Chinese open-weight models like GLM-5.2 and Kimi K3 have surpassed American ones and reached commercial agentic viability.
“Since about April 2025, Chinese AI companies have been the clear leader in open-weight models.”
Latest signals
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
OpenAI calls for coordinated global AI standards covering evaluation, reporting, and governance to improve safety.
OpenAI expands its Academy with new learning paths for employees, developers, leaders, educators, and students.
V7 uses GPT-5.6 to turn scattered company files into institutional memory for AI agents.
“V7 turns scattered company files into context agents can use to complete complex, source-linked work.”
Parse error
A new hire reports their big-company team is forced to ship Claude Code output nobody reads or understands.
“People are working 12 to 13 hours a day just to press enter. Nobody is reading anything.”
AppLovin's ML-driven ad platform inside mobile games powers a ~$50B gaming ad market growing 60% year-over-year.
“there's probably about $50 billion of advertising being spent every single year in this mobile gaming ecosystem”
Simon Willison released llm-keys-ui 0.1, a plugin for securely adding API keys to remote coding-agent machines without pasting them.
“I don't like pasting API keys into agent sessions, so I wanted a way to get those keys onto a machine without pasting them into the ChatGPT app directly.”
Salesforce's Dreampitch competition awards $100K to entrepreneurs building a better world with AI.
“entrepreneurs building a better world with AI. Nothing could be more impactful.”
a16z-backed Paid in Full Foundation aims to financially honor hip-hop's original creators.
“Hip-hop don't have social security. You have to work.”
A startup is betting on diffusion-based LLMs because they are inherently more parallel at inference time than autoregressive transformers.
“the bitter lesson is that the more parallel solution is the one that is eventually going to win”
Transformers.js 4.3 adds structured output that constrains generation to a JSON schema or regex in the browser.
“it makes sure that every new token actually follows the structure that we want”
datasette-explain 0.2.2 adds explain plan support on read-only stored-query pages.
“Explain plans now work on read-only stored-query pages.”
Simon Willison released datasette-auth-github 1.0, fixing a session cookie expiration bug.
“I'm trying to get better at promoting stable plugins to 1.0.”
will.i.am's FYI.AI builds personal AI agent architecture letting individuals supercharge their lives like an enterprise.
“we have the architecture for the individual to supercharge themselves with personal data and run their life their aspirations and and goals like an enterprise for the individual”
Two lines of algebra make transformer RMS norm cheaper and faster via weight folding and deferred division.
“the GPUs are not slow or bad at math, but they're bad at everything else around the actual math”
AI21 engineers traced rare 'one-in-a-thousand gibberish' output bugs unique to vLLM during Jamba RL training.
“there is no crash there's no warning no error and there's high confidence that's not a quality issue”
Open-weight models have reached frontier quality and run agentic workloads ~5.6x cheaper than closed models.
“Open weight models have crossed the quality threshold”
Small open-source models can match or beat frontier performance on specific tasks at far lower cost.
“actually for specific tasks you can be at frontier or beyond frontier performance”
A wildlife photograph of a California Sea Lion and Brandt's Cormorant at Pillar Point Harbor.
Baseten's Philip Kiely presents the first public addendum to his Inference Engineering book covering new advances since February 2026.
“This is the first public addendum of all new information since the book came out.”
Vercel CEO argues AI's killer app is code generation, but quality and picking the right things to build now differentiate and beget trust.
“the killer app of AI in my view is coding”
CoreWeave is building an inference platform offering serverless and dedicated consumption models to serve small to trillion-parameter workloads.
“another feature that we have on the serverless side is what we calling provisioned throughput”
Reese Witherspoon discusses storytelling and her media company Hello Sunshine at Dreamforce 2026.
“You've spent decades proving that being underestimated is usually just someone else's mistake.”