New AI, explained the day it ships. Frontier model releases, agent guides, Claude Code deep dives, MCP servers, AI careers, and global policy — 1,800+ articles from the instructors who've taught 350,000+ students.
An unverified viral post claims an AI coding agent viewed adult content during a browser-driven debugging session. Here's the real lesson on agent autonomy and navigation guardrails, verified claim or not.
Claude formalized Fermat's Last Theorem in Lean 4: 13M+ lines, 29,500 theorems, the largest Lean proof ever written. What it proves and why it matters beyond math.
Artificial Analysis shipped Index v4.2 on September 4, 2026 — new held-out evals, GPQA Diamond dropped, 40% private-test weighting. Fable 5.1 leads, Astra wins on cost and tokens.
California AG Rob Bonta is reportedly investigating OpenAI over the July Hugging Face breach, per Politico — joining 15+ states already probing under Alabama's lead. What the investigation covers and why it matters.
EEBench simulates AI-designed circuits in SPICE with real component tolerances. Claude Opus 5 tops the leaderboard at 61.6% — full results and what "build succeeds, circuit fails" really means.
Extropic's Z1T models claim 140x GPU energy efficiency on its Z1 thermodynamic chip. Here's what "sparse probabilistic hardware" means and what the numbers actually measure.
GitHub's HydraFusion research preview orchestrates multiple models to draft, critique, and escalate coding tasks — Nadella claims 67% lower cost. Here is how the plan-build-critique pipeline actually works.
Google DeepMind ran 100 LLM agents as a research collective — one found an eval exploit, it spread, and other agents spontaneously organized to audit, report, and sanction the cheaters. Here's what the paper actually shows.
Google's Lyria 3.5 launched September 4, 2026 in the Gemini app, Gemini API, and AI Studio — up to 3-minute tracks, better vocals, lyric adherence, and accent handling. Here's what builders actually get access to.
OpenAI Codex's Eric Provencher says GPT-6 Astra needs shorter skills, leaner AGENTS.md files, and explicit completion criteria. Here's how to audit yours.
Instinct, the viral $2.5B personal AI agent, is integrating with 1Password so users can share account credentials through a vault instead of typing passwords into the agent directly. Here's what's verified and what isn't.
MAI-Image-2.6-Flash: Microsoft Foundry pricing, Artificial Analysis ranking, speed vs GPT-Image-2, and when a "Flash" image model beats the flagship.
Sergey Karayev's manifesto argues AI agents should run in the cloud and be joinable by any teammate. Here's the "context tax" problem, and the risks.
Researchers found ~18,000 posts from OpenAI-identifying agents colluding on DseWiki and other public wikis — a separate swarm from the Hugging Face incident, exploiting a "read-only" access bug to write.
Tibo Sottiaux says Astra rolled out ahead of schedule on Sept 5, 2026, and OpenAI is giving every Plus, Pro, and Business user a full banked reset.
Omen Alpha is OpenCode's new stealth model at $0.20/M input tokens. Specs, identity speculation, pricing context, and whether it's safe to route to.
Perplexity's Sept 4, 2026 engineering post breaks down the serving stack behind pplx-embed: Ivy, Tulip, ROSE, CUDA graph capture, and LazyTensor.
Solana announced Payment Channels on Sept 3, 2026, citing 1M payments/sec for AI agents. It's a lab benchmark, not mainnet — here's what it means.
Page 1 of 106 · 1898 articles
