Itamar Iliuk

@googlegemma Amazing!!!👏🏻👏🏻👏🏻

Gemma 4 up to 3x faster, directly in your phone! 🚀 Check out the difference Speculative Decoding makes! Multi-Token Prediction (MTP) is supercharging inference speeds for Gemma 4.

@claudeai 👏🏻👏🏻👏🏻🥰🥰🥰

Claude for Word is now available on Pro and Max plans to use alongside Opus 4.7: claude.com/claude-for-word

@claudeai 👏🏻👏🏻👏🏻🎉🎉🎉

Computer use in Claude Cowork and Claude Code Desktop is now available on Windows.

@claudeai 👏🏻👏🏻👏🏻🥰🥰🥰

Computer use is now in Claude Code. Claude can open your apps, click through your UI, and test what it built, right from the CLI. Now in research preview on Pro and Max plans.

@HuggingModels Muito obrigada!!! Portuguese models are very important here in Brazil. 👏🏻👏🏻👏🏻

Meet the Portuguese language's new grammar detective. This BERT-based model specializes in Part-of-Speech tagging, automatically labeling each word in Portuguese text as nouns, verbs, adjectives, and more. It's a game-changer for NLP in Portuguese!

@steipete @OpenAI @openclaw You rock!!!👏🏻👏🏻👏🏻

I'm joining @OpenAI to bring agents to everyone. @OpenClaw is becoming a foundation: open, independent, and just getting started.🦞 steipete.me/posts/2026/ope…

@karpathy I really to discuss more about nanochat then molt/claw/openbot this week.

nanochat can now train GPT-2 grade LLM for <<$100 (~$73, 3 hours on a single 8XH100 node). GPT-2 is just my favorite LLM because it's the first time the LLM stack comes together in a recognizably modern form. So it has become a bit of a weird & lasting obsession of mine to train a model to GPT-2 capability but for much cheaper, with the benefit of ~7 years of progress. In particular, I suspected it should be possible today to train one for <<$100. Originally in 2019, GPT-2 was trained by OpenAI on 32 TPU v3 chips for 168 hours (7 days), with $8/hour/TPUv3 back then, for a total cost of approx. $43K. It achieves 0.256525 CORE score, which is an ensemble metric introduced in the DCLM paper over 22 evaluations like ARC/MMLU/etc. As of the last few improvements merged into nanochat (many of them originating in modded-nanogpt repo), I can now reach a higher CORE score in 3.04 hours (~$73) on a single 8XH100 node. This is a 600X cost reduction over 7 years, i.e. the cost to train GPT-2 is falling approximately 2.5X every year. I think this is likely an underestimate because I am still finding more improvements relatively regularly and I have a backlog of more ideas to try. A longer post with a lot of the detail of the optimizations involved and pointers on how to reproduce are here: github.com/karpathy/nanoc… Inspired by modded-nanogpt, I also created a leaderboard for "time to GPT-2", where this first "Jan29" model is entry #1 at 3.04 hours. It will be fun to iterate on this further and I welcome help! My hope is that nanochat can grow to become a very nice/clean and tuned experimental LLM harness for prototyping ideas, for having fun, and ofc for learning. The biggest improvements of things that worked out of the box and simply produced gains right away were 1) Flash Attention 3 kernels (faster, and allows window_size kwarg to get alternating attention patterns), Muon optimizer (I tried for ~1 day to delete it and only use AdamW and I couldn't), residual pathways and skip connections gated by learnable scalars, and value embeddings. There were many other smaller things that stack up. Image: semi-related eye candy of deriving the scaling laws for the current nanochat model miniseries, pretty and satisfying!

@claudeai 👏🏻👏🏻👏🏻

Claude in Excel is now available on Pro plans. Claude now accepts multiple files via drag and drop, avoids overwriting your existing cells, and handles longer sessions with auto compaction. Get started: claude.com/claude-in-excel

@claudeai Merry Christmas!!!!🥰🥰🥰

Starting at midnight PT tonight, all Pro and Max plans have 2x their usual usage limits through New Year's Eve.

@AnthropicAI 👏🏻👏🏻👏🏻👏🏻👏🏻👏🏻👏🏻

Anthropic is donating the Model Context Protocol to the Agentic AI Foundation, a directed fund under the Linux Foundation. In one year, MCP has become a foundational protocol for agentic AI. Joining AAIF ensures MCP remains open and community-driven. anthropic.com/news/donating-…

@antigravity 👏🏻👏🏻👏🏻👏🏻👏🏻

(1/3) Google AI Pro and Ultra users now receive priority access, featuring our highest, most generous rate limits with quotas that refresh every five hours.

@JasmineLeeTS Morning Jasmine! Have a wonderful day. 😘🥰💖

Good morning 🥰🥰🥰 If you have the chance to make someone happy, be like Nike, just do it 🥰🥰🥰 Sometimes people are struggling silently. Your simple, free act of kindness may make their day. Spread love instead of spreading lies 🥰🥰🥰 WE. Are. Worth. It. 💯*💯% 💖💖💖

@psipietra Perfeita!!!! 😍😍😍

Se não biscoitar, o shape não vem, disseram

Happy Birthday sweetheart! Shine bright like a Diamond in the sky. I miss you so much. 💜💜💜 @WhoisJessicaB

@recap_david Fantastic job. ASSISTANT

$24,000 per year from this simple AI Dentist Voice Agent (and why I'm crazy for giving it away for free) A dental practice was losing $6,000+ in revenue every month from missed after-hours calls. That's 20-25 potential patients walking away because no one was available to book their appointments. So I built an AI voice assistant that handles after-hours dental bookings 24/7 using n8n and ElevenLabs based on internal policies and scheduling availability. Here's what this system does: → Answers calls with a natural-sounding AI receptionist → Collects patient information and insurance details → Checks calendar availability in real-time → Books appointments automatically → Logs all patient details to a Google Sheet The result? This similar AI voice system was sold to a dental practice for $24k per year by another entrepreneur!! This isn't just about dental practices. Any service business losing money from missed calls can implement a similar system. Want the complete n8n workflow template? 1. Retweet & Like this post 2. Comment "ASSISTANT" I'll send you the entire system for free, a full setup walk-through video, including the ElevenLabs automation components.

@OpenAIDevs 👏🏻👏🏻👏🏻👏🏻

We've updated function calling to support files and images as tool call outputs. You can now call functions like `generate_chart` or `load_image` and return those files back to the model, rather than just JSON or text. 🌠 platform.openai.com/docs/guides/fu…

@kevinlu625 @budapp Fantastic.

@code 👏🏻👏🏻👏🏻

🚀 v1.104 of @code is here! Check out what's new: 🤖 Improved coding agent integration 📄 AGENTS.md file support for better context 🔍 New Auto mode (Preview) for smart model selection 🔑 Model flexibility via BYOK extension API …and more: aka.ms/VSCodeRelease Here are the highlights 🧵

@Alibaba_Qwen 👏🏻👏🏻👏🏻😍😍🏳️‍⚧️

Qwen3-Max-Preview is now live on OpenRouter! 🚀