Qwen3-next-80B-A3B 👀 Only 3B active parameters and almost as good (in benchmarks) as Qwen3-235B-A22B and Qwen3-32B.

Gemma 3 270M (Million not Billion) released. I keep a close-eye to small models, and this one is a great win. I've seen some tests and it is clever enough despite its size, but the main purpose is to fine-tune it to do specialized tasks.

Qwen3-coder seems to be great 👀

So many great announcements at Google I/O that is hard to put on a list. I like Gemma 3n. They are decent lightweight models to run offline on smartphones. They have 5B and 8B raw parameters but are comparable with the memory footprint of 2B to 4B models!

Companies should release their AI models/LLM as soon as possible. That way they don't have to compare themselves to Qwen3 or DeepSeek r2 and distill versions. The more they wait, the more embarrassing it will be when they don't compare their model to those two. 😅

Qwen3 is the ultimate GPU-Poor LLM! Qwen3 4B and 8B are good for many use cases. Even Qwen3 0.6B seems coherent. Qwen3-30B-A3 can be run with 4GB VRAM with enough system RAM. I ran it on 4GB VRAM and got 12tok/s!

Good to see smaller LLMs pushing the pareto frontier of size vs performance. Gemma 3 27B by Google DeepMind was released with performance between DeepSeek V3 and R1. OlympicCoder 7B by Hugging Face with better performance than Claude 3.7 in Olympiad-Level Programming Problems.

The debate around AI is polarized: either it's taking our jobs tomorrow or it's just hype. Anyone who's actually used AI for longer than a few minutes knows the truth is far less sensational.

One of the coolest AI use cases I've read about recently is DLSS 4 for NVIDIA GPUs. DLSS stands for Deep Learning Super Sampling. Games use DLSS AI to predict multiple frames and improve image quality/upscaling. That allows lower-spec GPUs to play at higher resolutions/FPS!

Made-up tech news headline #1: Tech CEO Replaces Dev Team with AI, Only to Be Ousted by Ex-Employee's AI-Powered SAAS One Month Later.

Interesting text-to-video model preserving transparency: TransPixar.

Nvidia announced Project Digits, a personal AI supercomputer launching in May for $3K: - Powered by GB10 Grace Blackwell Superchip. - 128GB of unified memory. - Can be linked to run bigger models. It's kind of the Linux version of Mac Mini with 2x memory but 1.5x price.

Made-up tech role #27: Senior Overengineer. Why use one line of code when you can use 100? - Turns simple tasks into complex microservices. - Believes every app needs AI, blockchain, and a custom-built framework. - Thinks “scalability” is more important than functionality.

Made-up tech role #26: Agile Stand-Up Philosopher. Turns 15-minute meetings into existential debates.

It seems that BlueSky is becoming my main social media platform. I seem to find more interesting content here and it is not a coincidence as the feeds are easy to customize.

I sometimes miss the simplicity of early web development. No build tools, node.js, npm... just write code and see it work in the browser. You could inspect the source and understand everything. vue.js at least still makes it possible to write an app without a build system.

2025 is a perfect square: 45² = 2025 This will happen again in 2116. It's also the sum of the cubes of all single digit numbers: 0³ + 1³ + 2³ + 3³ + 4³ + 5³ + 6³ + 7³ + 8³ + 9³ = 2025.

Happy New Year 2025! 🎆🎇

The Korean institute KAIST found a way to turn cancer cells into normal cells using a computer model. It works for colon cancer, could work for others, and may reduce side effects and recurrence. The technology has been transferred to BioRevert Inc. for practical development.

SemiKong LLM: First open-source LLM for semiconductors, built on Llama3.1. Key takeaways: - Fine-tuned with semiconductor specific datasets. - Collaborative and scalable for diverse use cases. - Attempts to address the industry challenge of knowledge gaps when experts retire.

Meta introduced Byte Latent Transformer (BLT), key takeaways: - Processes raw bytes, no tokenization - Dynamic patching reduces compute costs by 50% - Excels in multilingual, noisy, and low-resource tasks - Extends to image/audio applications Hopefully we see BLT models soon!

Today I learned about the GPU-poor LLM arena. It's interesting as the models ranked there usually fit inside 4GB VRAM or less and run relatively well without GPU as well. EXAONE from LG is on the lead. I tested the 3B model before and it did well for its size. Here's the link:

ghostty 1.0 has been released! It is a blazingly fast terminal emulator written by the legendary Mitchell Hashimoto in the Zig programming language. Check it out: ghostty.org

DeepSeek v3 LLM is out! According to benchmarks it is as good as GPT-4o/ Claude 3.5 Sonnet, but open source. Here's a summary:

Made-up tech role #25: Scrum Claus. 🎷 Making a list, checking it twice, gonna find out who's naughty and nice. - Runs daily stand-ups where everyone pretends they’re on track. - Believes every sprint should end with a Christmas miracle.

Merry Christmas! 🎄

Made-up tech role #24: Silent Night SysAdmin. Ensuring all is calm and all is bright (on the servers). - Works tirelessly to prevent downtime during the holiday rush. - Monitors logs while sipping eggnog and pretending everything is fine. - Secretly wishes for a silent night without pager alerts.

Made-up tech role #23: DevOps Daredevil. Deploying code to production on Friday afternoons, just for the thrill of it.

As attention centers on OpenAI and Google AI models, an attention-free model has been released! Pun intended 🤓 RWKV-7 World 0.1B debuts with impressive long context capabilities. Check the report below 👇 🧵

OpenAI's new language model, o3, has achieved breakthrough performance on the ARC-AGI benchmark, scoring 87.5% on the Semi-Private Evaluation set. Here's the full report: 👇🧵

Someone should trademark o7 before it's too late. 🫡

gemini-2.0-flash-thinking-exp-1219 released. Google's version of the OpenAI o1 model. 12 days of GoogleAI. 🤓

Do you remember when GPT-4 was the absolute best? We have come so far in a short period of time...

It happened folks. Free GitHub Copilot. 🧑‍💻

Made-up tech role #22: Cloud Cartographer. Mapping the ever-shifting landscape of cloud services, one API at a time.

Improving smaller LLMs by giving them more "time to think" similar to o1 from OpenAI. With longer thinking, Llama 1B outperforms Llama 8B and Llama 8B outperforms Llama 70B. It's great to see smaller models punching above their weight!

Google is on a roll lately! Gemini 2.0 flash model closed the gap to Claude 3.5 Sonnet while being a small model and super fast. Gemini-exp-1206 similarly good. Notebook-lm improvements. And now great video generation with #Veo2? 🤯

Made-up tech role #21: Chief Procrastination Enabler. Encouraging employees to "take a break" until nothing gets done.

Linux 6.6.66 LTS released. Kernel of the beast to please the Linux daemons. www.phoronix.com/news/Linux-6...

95% AI generated anime HinaHima is set to release in 2025. www.animesenpai.net/the-first-an...

Made-up tech role #20: Chief Overthinker of Simple Problems Turning "restart the router" into a 3-hour brainstorming session.

Developers, techies, what are your favorite knowledge management/note taking apps?

Nice to see many UK/EU-based developers in BlueSky! 🇬🇧🇪🇺

If you like writing/using CLI tools (command line interface), check out charm.​sh They are building useful CLI tools in golang.

He used to date a model, but...

You wanted to upgrade your software version huh? 😅

Internet discovery path #1: Ever start with one search and end up somewhere only loosely related? These bullet points show the sequence: - Smallest compilers/interpreters - Micropython - Punyforth - Forth (70's programming language) - Starflight (game written in Forth) - Mass Effect

Void-linux seems to be a good distro. I gave it a quick try and I like the scripts it provides and the package manager xbps. Void also supports musl which is good to see! It's hard to find distros that have a musl variant.

LG launched three open-source EXAONE-3.5 LLMs (2.4B, 7.8B, 32B). They claim to be significantly better than Qwen 2.5 and with large context capabilities. I tried the 2.4B model, and it's quite good considering its size. Soon we'll have LLMs running in our fridges! 🤖

Livebech, a more objective LLM benchmark, has been updated! 📊 Gemini-exp-1206 is almost at the top of the list only behind o1, well done Google! Claude 3.5 Sonnet is still the coding king. Llama 3.3 70b is placed between Claude3 opus and gpt-4 turbo.