Releases in this digest

Headlines link to our pages; synopses are MediaRelease.co summaries.

Introducing Teel – Turbo-Efficient Execution Library for Agents (but not only)

AI21 Labs · Original source

AI21 Labs has open-sourced Teel, a Python runtime that overlaps independent IO-bound calls in sequentially written code without async or await. Decorators mark cacheable functions and JIT orchestration while a shadow interpreter dispatches ready calls on background threads. AI21 reports about 83% less runtime on average on Princeton’s Tree of Thoughts and says Teel averaged 19% faster than LLM-generated async JavaScript inside a Deep Agents coding agent.

What It Took to Build Wafer-Scale, and Why It Lasts

Cerebras · Original source

Cerebras has responded to a Culper Research short-seller report that argued its wafers wear out after about two years. The company says that lifetime figure is wrong and that it does not use the PariPoser connector described in the report. Cerebras describes patented assembly techniques that moderate temperature changes across the wafer and says system lifetime will be limited by newer silicon generations rather than thermal failure.

LangSmith BYOC is now available on AWS, Azure, and Google Cloud

LangChain · Original source

LangChain says LangSmith Bring Your Own Cloud now runs on AWS in general availability and on Azure and Google Cloud in beta. Customer traces, evals, datasets, prompts and agent deployments stay in the customer’s own cloud account while LangChain manages the data plane, including orchestration, scaling and platform updates. LangChain runs the control plane for auth, org config and billing metadata; customers own the sensitive data plane.

Want these digests by email?

We are gauging interest in a daily or weekly newsletter for each area. Tell us which areas you would read.

Register interest