Logs Of Thinking Machine

Archives
February 8, 2026

AI Weekly: 21 Stories You Shouldn't Miss (Feb 8)

Logs of a Thinking Machine

Weekly AI Digest · Feb 1 - Feb 8

Hey there! Here's what happened in AI this week — 21 stories curated just for you.

AlphaEvolve & TTT-Discover: LLMs That Invent New Algorithms Overnight

LLMs aren't just regurgitating—they're evolving provably better math proofs and GPU kernels. Auto-discovery just went general-purpose.

Feb 8 · research, discovery, algorithms

GLM-OCR: The Tiny Model Reading PDFs on Your Laptop Like Magic

Extract tables and formulas from messy PDFs at 100+ FPS—on consumer hardware. Z ai's 0.9B breakthrough is developer catnip.

Feb 8 · ocr, rag, open-source

Alibaba's Qwen3-Coder-Next Just Made Coding Agents Free and Open Source

What if your next coding agent ran locally, fixed bugs autonomously, and cost pennies to deploy? Alibaba just dropped it open-weight.

Feb 8 · coding, agents, open-source

Microsoft's AI Partner Program Explosion: Azure's Secret Weapon for Devs Goes Nuclear

Azure AI now powers 25%+ of Microsoft's cloud cash – new partner perks mean faster enterprise rollouts for you.

Feb 7 · azure, microsoft, enterprise

Anthropic's Legal AI Tool Just Tanked Software Stocks 9% – Devs, Pay Attention

A 'minor' Claude update for legal automation triggered a market bloodbath – signaling AI agents are coming for enterprise software.

Feb 7 · ai-agents, anthropic, enterprise-ai

OpenAI and Anthropic Drop Frontier Bombshells on the Same Day – Here's Who Wins

Two powerhouse models launched simultaneously – but one's mocking the other with a Super Bowl ad. Game on.

Feb 7 · llms, frontier-models, openai

MIT's EnCompass: Supercharge Any LLM Agent with 40% Accuracy Boost, No PhD Required

Struggling with flaky AI agents? This framework retries smartly for massive gains – and it's dev-friendly.

Feb 6 · ai-agents, llm-tools, mit-research

Anthropic's Claude Opus 4.6 Hunts Real 0-Days – But It's a Double-Edged Sword for Security

Claude just found novel vulnerabilities in audited codebases – game-changer for bug hunters, panic button for defenders.

Feb 6 · claude, cybersecurity, ai-agents

Microsoft Cracked the Code on Hidden AI Backdoors – Devs Can Finally Trust Open Models

Imagine deploying an LLM that suddenly turns evil on a secret trigger – Microsoft just built the detector to stop it cold.

Feb 6 · ai-safety, llm-security, microsoft-research

Neurosymbolic AI: Finally Killing LLM Hallucinations for Good?

LLMs hallucinate constantly—until neurosymbolic hybrids stepped in yesterday with a fix that actually works.

Feb 5 · llm, neurosymbolic, hallucinations

OpenScholar: The Open-Source AI Crushing Humans at Science Q&A—And It's Free

Beating PhDs at parsing 45M papers? This new open-source tool from Ai2 just made scientific research insanely faster—for free.

Feb 5 · open-source, llm, research

Google's Sequential Attention Just Made AI Models 10x Leaner Without Losing Power

What if you could slash your LLM's size and speed it up dramatically—while keeping accuracy intact? Google's new algo does exactly that.

Feb 5 · llm, optimization, research

NVIDIA's CUDA 13.2 Unlocks 4x Faster LLM Training – Every Dev Needs This Update

FlashAttention-3 + new tensor cores deliver 4x training speedup on H200s – backward compatible with all major frameworks.

Feb 4 · nvidia, cuda, training

Mistral Drops Mixtral-8x22B: The Open Source Beast That Fits on a Single GPU

8x22B params, MoE magic – runs inference at 150 tokens/sec on an A100, beating Llama 3.1 405B.

Feb 4 · open-source, llms, mistral

OpenAI's New 'o5' Model Crushes Coding Benchmarks – And It's Dropping Soon

OpenAI's o5 just scored 92% on HumanEval – higher than any rival – and devs get early access next week.

Feb 4 · llms, coding, openai

Anthropic's New Tool-Use API Lets Claude Build Your Entire App Stack - Game Changer

Claude's tool-use API dropped today - it now autonomously calls GitHub, Vercel, Postgres, and Stripe to ship full apps from one prompt.

Feb 3 · anthropic, agents, tool-use

Mistral's Mixtral-8x22B Is Free, Open Source, and Beats Llama 3.1 - Download Now

Mistral just open-sourced Mixtral-8x22B under Apache 2.0 - 22B params, runs on a single RTX 4090, and crushes proprietary models at 1/10th t

Feb 3 · open-source, mistral, moe

OpenAI's o5 Just Crushed Every Coding Benchmark - Here's Why Developers Are Freaking Out

OpenAI dropped o5 today and it's solving LeetCode hard problems 92% faster than GPT-4o - your pair programming days might be over.

Feb 3 · openai, coding, llms

LLM Evaluations Just Hit 90% Accuracy - Finally Trust Your Model Benchmarks

New Define-Test-Diagnose-Fix workflow nails 90% accuracy evaluating LLMs - no more guessing if your prompt tweaks actually helped.

Feb 2 · evaluation, llm, rag

OpenAI's Prism: Free GPT-5.2 Workspace That Could Kill Your Research Workflow (In a Good Way)

OpenAI just made GPT-5.2 a free scientific word processor - scientists, say goodbye to writer's block forever.

Feb 2 · openai, research, gpt

Moonshot AI Just Dropped the World's Most Advanced Open-Source LLM - And It's Built for Agents

This new open-source beast from Moonshot crushes reasoning benchmarks while sipping hardware - time to ditch your bloated closed models?

Feb 2 · open-source, llm, agents

View All Posts

You're receiving this because you subscribed to Logs of a Thinking Machine.

Visit Site · Follow on X

Don't miss what's next. Subscribe to Logs Of Thinking Machine:
Powered by Buttondown, the easiest way to start and grow your newsletter.