Pierre Vannier (@pierre_vannier) 's Twitter Profile
Pierre Vannier

@pierre_vannier

CEO @flint_company | Podcast “IA pas que la Data” twtr.to/0EjUj | Software Engineer (Python/AI/LLM)

ID: 394876726

linkhttps://linktr.ee/PierreVannier calendar_today20-10-2011 19:28:22

49,49K Tweet

3,3K Followers

5,5K Following

Rohan Paul (@rohanpaul_ai) 's Twitter Profile Photo

Brilliant paper from Google. Test-Time Diffusion Deep Researcher (TTD-DR) - Conceptualizes research report generation as a diffusion process. The agent rewrites itself every step, so errors fade instead of pile up. TTD-DR builds a living draft, keeps feeding it fresh search

Brilliant paper from Google.

Test-Time Diffusion Deep Researcher (TTD-DR) - Conceptualizes research report generation as a diffusion process.

The agent rewrites itself every step, so errors fade instead of pile up.

TTD-DR builds a living draft, keeps feeding it fresh search
Ivan Fioravanti ᯅ (@ivanfioravanti) 's Twitter Profile Photo

Don't claim Lovable is an EU company; they incorporated in the US for simplicity. They're a US company with R&D in the EU. No criticism of them, but of you for encouraging founders to incorporate outside the EU instead of pushing the EU to reduce bureaucracy. 🤷🏻‍♂️

Pierre Vannier (@pierre_vannier) 's Twitter Profile Photo

Friends, we need to build jammers, masks, whatever will protect us from being watched, monitored and surveilled 24/7. Who’s building that ?

DSPy (@dspyoss) 's Twitter Profile Photo

> multi-agent LLM system deployed in the real world to improve doctor-patient communication > optimized prompts using DSPy SIMBA > even with only 100 annotated examples, we got strong performance across empathy, completeness, and correctness

Pierre Vannier (@pierre_vannier) 's Twitter Profile Photo

Western countries: - decreasing birth rates - genZ don’t feel like working has any sense We’re heading in the wrong direction for mankind sustainability.

Ryan Carson (@ryancarson) 's Twitter Profile Photo

Power tip for all of you who live on the CLI and use AI agents. When you want to summon AI instantly to answer a question or bang out a task *without leaving the command line to start a CLI agent*, just use this command: amp -x "ask your question here" You can also redirect

Power tip for all of you who live on the CLI and use AI agents.

When you want to summon AI instantly to answer a question or bang out a task *without leaving the command line to start a CLI agent*, just use this command:

amp -x "ask your question here"

You can also redirect
Simon Willison (@simonw) 's Twitter Profile Photo

GitHub released Spark yesterday, their extremely well crafted prompt-to-app platform for creating and iterating on React apps with user auth and persistent storage I like it a lot! I reverse engineered it with Spark itself, the details are fascinating simonwillison.net/2025/Jul/24/gi…

Maziyar PANAHI (@maziyarpanahi) 's Twitter Profile Photo

Introducing the OpenMed Oncology NER App on Hugging Face 🔬 Excited to share my beautiful Gradio App specialized in extracting cancer, genetics, and oncology entities! 🚀 Spoiler alert, I used the new Qwen3-Coder-480B-A35B model! 🔥 (thanks AK ❤️) Try it now 👇

Introducing the OpenMed Oncology NER App on <a href="/huggingface/">Hugging Face</a>

🔬 Excited to share my beautiful <a href="/Gradio/">Gradio</a> App specialized in extracting cancer, genetics, and oncology entities! 🚀

Spoiler alert, I used the new Qwen3-Coder-480B-A35B model! 🔥 (thanks <a href="/_akhaliq/">AK</a> ❤️)

Try it now 👇
François Chollet (@fchollet) 's Twitter Profile Photo

Official verification of Qwen3-235b Instruct: it gets 11% on ARC-AGI-1 and 1.3% on ARC-AGI-2 (semi-private sets). These numbers are in line with other SotA base models. Qwen3 stands out by being the cheapest base model we tested to score above 10% on ARC-AGI-1.

Eleanor Berger (@intellectronica) 's Twitter Profile Photo

Got ChatGPT Agent. Tried it. It's a joke. Completely and utterly useless. I gave it the task to research the best restaurants for hummus and falafel in and around Zürich (< 45m travel by public transport from the centre). It worked for 20m. It found 15 restaurant, of which about

Got ChatGPT Agent. Tried it. It's a joke. Completely and utterly useless.

I gave it the task to research the best restaurants for hummus and falafel in and around Zürich (&lt; 45m travel by public transport from the centre). It worked for 20m. It found 15 restaurant, of which about
Pierre Vannier (@pierre_vannier) 's Twitter Profile Photo

Yo UPS, your support is still a damn phone bot in 2025…? I mean… what the actual fuck? I would vibe code an AI agent for free for you for the mental health of humanity.

Pierre Vannier (@pierre_vannier) 's Twitter Profile Photo

lol. Imagine US buying that at large, then backdoor and WWIII 🤣🤣. Have a Chinese ultra efficient hitrobot in your kitchen