Sign in

Axel 👨‍💻 Developer

@axelgarciak.bsky.social
6.1K followers 2.6K following 688 posts

👨‍💻 Software Engineer 💾 Software minimalist/retro 🤖 AI tinkerer 🏗️ Building tech communities 🇪🇺 UK 🇬🇧🇩🇪🇻🇪 | Check bio: axelgarciak.com/bio

PostsRepliesMedia
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 16/04/2026
Hermes Agent + Qwen3.6 MoE will be a great combo Qwen3.6 seems to be the best open weights model (on benchmarks) for its size, better than Qwen3.5 MoE and 27B dense and better than Gemma 4 both dense and MoE. I hope more Qwen3.6 sizes are released soon!
11130
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 05/01/2026
Starting 2026 with smol LLMs for the GPU-poor: - tencent/Youtu-LLM-2B - tiiuae/Falcon-H1R-7B And medium-sized LLM: - miromind-ai/MiroThinker-v1.5-30B Links below:
380
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 31/12/2025
Goodbye to 2025 👋 Happy New Year 2026! Small LLMs highlights in 2025: - Qwen3-4B - LFM2–8B-A1B / LFM2-2.6B-Exp - granite-4.0-h-tiny (7B A1B) - gemma-3n-E4B / gemma-3-4b-it - SmolLM3-3B - Ministral-3-3B They run at decent speeds on the CPU! Looking forward to 2026's smol LLMs!
060
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 11/09/2025
Qwen3-next-80B-A3B 👀 Only 3B active parameters and almost as good (in benchmarks) as Qwen3-235B-A22B and Qwen3-32B.
020
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 16/08/2025
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.
120
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 22/07/2025
Qwen3-coder seems to be great 👀
152
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 20/05/2025
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!
040
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 01/05/2025
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. 😅
260
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 01/05/2025
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!
160
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 12/03/2025
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.
280
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 19/01/2025
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.
6302
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 13/01/2025
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!
3131
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 12/01/2025
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.
2180
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 10/01/2025
Interesting text-to-video model preserving transparency: TransPixar.
0232
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 07/01/2025
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.
2261
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 04/01/2025
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.
1302
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 03/01/2025
Made-up tech role #26: Agile Stand-Up Philosopher. Turns 15-minute meetings into existential debates.
6426
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 02/01/2025
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.
3521
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 01/01/2025
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.
12540
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 01/01/2025
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.
1417
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 01/01/2025
Happy New Year 2025! 🎆🎇
1170
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 30/12/2024
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.
onlinelibrary.wiley.com
Control of Cellular Differentiation Trajectories for Cancer Reversion
A computational framework, single-cell Boolean network inference and control (BENEIN), is presented. Applying BENEIN to human large intestinal single-cell transcriptome, MYB, HDAC2, and FOXA2 are ide...
2488
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 29/12/2024
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.
semikong.ai
SemiKong | The World's First Open-Source LLM for Semiconductor
SemiKong outperforms general-purpose LLMs in semiconductor-specific tasks. An AI Alliance core project. Led by industry & AI specialists from Aitomatic, TEL, FPT
070
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 29/12/2024
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!
1411
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 28/12/2024
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:
3401
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 27/12/2024
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
ghostty.org
Ghostty
Ghostty is a fast, feature-rich, and cross-platform terminal emulator that uses platform-native UI and GPU acceleration.
4405
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 27/12/2024
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:
4403
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 25/12/2024
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.
2260
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 25/12/2024
Merry Christmas! 🎄
1170
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 24/12/2024
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.
2191
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 22/12/2024
Made-up tech role #23: DevOps Daredevil. Deploying code to production on Friday afternoons, just for the thrill of it.
7381
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 21/12/2024
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 👇 🧵
3150
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 21/12/2024
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: 👇🧵
3444
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 21/12/2024
Someone should trademark o7 before it's too late. 🫡
1100
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 19/12/2024
gemini-2.0-flash-thinking-exp-1219 released. Google's version of the OpenAI o1 model. 12 days of GoogleAI. 🤓
2181
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 19/12/2024
Do you remember when GPT-4 was the absolute best? We have come so far in a short period of time...
3170
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 19/12/2024
It happened folks. Free GitHub Copilot. 🧑‍💻
2162
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 19/12/2024
Made-up tech role #22: Cloud Cartographer. Mapping the ever-shifting landscape of cloud services, one API at a time.
2271
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 17/12/2024
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!
1203
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 17/12/2024
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? 🤯
2131
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 16/12/2024
Made-up tech role #21: Chief Procrastination Enabler. Encouraging employees to "take a break" until nothing gets done.
4200
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 16/12/2024
Linux 6.6.66 LTS released. Kernel of the beast to please the Linux daemons. www.phoronix.com/news/Linux-6...
phoronix.com
Linux 6.6.66 LTS Kernel Released With New Hardware Support & Many Fixes
Linux 6.6.66 was released today alongside other updated Long Term Support (LTS) kernel versions.
1476
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 15/12/2024
95% AI generated anime HinaHima is set to release in 2025. www.animesenpai.net/the-first-an...
animesenpai.net
The First Anime Created With 95% AI Use to Release in Spring 2025 - Anime Senpai
The upcoming Twins HinaHima, based on the twin TikTokers (and YouTubers) Himari and Hinana, is being promoted as the first anime to use AI for over 95% of its production. It’s set to release in Spring...
1103
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 15/12/2024
Made-up tech role #20: Chief Overthinker of Simple Problems Turning "restart the router" into a 3-hour brainstorming session.
3291
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 14/12/2024
Developers, techies, what are your favorite knowledge management/note taking apps?
30230
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 14/12/2024
Nice to see many UK/EU-based developers in BlueSky! 🇬🇧🇪🇺
1340
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 14/12/2024
If you like writing/using CLI tools (command line interface), check out charm.​sh They are building useful CLI tools in golang.
4473
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 13/12/2024
He used to date a model, but...
0431
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 13/12/2024
You wanted to upgrade your software version huh? 😅
5570
Axel 👨‍💻 Developer @axelgarciak.bsky.social · 12/12/2024
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
4191