Archive
Every issue of the AlphaSignal newsletter, newest first.
649 issues total
October 2026
26 issues- Oct 30 NVIDIA ships 4B speech model, DeepSeek runs 1M tokens under 1GB RAM, Mistral tops FrontierCode, new retrieval beats dense search by 17pts
- Oct 29 Anthropic ships Claude for Legal, Liquid AI releases 2B on-device models, NVIDIA drops open speech model, REA open-sources reverse engineering studio
- Oct 28 Google releases Genie 4, OpenAI cracks Millennium Prize math, LiquidAI ships 2B models beating 12B rivals, IBM RAG hallucinates 65x less
- Oct 27 OpenAI ships GPT-6.1 Terra at 1/10th price, Cohere catches Qwen3 probing sandbox, Pomerado turns sites into APIs, OrcaRouter cuts refusals 91.5%
- Oct 26 Kimi drops 104B open model, OpenAI model escapes sandbox via DNS, modders unlock RTX compute, Barron finds softmax bugs in major ML frameworks
- Oct 25 Stronger models often beat costly memory layers when given raw history. Hereβs when agent memory is worth paying for.
- Oct 23 Anthropic launches Claude Security, Meta proves byte models beat Llama with 1/6 data, Grok controls your phone, Recursive ships laptop coder
- Oct 22 DeepSeek open-sources inference engine, Anthropic previews Claude Duet, Unitree demos kung fu humanoid, OpenAI finds PHP flaw
- Oct 21 DeepSeek open-sources inference stack, Anthropic designs 300k proteins, drops 400+ free agents, researchers cut Qwen3 censorship from 58% to 2%
- Oct 20 DeepSeek ships Flash model, Anthropic scanner finds 77 bugs, Nvidia robot framework hits 99%, Google finds bad guesses halve AI success
- Oct 19 DeepSeek ships faster open model, Claude designs quantum circuits 10x efficient, Grok Bot gets 3x speed boost, image gen cut to 8 steps
- Oct 18 What if AI planned the whole answer before writing a single word? Stanfordβs NCM does, with major speed gains.
- Oct 16 Meta ships 816B coding model, GPT-6 solves 98% of expert physics problems, Google releases Gemini 4 Argon, Black Forest Labs drops 5.6B image model
- Oct 15 Claude Code runs parallel sessions, Google releases Gemini 4 Argon, PS5 runs 27B model at 20 tok/s, fork-flush boosts agents 66%
- Oct 14 Anthropic builds custom version control, Fish Audio ships Drama 3, LiquidAI releases 24B on-device model, Google Gemini 4 Argon cuts bad code 30%
- Oct 13 Claude routed through 49 free providers, Z.ai drops GLM-Legal, NVIDIA open-sources duplex voice model, Mistral Large 4 beats closed models
- Oct 12 Anthropic launches 1,000-agent orchestration, AirLLM runs Llama 405B on 8GB VRAM, Codex predicts your next message, Baidu open-sources 3B OCR model
- Oct 11 Multi-model agents waste compute on every handoff. New KV cache techniques cut that cost, with speedups of up to 25x.
- Oct 9 Anthropic ships Claude Dashboards, OpenAI opens Decisions API beta, dev rewrites TypeScript compiler in Rust using $20k of Claude
- Oct 8 Anthropic releases Claude Haiku 5.5, OpenAI ships GPT-6 with UI tools, Stanford diffusion runs 4.6x faster, robot planner hits 73% success
- Oct 7 Mistral drops 1T multimodal model, Claude integrates into Google Docs, OpenAI releases math results, Anthropic launches $45k credits
- Oct 6 OpenAI watermarks ChatGPT, Reflection ships 501B Beam model, DeepSeek cuts token costs 98%, FlashAttention-3 has silent BF16 bug
- Oct 5 Aleph Alpha open-sources 78B model, photo-to-3D tool ships, Cua runs AI agents on Mac desktop, Wavestone maps 7-part coding agent blueprint
- Oct 4 Skills, harnesses, and models can now rewrite themselves. Here's which part of your agent stack to evolve first.
- Oct 2 Anthropic opens Claude Code to TypeScript mods, OpenAI scans full GitHub repos, 1.58-bit 27B model fits 16GB GPU, Cursor adds inline charts
- Oct 1 Google unveils Gemini 4 Argon 1M output, OpenAI launches Ultrafast GPT-6 tier, Fermion ships 164MB speech model, researchers crack AI reasoning blocks