AI Daily Brief — 19 April 2025
Western Easter Saturday — genuinely quiet day. No frontier-lab releases, no funding rounds, no papers. The NVIDIA H20 export-control fallout continues to dominate weekend commentary. The…
Latest
The State of Reinforcement Learning for LLM Reasoning
Understanding GRPO and New Insights from Reasoning Model Papers
AI Daily Brief — 18 April 2025
Good Friday — but not quiet. Jensen continues his Beijing meetings. Perplexity inks a Motorola Razr distribution deal — its first major smartphone OEM win. OpenAI…
Claude Code: Best practices for agentic coding
Claude Code is a command line tool for agentic coding. This post covers tips and tricks that have proven effective for using Claude Code across various…
AI Daily Brief — 17 April 2025
Google launches Gemini 2.5 Flash in preview — first fully hybrid reasoning model with togglable thinking and budget control. Jensen Huang lands in Beijing one day…
AI Daily Brief — 16 April 2025
OpenAI's biggest single day of 2025 so far: o3 and o4-mini ship with agentic tool use across web search, Python, image gen and image manipulation inside…
AI Daily Brief — 15 April 2025
NVIDIA discloses $5.5B Q1 FY26 charge in 8-K after-hours — H20 China export license is now indefinite. Stock plunges after-hours. AMD files 8-K for up to…
AI as Normal Technology
A new paper that we will expand into our next book
Now in Preview: Groq’s First Compound AI System
Now in Preview: Groq’s First Compound AI System
AI Daily Brief — 14 April 2025
OpenAI launches GPT-4.1 family — three API-only models with 1M context. GPT-4.1 hits 54.6% SWE-Bench Verified (vs GPT-4o 33.2%) and 90.2% MMLU. Mini ~83% cheaper than…
AI Daily Brief — 13 April 2025
Sam Altman teases 'a lot of good stuff' coming this week — kicking off Monday. UNCTAD's 2025 Tech & Innovation Report continues to circulate, projecting AI…
AI Daily Brief — 12 April 2025
Quiet Saturday between Llama 4 and GPT-4.1. ChatGPT memory rollout continues for Pro and Plus subscribers globally. Mira Murati's Thinking Machines Lab $2B seed reporting dominates…
AI Daily Brief — 11 April 2025
Sam Altman live at TED2025: ChatGPT crosses 800M weekly active users. Vanilla Llama 4 Maverick lands #32 on LMArena — far below GPT-4o, Claude 3.5 Sonnet,…
Defending against Prompt Injection with Structured Queries (StruQ) and Preference Optimization (SecAlign)
Recent advances in Large Language Models (LLMs) enable exciting LLM-integrated applications. However, as LLMs have improved, so have the attacks against them. Prompt injection attack is…
AI Daily Brief — 10 April 2025
Markets digest the historic +9.5% S&P / +12.2% Nasdaq rally. ChatGPT memory now references all past conversations — Pro first, Plus next. Sam Altman teases o3…
Add MiniMax MCP repo link to README.md
Add MiniMax MCP repo link to README.md Add MiniMax MCP repo link to README.md
AI Daily Brief — 9 April 2025
Google Cloud Next megaday: Ironwood TPU (192GB HBM, 9216-chip pods, 42.5 ExaFLOPs), Agent2Agent protocol with 50+ partners, Gemini 2.5 Flash, Cloud WAN, Agent Development Kit, Firebase…
Merge pull request #666 from codinglover222/deepseek-doc-fix
Merge pull request #666 from codinglover222/deepseek-doc-fix fix an args description.
AI Daily Brief — 8 April 2025
Amazon launches Nova Sonic — unified speech-to-speech model in Bedrock with a new bidirectional streaming API. Direct challenger to OpenAI Realtime API and Google voice. Llama…
Merge pull request #736 from shihaobai/main
Merge pull request #736 from shihaobai/main Docs: add LightLLM as supported engine
Repurposing Protein Folding Models for Generation with Latent Diffusion
PLAID is a multimodal generative model that simultaneously generates protein 1D sequence and 3D structure, by learning the latent space of protein folding models. The awarding…
Merge pull request #816 from KPCOFGS/main
Merge pull request #816 from KPCOFGS/main Update README.md
Merge pull request #720 from xiaokongkong/main
Merge pull request #720 from xiaokongkong/main modify the explanation of MLA
AI Daily Brief — 7 April 2025
Runway releases Gen-4 Turbo — 10-second video in 30 seconds, 5x faster than Gen-4. Meta VP Ahmad Al-Dahle denies Llama 4 benchmark cheating. Stanford HAI ships…