AI News - Live AI News & Trends | Blockchain.News

Tencent Hunyuan AIG boosts ClawScan security

According to @openclaw, Tencent Hunyuan AIG now powers ClawScan reviews, scanning all ClawHub skills and plugins for vulnerabilities. (Source)

Share More from OpenClaw 10-02-2026 15:22 UTC

Frontier agents surge: 1.9B capacity analysis

According to KyeGomezB, chips through 2027 could run hundreds of millions of frontier agents; DeepSeek V4 Pro benchmarks suggest up to 1.9B. (Source)

Share More from Kye Gomez (swarms) 10-02-2026 13:09 UTC

Claude Opus 5.5 Builds working watermill showcase

According to @claudeai, creators used Opus 5.5 to design a functional watermill, highlighting rapid prototyping and physics reasoning in real projects. (Source)

Share More from Claude 10-02-2026 12:53 UTC

Intelligence Explosion Analysis Warns of Near-Term Risk

According to Geoffrey Hinton, a new CASP paper analyzes recursive self-improvement and finds intelligence explosion could arrive soon, impacting safety policy. (Source)

Share More from Geoffrey Hinton 10-02-2026 12:42 UTC

Muse Spark powers 6 open-problem breakthroughs

According to AIatMeta, mathematicians used Muse Spark 1.1 and 1.2 to co-author six papers solving open problems via meta.ai chat, with peer review. (Source)

Share More from AI at Meta 10-02-2026 11:11 UTC

SWE-chat Dataset Expands to 230K Prompts

According to @StanfordAILab, SWE-chat v2 publishes 230K real-world coding prompts on HuggingFace, enabling training and evaluation of AI code agents. (Source)

Share More from Stanford AI Lab 10-02-2026 11:00 UTC

Tokenizer Myths Debunked: 7 Findings

According to StanfordAI Lab, slides by Rylan Schaeffer reveal surprising tokenizer behaviors and efficiency trade offs; a preprint is forthcoming. (Source)

Share More from Stanford AI Lab 10-02-2026 10:17 UTC

GLM5.3 Matches Claude Mythos in vuln exploits

According to DeepLearning.AI, GLM-5.3 reached 12% vulnerability exploits vs Claude Mythos at 14%, plus updates on Xiaomi leader, Gemini 3.8 Live, and more. (Source)

Share 10-02-2026 06:59 UTC

Tavus Griffin fools video calls, Google watermarks

According to TheRundownAI, Tavus’ Griffin mimics humans on calls, Google extends AI watermarks to labs, and ChatGPT Dots raises Pro value questions. (Source)

Share More from The Rundown AI 10-02-2026 02:30 UTC

KV-streams Accelerates agentic RL 2x

According to @_avichawla, KV-streams prunes KV cache in-place to cut re-prefill costs, nearing 2x faster RL training on SWE-bench with matched accuracy. (Source)

Share More from Avi Chawla 10-02-2026 02:22 UTC

OpenAI launches dots and agents computers

According to @AINewsOfficial_ and YouTube, OpenAI launches dots to give agents computers; Gemini 4 Argon reaches 1M tokens; GPT-6.1 Sol nears Astra, per sources. (Source)

Share More from AI News 10-02-2026 01:45 UTC

DeepSeek Harness launches desktop apps now

According to DeepSeekHarness, packaged desktop apps for macOS and Windows are live, with Linux via npm @ deepseek-ai/dsh, enabling local AI workflows. (Source)

Share More from DeepSeek 10-01-2026 23:00 UTC

SynthID Bio Watermarks Elevate Biosecurity

According to sundarpichai, SynthID Bio watermarks AI protein designs with invisible signatures to boost integrity and biosecurity. (Source)

Share More from Sundar Pichai 10-01-2026 17:12 UTC

Google Compute-in-Space Launch Boosts AI

According to Sundar Pichai, Google and Planet launched a prototype compute-in-space satellite with SpaceX, advancing on-orbit AI processing. (Source)

Share More from Sundar Pichai 10-01-2026 16:30 UTC

PromptScan Secures Swarms Marketplace Prompts

According to KyeGomezB, Swarms now scans all marketplace prompts for injections and risks, adding a protective layer as listings near 1,000. (Source)

Share More from Kye Gomez (swarms) 10-01-2026 13:52 UTC

Claude Mods Unlock Custom Workflows

According to @bcherny, Anthropic now lets users mod Claude to change behavior, UI, and features, shipping as plugins via CLI or desktop. (Source)

Share More from Boris Cherny 10-01-2026 12:27 UTC

Google Cloud launches 31-day AI agents series

According to @_avichawla, Google Cloud’s free 31-day series teaches secure, governed AI agents with daily code tutorials for production deployment. (Source)

Share More from Avi Chawla 10-01-2026 11:19 UTC

Claude3 Accelerates Exact Science Calculations

According to AnthropicAI, Claude reveals cross-domain links and speeds exact quantitative calculations with a new toolkit by Harvard’s Matthew Schwartz. (Source)

Share More from Anthropic 10-01-2026 10:57 UTC

Claude Sonnet 5.5 halves usage for design work

According to @claudeai, starting decks, docs, or designs in Claude cuts usage by 50% for two weeks and Sonnet 5.5 builds slides needing minimal edits. (Source)

Share More from Claude 10-01-2026 10:08 UTC

Training Witnesses Debuts to Fix ML Reviews

According to StanfordAILab, Training Witnesses offers verifiable “honesty” certificates to ease ML conference submission overload and rebuild trust. (Source)

Share More from Stanford AI Lab 10-01-2026 09:35 UTC