Archive
Every issue of the AlphaSignal newsletter, newest first.
649 issues total
July 2026
27 issues- Jul 31 Claude hacks real orgs in safety tests, Gemini Robotics 2 ships, Microsoft open-sources image-to-3D 4B model, NVIDIA finds Muon beats AdamW
- Jul 30 OpenAI ships security CLI, Fish Audio raises $52M, robot policy hits 97.3% accuracy, PixelRAG beats text RAG by 18%
- Jul 29 MCP goes stateless, Claude cracks encryption in days, Cursor launches India plan, Microsoft open-sources graph RAG, AI books hit 20% Amazon sales
- Jul 28 NVIDIA backs SSI compute push, Kimi K3 ships with corp lockout, MCP server cuts token costs 95%, 3B model beats Qwen 9B
- Jul 27 Claude Opus 5 builds game one-shot, scores 30.2% on ARC-AGI-3, ChatGPT Work agent logs into sites, Swiss AI releases 262K context model
- Jul 26 Standard tokenizers shred your CSVs. A new class of tabular foundation models finally reads tables the way ML needs.
- Jul 24 OpenAI brings voice agents to desktop, Anthropic upgrades Claude voice with Opus, FLUX 3 unifies image/video/audio, GPT cracks 6 Erdős problems
- Jul 23 Anthropic ships Claude Code security plugin, Cursor cuts AI costs 60%, tokenizer hits 24GB/s, Microsoft open-sources prompt rewriter
- Jul 22 Claude learns by watching, OpenAI breaches Hugging Face, Claude/GPT/Kimi ace IMO 2026, Google drops Gemini Flash cybersecurity model
- Jul 21 Dev clones Palantir war room free, Cursor rebuilds SQLite in Rust, Baidu releases open-source OCR, one question fingerprints any LLM
- Jul 20 Phone video becomes 3D world at 60fps, Baidu ships 3B OCR model, Cline benchmarks Kimi K3 vs Fable, dots unlock hidden LLM reasoning
- Jul 19 Why 90% of your tasks don't need a frontier model and what to run instead.
- Jul 17 Kimi drops 2.8T-param model beating GPT-5, Gemma 4 gets 70% faster prefill, Claude Code gains MCP live data, China open-sources 20 FPS 3D scene
- Jul 16 Thinking Machines releases Inkling multimodal, PrismML ships 27B at 3.9GB, Ant Ling trains 1T param model zero annotations
- Jul 15 PrismML ships 27B model at 3.9GB, Nvidia releases 1B multilingual RAG embeddings, GLM 5.2 runs on CPU, LLMs gain sleep-based memory consolidation
- Jul 14 GPT-4o renders Blender scenes hands-free, 744B model runs on 25GB no GPU, Codex fork hits 60K stars, Llama 8B beats GPT-4 on agents
- Jul 13 GPT-5.6 Sol cracks 50-year math conjecture, Cursor ships agent history, Kimi K2 releases open agentic model, Kyutai transcribes songs to MIDI
- Jul 12 Scaling every request to the frontier is a fast track to API bankruptcy. Here's how routers fix it.
- Jul 10 OpenAI ships ChatGPT Work agent, Cognition releases SWE-1.7 at $1.97/task, Cohere drops Aya Vision 8B, Claude Code hides /advisor command
- Jul 9 Anthropic cuts costs 54% with managed agents, OpenAI ships duplex voice model, Prime Intellect raises $130M, Tokyo LLM sells ebook on real internet
- Jul 8 Anthropic cuts costs 50% with agent trick, Google releases Gemma 4, Stanford builds Git for AI agents, Microsoft ships regex LLM control
- Jul 7 Anthropic finds Claude's hidden workspace, NVIDIA ships GR00T N1.7, Tencent releases 295B MoE, OpenAI cuts Realtime latency 25%
- Jul 6 Dev hacks Claude Code costs 70%, Mistral solves 587 Putnam problems, Sakana hits 93% on Sudoku, OpenMed ships on-device PII redaction at 755 tok/s
- Jul 5 Manual harness engineering is brittle. New research lets the model fix its own scaffolding instead.
- Jul 3 Anthropic launches Claude Science, NVIDIA splits 30B model for 2.42x speed, GitHub releases spec-kit, 35B model beats trillion-param giants
- Jul 2 Anthropic ships Claude Sonnet 5, NVIDIA splits 30B model for 2.42x speed, Allen AI releases PDF-to-LLM toolkit, Rust browser loads in 85ms
- Jul 1 NotebookLM adds 60s videos, 7M model beats LLMs on logic puzzles, robot control hits 27x compression, uncensored 27B writes exploits