Pierre Alex @pa_schembri

Turn your existing strengths into an unbeatable advantage with AI | Python Expert and AI builder advanced-stack.com Paris Joined August 2015

Tweets

2K
Followers

447
Following

360
Likes

1K

Georgi Gerganov @ggerganov

4 weeks ago

whisper.cpp is coming to ffmpeg github.com/FFmpeg/FFmpeg/…

45 183 2K 176K 530

Pierre Alex @pa_schembri

a month ago

Oops I did it again... Guessing what's next ?

0 0 0 41 0

Download Image

Pierre Alex @pa_schembri

a month ago

Is there such a thing as a K8S / N8N / Redis / React-Frontend (replit-based) / Python-backend / Tilt-enabled boilerplate starter ?

0 0 0 51 0

LFM2 🤝 Outlines Constrained generation is amazing to squeeze more performance from tiny models. We just released a blog post with @remilouf with a smart home control demo. We used LFM2-350M to save notes, close blinds, and handle displays.

4 13 36 3K 13

Download Image

Dev Valladares @dev_valladares

2 months ago

Infinite Wiki ⁕ Every word is a hyperlink. Every description is generated in real-time (in ~1 second) ⁕ Runs on Gemini 2.5 Flash Lite. ASCII diagrams using 2.5 Flash

106 232 3K 252K 2K

Download Video

Pierre Alex @pa_schembri

2 months ago

Interesting to see that one can train a 1 014 parameters model for keyword spotting and making it work. I might have to write an article for this one

0 0 0 36 0

Pierre Alex @pa_schembri

2 months ago

The untold joy of seeing loss decrease with a massive training dataset

0 0 0 26 0

Download Image

Pierre Alex @pa_schembri

2 months ago

So far my first attempts to train a keyword spotter model from scratch have failed. It’s like learning how to bike all over again. You can vibe code a huge SaaS app in a few hours but one can’t vibe train.

0 0 1 37 0

Pierre Alex @pa_schembri

2 months ago

0 0 1 31 0

Download Image

Pierre Alex @pa_schembri

2 months ago

The American mind cannot understand this

0 0 0 30 0

Download Video

Pierre Alex @pa_schembri

2 months ago

Tested the prompt API embedded in @googlechrome today… It supports streaming and structured outputs locally. It’s like having @ggerganov llama.cpp within chrome to build extensions. Can’t stop thinking about use cases 🤯

0 0 0 66 0

Pierre Alex @pa_schembri

2 months ago

I think the next logical step for @realsohamparekh is to work at Meta.

0 0 0 80 2

Download Video

Pierre Alex @pa_schembri

3 months ago

Release dates keep slipping, but code quality isn’t the real issue. My recent CTO chats point to repeatable fixes. Which one would shorten your timeline ?

0 0 0 57 0

Pierre Alex @pa_schembri

3 months ago

So I finally managed to find 1 hour to test on the text-to-lora from @SakanaAILabs Generated my first lora adapter on a remote GPU-equipped notebook. Now downloading it with magic wormhole, but my mac be like "Installing World..."

0 0 1 179 0

Download Image

Pierre Alex @pa_schembri

3 months ago

brew decided to upgrade itself... my mac is downloading the entire internet

0 0 1 37 0

Sakana AI @SakanaAILabs

3 months ago

We’re excited to introduce Text-to-LoRA: a Hypernetwork that generates task-specific LLM adapters (LoRAs) based on a text description of the task. Catch our presentation at #ICML2025! Paper: arxiv.org/abs/2506.06105 Code: github.com/SakanaAI/Text-… Biological systems are capable of…

45 391 2K 398K 1K

Download Video

Pierre Alex @pa_schembri

3 months ago

Prompt → Workflow with Local LLMs What does it take to turn: “When the electricity price is below $0.4/kWh and my EV is plugged in…” into an executable workflow? It’s not trivial—especially for sub-8B models. advanced-stack.com/resources/how-…