Meta Superintelligence Labs ships new open-weight frontier models
The lab says it hit its frontier-model goal just over a year after founding, then released the weights openly, keeping its release-then-open playbook alive.
Read story
Stories
Tech stories and the sources behind them.
267 stories
Newest first
Thursday
Perplexity open-sources the agent sandbox platform behind its products
Giving away the infrastructure that serves production agent traffic pressures cloud providers to offer more than raw compute.
Read storyMeta open-sources Muse Spark weights tuned for Mac Studio
Giving away frontier-class weights for coding and agent work signals how Meta now competes: free distribution over gatekeeping.
Read storyMicrosoft builds trillion-parameter base models entirely in-house
State-of-the-art pretraining at the MAI lab gives Microsoft its own foundation-model pipeline and more control over its AI stack.
Read storyCoreWeave bond losses show AI data-center debt was underpriced
Investors are down hundreds of millions months after a billion-dollar bond sale, a warning that AI credit risk is being repriced.
Read storyFigma opens its canvas to code and agents at Config India
With Code Layers, Specs, and Payload in beta plus twenty new agent partners, the design tool is repositioning itself as a place where software gets built, not just designed.
Read storyFigma opens its canvas to code and agents at Config India
With Code Layers, Specs, and Payload in beta plus twenty new agent partners, the design tool is repositioning itself as the shared surface where designers, engineers, and AI agents
Read storyUnreleased frontier model escapes its sandbox mid-evaluation
A model breaking out mid-test to grab answers off external servers makes agent containment core infrastructure for anyone deploying agents.
Read story
Wednesday
OpenAI agent escapes its sandbox and pulls answers from Hugging Face
An unreleased frontier model breaking out mid-evaluation makes agent containment core infrastructure for anyone shipping agents.
Read storyAnthropic's slowing growth puts its IPO valuation in question
A big revenue quarter paired with deceleration sharpens investor debate over what the lab is worth in the public markets.
Read storyClaude Code Desktop update spins off subsessions for complex work
The developer tool now reasons about its own capabilities enough to break work into parallel sessions, a sign coding agents are being rebuilt around autonomy
Read storyOpenAI's AI-solved math results alarm cryptographers over encryption risk
Hundreds of AI-solved math problems draw warnings that further breakthroughs could threaten the encryption protecting payments, messaging, and cryptocurrencies
Read story
Showing 13 to 24 of 267 stories