ToolsVercel CEO on the Shift Toward Separating Models and AgentsVercel CEO Guillermo Rauch discussed how the company’s platform has become central to AI software, citing 6 million daily deployments and over 1 trillion tokens flowing through its AI gateway.4 min read
ToolsAI model Fable produces a high-performance GPU megakernel, signaling increased automation in AI R&DFable autonomously generated a CUDA megakernel that set a top score on KernelBench-Mega, delivering an 18.71× speedup on an RTX PRO 6000 Blackwell relative to an optimized PyTorch baseline.2 min read
ToolsMicrosoft and NVIDIA unveil a joint platform to industrialize agentic enterprise AIMicrosoft and NVIDIA announced a joint platform approach to help enterprises move agentic AI from prototypes into production, arguing that agent-based systems require new infrastructure for orchestration, observability, identity and governance.3 min read
ToolsHugging Face Kernels adds repo type, tightened security and agent-focused toolingHugging Face’s Kernels project introduced a new “kernel” repository type on the Hub, strengthened security with trusted publishers and code signing, and reworked CLIs and tooling to better support multiple frameworks and agent-driven kernel development.5 min read
Toolssqlite-utils 4.0rc2 fixes transaction bugs after Claude Fable review; session cost estimated at $149.25The sqlite-utils 4.0rc2 release fixes several critical transaction-handling bugs discovered with the help of Anthropic’s Claude Fable and was finalized after additional review by GPT-5.5.6 min read
ToolsNewer Claude models send malformed edit-tool calls to Pi, causing compatibility issuesArmin Ronacher discovered that recent Anthropic Claude models (including Opus 4.8 and Sonnet 5) sometimes call Pi’s edit tool with invented extra fields in the edits[] array, so Pi rejects the tool invocation.3 min read
ToolsAI 'hyper‑communication' connected 277 Americans to deliberate the country's top three contributionsUsing a hyper‑communication platform run by Unanimous AI, 277 randomly selected Americans debated for twenty minutes to identify the top three American contributions over 250 years.3 min read
ToolsKyndryl launches enterprise AI orchestration to align agent-driven operationsKyndryl introduced Kyndryl AI Orchestration for Business, a solution built on the Kyndryl Agentic AI Framework that coordinates AI agents across supply chain, commerce, finance, IT and customer service.4 min read
ToolsCurrent AI publishes Open Source AI Gap Map v0.1 and releases dataset on GitHubCurrent AI, a non-profit partnership founded at the AI Action Summit in February 2025, has published the Open Source AI Gap Map v0.1 to index the open-source AI ecosystem.2 min read
ToolsBuilt-in observability for agents' self-improvement on Vercel deployAccording to the announcement, agents can now introspect their previous runs, identify efficiency issues, bugs and unnecessary tool calls, and generate new prompts and skills; this observability is available built-in when the service is deployed to Vercel.1 min read
ToolsDelegating routine coding to lower-power subagents reduces Fable token use, Claude Code users reportParticipants in a Fireside Chat with Cat Wu and Thariq Shihipar from the Claude Code team recommended letting Fable decide when to run tests and which model to use for coding subtasks.3 min read
ToolsMicroVM-based Sandbox now runs Docker and FUSE without restrictionsThe environment called Sandbox can now run Docker and FUSE applications without restrictions, leveraging MicroVMs' instant startup time and flexible runtime options; additionally, it can be integrated with agents using an S3-based filesystem with just 10 lines of code.1 min read
