https://ai-tldr.dev/releases/google-gemini-irregular-breakout/ AI/TLDR en 2026-09-19T12:00:00+00:00 Gemini hacked three real companies — a breakout during Google's security test https://ai-tldr.dev/releases/sglang-v0-5-20/ AI/TLDR en 2026-09-18T12:00:00+00:00 SGLang v0.5.20 — CUDA 12 wheels retired, radix cache covers every model https://ai-tldr.dev/releases/anthropic-claude-code-2-1-278/ AI/TLDR en 2026-09-19T12:00:00+00:00 Claude Code 2.1.278 — auto mode's safety checks stop costing you tokens https://ai-tldr.dev/releases/prinz-astra-wwi-cipher/ AI/TLDR en 2026-09-17T12:00:00+00:00 GPT-6 Astra reads a 1918 German radio cipher — matched to Royal Navy logs https://ai-tldr.dev/releases/convai-laya/ AI/TLDR en 2026-09-18T12:00:00+00:00 Laya — an open 421M decision model that replies in 33 milliseconds https://ai-tldr.dev/releases/hacktron-openai-claude-exploit/ AI/TLDR en 2026-09-18T12:00:00+00:00 Hacktron used Claude to reach OpenAI's internal code repository https://ai-tldr.dev/releases/google-cc-families/ AI/TLDR en 2026-09-17T12:00:00+00:00 Google CC opens to families — one AI agent for up to six people https://ai-tldr.dev/releases/meta-muse-mac/ AI/TLDR en 2026-09-17T12:00:00+00:00 Muse comes to Mac — Meta's agent acts inside your desktop apps https://ai-tldr.dev/releases/wes-roth-openai-hacked-sep19/ AI/TLDR en 2026-09-19T12:00:00+00:00 Wes Roth — 'OpenAI JUST got HACKED...' https://ai-tldr.dev/releases/theoleecj-semif/ AI/TLDR en 2026-09-18T12:00:00+00:00 SemIf (formerly OpenJev) — typed decisions without generating JSON https://ai-tldr.dev/releases/cactus-needle-3/ AI/TLDR en 2026-09-18T12:00:00+00:00 Needle 3 — an 8-29 MB model that picks tools on a Raspberry Pi https://ai-tldr.dev/releases/us-military-ai-false-intel-report/ AI/TLDR en 2026-09-18T12:00:00+00:00 False AI intel report — US military nearly boarded a Chinese ship https://ai-tldr.dev/releases/deepseek-v4-1-flash-report/ AI/TLDR en 2026-09-17T12:00:00+00:00 DeepSeek V4.1 Flash report — the KV cache drops to 890 bytes a token https://ai-tldr.dev/releases/zai-zcode-workspace-upload/ AI/TLDR en 2026-09-18T12:00:00+00:00 ZCode uploads your whole repo — .git history included, keys held by Z.ai https://ai-tldr.dev/releases/gaearon-conway-refinement-proof/ AI/TLDR en 2026-09-18T12:00:00+00:00 Conway's refinement conjecture — Dan Abramov got AI to write a Lean proof https://ai-tldr.dev/releases/tptacek-how-to-write-with-an-llm/ AI/TLDR en 2026-09-17T12:00:00+00:00 How to Write with an LLM — Thomas Ptacek's rules for keeping your voice https://ai-tldr.dev/releases/anthropic-biomolecular-modeling-kits/ AI/TLDR en 2026-09-17T12:00:00+00:00 Claude optimized 36 biology models — Anthropic releases the speedup kits https://ai-tldr.dev/releases/anthropic-claude-code-2-1-277/ AI/TLDR en 2026-09-18T12:00:00+00:00 Claude Code 2.1.277 — AGENTS.md works when there's no CLAUDE.md https://ai-tldr.dev/releases/sam-witteveen-jev-sep18/ AI/TLDR en 2026-09-18T12:00:00+00:00 Sam Witteveen — 'Jev - The Ultimate Classification Model?' https://ai-tldr.dev/releases/china-telecom-xing4-0-29b-a4b/ AI/TLDR en 2026-09-16T12:00:00+00:00 Xing4.0-29B-A4B — China Telecom's agent model trained on Ascend NPUs https://ai-tldr.dev/releases/anthropic-claude-code-2-1-275/ AI/TLDR en 2026-09-17T12:00:00+00:00 Claude Code 2.1.275 — claude.ai skills and plugins sync to the terminal https://ai-tldr.dev/releases/openai-codex-cli-0-155/ AI/TLDR en 2026-09-17T12:00:00+00:00 Codex CLI 0.155.0 — voice conversations arrive in the terminal https://ai-tldr.dev/releases/qwen-3-8-omni-flash/ AI/TLDR en 2026-09-18T12:00:00+00:00 Qwen3.8-Omni-Flash — Alibaba's omni model gets a 1M-token context https://ai-tldr.dev/releases/openai-misalignment-reporting-framework/ AI/TLDR en 2026-09-16T12:00:00+00:00 OpenAI publishes misalignment reports — six cases from its own training runs https://ai-tldr.dev/releases/two-minute-papers-deepseek-kv-cache-sep18/ AI/TLDR en 2026-09-18T12:00:00+00:00 Two Minute Papers — 'DeepSeek Just Made AI Memory 4x Smaller!' https://ai-tldr.dev/releases/wes-roth-astra-self-jailbreak-sep18/ AI/TLDR en 2026-09-18T12:00:00+00:00 Wes Roth — 'OpenAI's Astra class model JAILBROKE ITSELF...' https://ai-tldr.dev/releases/prismml-bonsai-2-27b/ AI/TLDR en 2026-09-17T12:00:00+00:00 Bonsai 2 27B — ternary model keeps 98.2% of full precision in 5.9 GB https://ai-tldr.dev/releases/nvlabs-sol-pi/ AI/TLDR en 2026-09-17T12:00:00+00:00 SoL-Pi — NVIDIA's harness extension cuts coding-agent tokens by about half https://ai-tldr.dev/releases/anthropic-pace-measurements/ AI/TLDR en 2026-09-17T12:00:00+00:00 Anthropic publishes pace metrics — Claude leads 26% of its own AI research https://ai-tldr.dev/releases/openai-astra-for-law/ AI/TLDR en 2026-09-17T12:00:00+00:00 Astra for Law — OpenAI ties GPT-6 Astra to a 230M-URL legal index https://ai-tldr.dev/releases/bendlang-bend-2/ AI/TLDR en 2026-09-17T12:00:00+00:00 Bend 2 — a language that makes an AI prove its code obeys your laws https://ai-tldr.dev/releases/tim-gowers-fields-letter-dissent/ AI/TLDR en 2026-09-17T12:00:00+00:00 Tim Gowers — why he didn't sign the Fields medallists' AI letter https://ai-tldr.dev/releases/fireship-dream-rsi-sep17/ AI/TLDR en 2026-09-17T12:00:00+00:00 Fireship — 'Did Google just kickstart the intelligence explosion?'