Dad of https://xitter.cf/t.co/PvsS2MQczg, PicoGopher, and two amazing kids. Genetically a teacher, forever a student. MLE at https://xitter.cf/t.co/n8sHHWNpFg. He/him
London
Joined April 2008
- Tweets2.4K
- Following471
- Followers451
- Likes5.3K
Davide Eynard retweeted
new Earendil Pi extension, pi-voice, built by CJ Pais (creator of Handy)
npm install -g @earendil-works/pi-voice
npmjs.com/package/@earendil-…
Played with Qwen3.8 Flash Next and @pidotdev on my 64GB Mac, as a result I have another episode of "Vibe reversing" ready 🙃 The crackme was quite simple, but I was surprised to see how well the agent behaved (also ran the exe w Unicorn + wrote a keygen!) pi.dev/session/#d98e0e0db4f7…
Qwen3.8 Flash Next is now supported in DwarfStar, covering 64GB Mac systems very well and with very fast inference of 50~70 t/s and > 1400 t/s prefill. For now this is Metal only. Thanks to @ivanfioravanti for all the cool work in the PR. N-grams on SSD like for DS4.1F.
Davide Eynard retweeted
New post: We need to be deleting more code. 💆♀️
vickiboykis.com/2026/09/01/b…
Davide Eynard retweeted
Come say hi and lets chat open source & AI!
round two, august 25 at 11am ET.
@Marcus_Rein_ hosts Richard Tweed of Tessl.io and @natebrake from @MozillaAI. these events are opportunities for the people actually building with open source to gather and share ideas.
this event is exclusive to ONR subscribers. join the community today: newsletter.ownersnotrenters.…
keep an eye on your inbox for more information.
Davide Eynard retweeted
Your daily reminder that otari is a router that is actually open github.com/mozilla-ai/otari
Weekend project: I started working on a new llamafile release (0.10.6 ) with support for MuseGlimmer:30B and Qwen3.8:27B. Yesterday I uploaded WIP llamafiles in the alpha HF repo huggingface.co/mozilla-ai/ll… in case you want to play with them 🙃 (metal+linux cuda and vulkan only atm)
Davide Eynard retweeted
How do you reduce your reliance on subscription plans and token subsidies if you're not actively measuring how much you benefit from them? Awareness is the crucial first step for change
Davide Eynard retweeted
Over the past week I've used the @poolsideai Laguna S 2.1 model for 54 million tokens of work, almost 1000 requests (all running through @pidotdev and @AgentofEmpires 🚀) Probably saved me about ~$50 of API Costs. (tracked via @MozillaAI's otari)
Right now I'm using it almost exclusively as an orchestrator agent: I work with it to launch and monitor a fleet of coding agents (Opus 4.8). It is something that I used to use Claude code to do, but the Laguna model is completely capable of doing this for the unbeatable price of ~$0.00.
Davide Eynard retweeted
Human coders can't think of all the edge cases, AI coding agents can't forecast all the edge cases: users find the edge cases. Users reporting and helping fix them are what makes open source software still an unstoppable force in 2026.
I love @pidotdev’s trace sharing, and I use it a lot to show things during talks or in blog posts. While having open datasets with traces is super important, I also value sharing with individuals. So I have decided to start publishing them as a feed: aittalam.github.io/traces/fe…
This is my very first experiment, title/descriptions are still quite crappy but I wanted to get a sense of how this could work. I like the idea of following other ppl’s work, seeing examples of new model / tool capabilities, learning new tricks this way.
Davide Eynard retweeted
CJ is good people. transcribe.cpp, ggml based STT engine supporting tons of models and platforms. Love it.
workshop.cjpais.com/projects…
On AI-Assisted Web Search: aittalam.github.io/posts/202…
This is Part 2 of my series on Web Search: after showing how to run an agent with #Pi, #llamafile, and #SearXNG in Part 1, here I talk about:
- running realistic agentic workloads with limited resources
(1/4) 🧵
- why bother? I think AI-assisted search and coding are two different beasts, but with some points in common: here I share some thoughts about this experiment and the possible consequences of having agents run automatic searches for us
(3/4) 🧵
Searching the Web with agents: aittalam.github.io/posts/202…
Or: build your own search agent with @PidevCoding , #SearXNG, and #llamafile
Open, small, custom agents can be quite good at searching. This feels as mindblowing as AI-assisted coding, but (1) useful to more ppl, (2) using cheaper compute, (3) easier to verify (4) providing you with more knowledge than what you started with, instead of atrophying yours.
ALT A handdrawn image of a man sitting on a couch in front of a TV showing white noise. The couch and the man are seen from behind and some smoke is coming out of the man's head. The man looks collapsed and a can has fallen from his right hand on the floor. The TV static does not look hand-drawn. On the top-left corner, two numbers: 120-153. On the bottom right, two more: 201-234. Image taken from Ovid's riddle on http://3564020356.org/ , September 2001