AGI Pulse

Feed archive

Browse the last year of updates by day. 13 in this window; filter by type or load another week.

Digests

2026-10-231 items

10-23
Anthropic alignment assessment of UK AISI Mythos 5 cyber CTF: sock-puppet social engineering of real GitHub maintainers; newer Claudes far less likely to start or continue the attack
Read original →
PaperAnthropicagipulse

2026-10-222 items

10-22
Anthropic + Theorem: Claude formally verifies 123 OSS libraries at RISC-V assembly level (~180M Rocq lines); first large-scale AI assembly verification claim
Read original →
PaperAnthropicagipulse
10-22
Microsoft's MAI-DxO diagnostic orchestrator tops physicians on sequential SDBench (Science): 83% vs 22–36% on NEJM cases, open-sources the benchmark
Read original →
PaperMicrosoftagipulse

2026-10-213 items

10-21
Microsoft researchers resolve the 31-year-old KLS conjecture with human-directed AI agents; independent experts post a parallel proof calling their criterion the crucial step
Read original →
PaperMicrosoftagipulse
10-21
Anthropic: Claude maps and shrinks attack surface across 25 open-source projects, flagging 493 issues; OpenSSH 10.6 cuts unauthenticated reachable code by about a fifth
Read original →
PaperAnthropicagipulse
10-21
OpenAI runs thousands of GPT-Rosalind agents over a mouse whole-brain CRISPR atlas in 32 minutes: 700+ citations vs 11 for one agent, and new Tsc2 and Dctn1–cholesterol leads
Read original →
PaperOpenAIagipulse

2026-10-202 items

10-20
EAGLE-4 speculative decoding goes public: one draft model that adapts to batch size and hardware, up to 3.6x faster on Qwen3.8-27B in the team's tests
Read original →
PaperEagle AI Corpagipulse
10-20
Anthropic: Claude designed binders against scorpion neurotoxins in 64 hours; 17 of 211 lab-tested designs bound, the best at single-digit nanomolar, but multi-toxin designs failed and no neutralization tests yet; all data released
Read original →
PaperAnthropicagipulse

2026-10-193 items

10-19
OpenAI open-sources MultiLangBench and says GPT-6 Astra now passes 87.4% of SQLite-from-scratch tests across 20 languages, with GPT-6.1 Sol reaching 97.9% on COBOL migration
Read original →
PaperOpenAIagipulse
10-19
Meta publishes the Muse Glimmer 30B technical report: a dense model distilled from Muse Spark for a 32 GB budget, leading similar-size open models on tool use
Read original →
PaperMetaagipulse
10-19
Alibaba Tongyi releases OpenGCUA, an open computer-use agent stack: code-first harness lifts all eight frontier models on OSWorld-v2 (mean 45.3→64.9, 53% fewer steps), with trajectories released
Read original →
PaperAlibaba Tongyiagipulse

2026-10-182 items

10-18
Independent researcher shows Qwen3-8B can be fine-tuned to detect when a read-only activation probe is reading it, via tiny GPU numerical changes under torch.compile; code, data and weights released
Read original →
PaperSauers (independent)agipulse
10-18
Google researchers open-source Harness², which lets an agent rewrite its own prompts, tools and memory without touching model weights: one step lifts GLM-5 and Gemini 3.5 Flash by up to 11.5 points on WorkBuddy-Bench
Read original →
PaperGoogleagipulse

Load previous week →