AI — Page 62

The Northeastern and Stanford team open-sourced a Python framework for meta-agents that can inspect, fork, replay and repair agent runs.

Meta's reported compute business would monetize overflow capacity, but the scarce assets are power, frontier clusters, and delivery timing.

The Scale AI and Center for AI Safety benchmark tests agents on freelance projects judged against accepted human work.

Ben Colman's 2021 bet is becoming a line item for fraud, identity, hiring, and communications teams.

A reported Meta Compute business would sell GPU capacity and hosted models into a market already controlled by AWS, Azure and Google Cloud.

The Broadview moving operator is adding a consumer marketplace and AI concierge while News Corp's Move sharpens its focus around Realtor.com.

The rollout puts Quinn Litherland's AI operator model inside a 27-brand roofing platform where missed calls become missed revenue.

Cloudflare is separating search indexing from AI use, adding default blocks for training on ad-supported sites and building x402 payments into its edge network.

BMW says Figure 02 supported production of more than 30,000 X3s, while Figure 03 moves next into Spartanburg logistics work.

The Prague lab raised an undisclosed Series A led by Creandum after taking reinforcement learning from games into live markets.

The beta turns inbox retrieval into a paid Gemini workflow across Gmail, Docs and Keep, tightening Google's grip on AI productivity.

The San Jose chip startup says it has working silicon, first racks built and more than $1B in signed customer contracts.

The new Gemini media models put faster image generation and conversational video editing into AI Studio, the Gemini API and enterprise tooling.

Anthropic has now announced Claude Sonnet 5 through the official Claude account, shifting the story from leak-watch to whether the new Sonnet tier delivers the claimed price cut, 1M context and near-Opus positioning.

A reverse-engineering post claims Anthropic's CLI mutates hidden prompt text when proxies, time zones, or China-related domains appear, raising a bigger question: what else is your coding agent saying before you ever type?

The MIT-licensed model family spans 9B to 397B parameters, but its benchmark lead rests on DeepReinforce's own harness-heavy evaluations.

Scoble says his lists include 13,000 AI founders, but the more important bet is packaging a hand-curated X graph as startup discovery infrastructure.

LongCat-2.0 is billed as an open-source MoE model, but the release still says full weights are coming soon.

WIRED says contractors on Meta's Cannes project used dummy under-18 accounts to probe ChatGPT, Gemini and Character.AI.

Niklas Gustavsson's Honk system shows why enterprise AI coding is becoming a verification and workflow problem, not just a model problem.

Piotr Migdal's June 29th writeup turns April's Qwen 3.6 buzz into a practical guide for running coding work locally.

The former Google and eBay engineering leader built Zettair around an early-2000s IR system, making expertise the point, not the footnote.

Leonhard Soenke and Patrice Becker say TAR raised $27 million to build behind-the-meter energy systems for data centers.

The Cyprus-based builder is arguing for travel tools that plug into Hermes, OpenClaw and user-owned personal agents.