OAuth can limit a server’s access to the connected user, but that doesn’t mean search results will be filtered exactly as they were in the outgoing index.

Deletion timeline still unresolved

The company says disabling a Knowledge Connector results in permanent deletion of the indexed data, but the deprecation notice does not say whether the August shutdown will trigger that process automatically or how long the deletion will take. It also leaves administrators unsure whether they should disconnect the old connectors themselves before the deadline.

Because the same connectors work in Vibe Code and Mistral’s workflow system, teams can use the same approach to expose external data across chat, coding and automated jobs — a pattern that is becoming more common as MCP connectors spread from chatbots into production infrastructure.

Mistral also recommends checking server output for signs of prompt injection — advice that underscores the still-emerging challenge of securing the space between AI agents and the external services they access.

The post What happens to your indexed data when Mistral flips the switch? appeared first on The New Stack.