A cheeky little article I wrote on why we're building Nanocoder :) nanocollective.org/blog/why-were-…

Claude Sonnet 4.5 didn't do as well as I expected one shotting a landing page in comparison to Kimi K2.5 - only after 10 mins of back and fourth did we get somewhere

Working on sexy new git tools for Nanocoder. Makes git operations intelligent, simple and intuitive.

We went open source with Whisper • AI last week - a huge step. We want Whisper to be a community project where we can build private AI for everyone and open sourcing what we have is a cornerstone to this!

When the coding agent you started is good enough to be your daily driver. Open source made this happen.

A very cool release of Nanocoder. Lot's of improvements all around 🥳

Kimi K2.5 mid-creation of 5x6 chess. Fast. Efficient. Video out soon.

Kimi K2.5 one shot this chess game. I gave it two more prompts after this, one to amend colours and one to add a full AI opponent. Which it did easily. Mad. Full video soon.

Thanks Kimi.ai - love how keen you are to help with coding 😳

I recently gave GLM-5 a spin on their coding plan. It was very slow creating a few section landing page in about 20 minutes. It got me thinking about the efficiency gain though. Even just a few years ago, creating a landing page of the calibre GLM-5 did would have taken hours

npm auditing is down. no status. 500 error.

Fine-tuning tiny models is one of our strategies to bring private, local AI up alongside the models that come from big tech. The output of this research will bring direct value to both Whisper • AI and the software being built in Nano Collective. There's so much R&D to do in

Modeled from my very own cat Tiddles. He deserves to be centre-stage 😜

Ben Parry ⚡ Sovereign AI asked last week "What's been the hardest technical problem so far" for Whisper... After the last few days, I'd say how we safely update user's JSON data between App versions is a contender Ended up creating `json-up` - a new NPM package to solve migrating JSON

Bless it's little 0.8b socks. Qwen3.5:0.8b is really cool but, this will forever get it lol

Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI

Very proud of this Nanocoder release - big contributions from all corners of the community and the software is getting reallllyyyy good 🔥

An epic feature going into Nanocoder - sometimes it's the small things in the UI that add that polish!

Very proud of this release. Agent workflows are so powerful...