Is AI lock-in coming for enterprises?
Palantir's Alex Karp and Mistral's Arthur Mensch are making the same case from different angles: Don't let closed AI providers control your data and deployment.
/AI Weekly/Issue 158
Is AI lock-in coming for enterprises?
Published Tuesday, July 14, 2026
Palantir's Alex Karp and Mistral's Arthur Mensch are making the same case from different angles: Don't let closed AI providers control your data and deployment.
In recent years, as generative AI has shifted from model training toward large-scale inference, an increasing number of AI model developers have turned
One of our chapter leads, Veronika Kolesnikova, wrote an excellent summary of Anthropic's latest interpretability research on the newly discovered "J-space" in Claude. The research explores how language models appear to maintain an internal workspace that supports deliberate reasoning and coordination between concepts. Veronika breaks down the findings into an accessible overview, explains the connection to Global Workspace Theory, and discusses why this research is an important step toward more interpretable, trustworthy, and safer AI systems.
Recursive language models (RLMs) fix context rot by having agents write code that dispatches subagents over context chunks instead of pumping everything in one context window. Deep Agents now implements this through dynamic subagents and a lightweight code interpreter, letting agents programmatically fan out work like grep, map, and reduce over large inputs. We benchmark the approach on OOLONG, a long-context reasoning task, and show it holds up where turn-by-turn agents start to break down.
Agents are changing fast. They are moving from basic chat tools to systems that can plan work, use integrations, hand tasks to other agents, and operate inside real apps and workflows. We’ll look at the new way to build and use agents today through concrete demos, not slides, and show what actually makes them useful in practice. We’ll also see how building agents is becoming accessible to a much broader set of people, not just developers, as these capabilities show up directly in everyday tools. Along the way, we'll introduce the use of rubrics to reason about agent behavior, reliability, and control as agents move into real-world use.
Amy Lu from the LangChain product team walks through configuring Claude Code so every session traces straight into LangSmith. You'll see the exact plugin install and settings file needed to capture every message, tool call, and sub agent run, then how to read that trace and follow a full multi-turn session in the Threads tab. Chapters: 0:00 Why your Claude Code agent's black box moments cost you debugging time 0:29 Three things you need before you start 0:44 Installing the LangSmith tracing plugin 1:10 Pointing the plugin at LangSmith with one settings file 1:21 The settings file, line by line 1:39 Running Claude Code with tracing turned on 1:49 Viewing your first trace in LangSmith 2:06 Following a full session in the Threads tab 2:17 Recap: three commands and one JSON block 2:25 Troubleshooting and what's next Docs: https://docs.langchain.com/langsmith/trace-claude-code
LangChain and NVIDIA launch the NemoClaw Deep Agents blueprint, combining Deep Agents Code, Nemotron 3 Ultra, and OpenShell for open, governed enterprise agents.
Today, we’re excited to share that AlphaEvolve is generally available (GA). AE is a code optimization and discovery agent built on Gemini that helps solve some of the hardest business problems organizations, researchers and developers face.
From local meetups to full-day conferences, the Global AI Community brings developers, architects, founders, and AI enthusiasts together to learn, build, and connect. Discover hands-on workshops, expert talks, community gatherings, and flagship events like AgentCon, AgentCamp, and Global AI Construct. Whether you are exploring AI for the first time or building production-ready agentic systems, you will find opportunities to sharpen your skills, meet like-minded people, and become part of a thriving global community.
A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Go. - microsoft/agent-framework-go
Today, we are launching an open source Model Context Protocol (MCP) server that brings AI-powered dataset discovery to Registry of Open Data on AWS (RODA). As of today, RODA hosts over 1,100 high-value datasets from more than 400 organizations, spanning satellite imagery, life sciences, climate, geospatial, and more. Ask a research question in Kiro, Claude […]
This week's Last Week in AI covers several of the biggest developments across the AI landscape. Highlights include Anthropic's rollout of Claude Sonnet 5 with stronger coding capabilities and enhanced safety measures, Google's new NotebookLM video summaries and Nano Banana 2 Lite image model, DeepSeek's continued expansion, and Etched's push into AI inference hardware. The episode also explores emerging open-source models and benchmarks shaping the next generation of AI systems.
The AI links worth your time, curated by the community and always free.