Desk
Developer tools, platforms, and product launches.
After a sandbox agent used DNS to reach a public chatbot on September 20, OpenAI paused training, evaluation, and tool-use inference on its most capable models—and the stop is still on.
Meta is wiring prompt-to-game tools into Horizon and then into Facebook and Instagram feeds — a distribution bet as much as a creativity bet.
Meta says Muse was built from scratch but admits it was heavily inspired by OpenClaw — and the security story is still unfinished.
MarkTechPost’s Jev tutorial shows System One as typed Choice, Score, and Noul answers your code branches on — confidence gates, composite weights, and speculative fan-out included.
Google’s Gemini 3.8 Flash TTS and Flash-Lite TTS add prompt-built voices, 2,000+ library options, and line-by-line stage directions — API-only, with SynthID on every clip.
OpenAI just hired Patreon’s cofounder and two product/engineering leaders — the blueprint team for creator subscriptions — days before DevDay.
OpenAI fills the GPT-6 mid-tier with Sol and Luna — half the prior API rates versus 5.6 promo pricing, stronger caching, and a cost-per-task pitch aimed squarely at Opus and Fable.
EvilTokens turned inbox compromise into a subscription product — AI for target picking and lure drafting — until Microsoft and partners tore down the shop.
NVIDIA’s SoL-Pi trims coding-agent token traffic nearly in half by evolving the harness — same agent brain, fewer wasteful round trips.
NVIDIA’s eval guide says stop grading vibes and tool-call trivia — gate releases on whether the environment actually reached the goal state.
StepFun’s Step 5 Preview bets on a 600B MoE with 27B active and a 1M context — selling agentic work on price, with open weights promised mid-October.
Qwen’s LiveTranslate cuts average interpretation lag to 2.3 seconds across 60 languages — real-time speech translation that aims to feel like a meeting, not a tape delay.