Bruno Marnette

CEO at @trythoughtly, building AI tools to help us think and write better.

[Bruno Marnette @onurbon]

I recently started a podcast on AI and "taste".

Link: https://t.co/no1WyvCPDu
Also on Spotify and Apple Podcast.

With my co-host Philipp Zahn, we explore how the next generation of AI might develop the ability to judge what is good, choose what matters, and create work worth caring about.

Broader project: https://t.co/6IxM6SmmiQ

[Bruno Marnette @onurbon]

The word "taste" was used 23 times in the latest Lenny's Podcast episode with Andrew Ambrosino from OpenAI. That's about once every 3 minutes.

[Despoina Magka @MarlaMagka]

(🧵) Happy to release AIRS-Bench, a benchmark to test the autonomous machine learning abilities of AI research agents 🤖

AIRS-Bench includes 20 tasks sourced from machine learning papers that assess the autonomous research abilities of LLM agents throughout the full research lifecycle, from hypothesis generation 💡 and implementation 🛠️ to experimentation 🧪 and analysis 📊

Each task is extracted from a paper with a state-of-the-art result and consists of a:
📝 problem description (e.g. text similarity)
🗂️ a dataset (e.g. SICK) and
📏 a metric (e.g. Spearman correlation) to optimise over

The agent is then given a GPU and 24 hours to develop and submit a Python solution that matches or exceeds the paper SOTA 📈

Read on for baseline results and examples of agents surpassing human SOTA ���

🌱We open-source the AIRS-Bench task definitions and evaluation code to accelerate in autonomous scientific research:
💻 GitHub: https://t.co/UXzNXyGdU5
📜 ArXiv: https://t.co/badN0jq0IA
🤗 HF paper: https://t.co/6FIWxF0Bsw
📊 Meta AI website: https://t.co/wcIWLrlYBU

Huge shoutout to the team from Meta FAIR who painstakingly crafted, debugged and inspected every single of these tasks and its runs across more than a dozen of agents @alisia_lupidi, @_tomwithanh, @BhavulGauri, @basselralomari, @albertomariape, Alexis Audran-Reiss, Muna Aghamelu, Nicolas Baldwin, @LuciaCKun, @GagnonAudet, Chee Hau Leow, Sandra Lefdal, Abhinav Moudgil, Saba Nazir, Emanuel Tewolde, Isabel Urrego, @mahnerak, @ishitamed, @EdanToledo and @rybolos, @alex_h_miller, @j_foerst, @yorambac for their leadership and support

[Bruno Marnette @onurbon]

@Hadley Congrats! Well deserved

[Bruno Marnette @onurbon]

I also believe strongly in vibe-writing.

Curious, @thulme, did you just use a mainstream AI tool like Claude or ChatGPT (voice + Canvas) for this use case? Or did you feel the need for a more powerful interface?

Asking because we're building something in that space at https://t.co/3NA8PLeVFT, taking inspirations from Cursor/Windsurf to give writers more control.

[Bruno Marnette @onurbon]

@FlynnDevine website and preview video -> https://t.co/Ip11DRiFmZ

[Bruno Marnette @onurbon]

Excited to unveil my latest project!

Over the last few months, my co-founder @FlynnDevine and I have been researching ways to make AI work better for thinkers and writers who care deeply about quality and intellectual rigor.

There's a lot of free AI chat applications out there, and these can be really helpful as a research tool, but when it comes to writing, a lot of people are underwhelmed by the bland and generic feedback they get.

Then there's a lot of specialised apps helping marketing teams generate large volumes of SEO-optimised blog posts. But these are often just slop generators, and you'd never want to use them to produce any serious piece of writing.

We're exploring a new angle: giving people a single space to think and write, with an AI assistant helping them (or just observing) at every step, to try and really understand the user's taste and intentions.

This is still early days, but we'd love your feedback on the direction we're exploring at @TryThoughtly , so we're launching a first website today with a quick demo video. Please see link in comment to sign up for early access!

[Bruno Marnette @onurbon]

@dabit3 Why considering only 3 when they are at least 50 options out there? ;) See https://t.co/IWXZFc94cz for a fairly exhaustive list (please make a PR if something's missing or inaccurate!)

[Bruno Marnette @onurbon]

Reenacting the dining philosophers problem with @akatov, @SamirTalwar et al.

[TechCrunch @TechCrunch]

Enki Launches Snackable Mobile Learning App For Developers https://t.co/VVmaxVxDlX by @mikebutcher

[Forbes @Forbes]

This app quizzes you on code each day for about five minutes https://t.co/xaWoEeCtgW

[Bruno Marnette @onurbon]

Enki: The 5-minute daily workout for your dev skills https://t.co/PgxY3sO6AG via @BetaList

[Enki @enkidevs]

Coding is boring, unless... https://t.co/YMtLxuzzBg

[Andrew Ng @AndrewYNg]

Employee training can't be a one-time event. It should be always happening. How Baidu does it: http://t.co/UPVnsyyExv

[I Am Devloper @iamdevloper]

Samsung just put out a statement saying that they're also proud to be gay.

[Bruno Marnette @onurbon]

Just registered for hack.summit(), a virtual conference of top developers. Check it out! http://t.co/u741CxKNer

[Bruno Marnette @onurbon]

"How to end the government shutdown: if you hold down Texas and Main at the same time, it automatically reboots" - @ColbertReport, Oct 02

[Dmitri Akatov @akatov]

Check out our entry http://t.co/DEmG4NmdJr to the Clojurecup and vote for it here: http://t.co/3xwphEX8YK. Find foosball tables near you!

[Bruno Marnette @onurbon]

Foosball and Clojure enthusiasts: check out our #clojurecup app http://t.co/KZUphZrPMQ (and vote for it if you like it!)