Arnaud Benard (@arnaudai) header
Arnaud Benard (@arnaudai) profile picture

working on @stitchbygoogle. coding agent, harness, llms. acquired by google.

🌁
Joined February 2021
Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)Photo by Arnaud Benard (@arnaudai)
We're hiring research scientists at Stitch. Stitch is building the most talent-dense research team. If you've built your own agent harness for fun or have run tens to hundreds of agents in parallel to solve coding tasks, you will feel right at home here. We're working at the bleeding edge of agent harnesses, generative media and creativity. Apply here: google.com/about/careers/app…
4
8
2
55
10,830
Me reviewing agent traces
Léa Seydoux tries some of the finest French and British delicacies for ELLE UK.
1
542
Carbon is part of my daily workflow and is a great coding model.
Introducing Gemini 4 Carbon, it's a big jump over Argon which we released just 2 weeks ago. Gemini 4 Carbon is SOTA on cyber, long horizon tasks, ranks #1 on Artificial Analysis V5, significantly improves on usability in domains like coding, and is more aligned.
2
11
707
Arnaud Benard retweeted
switzerland in shambles
ETH is false. Still wrapping my head around this one. We’ve spent decades trying to understand the limits of computation. Seeing AI challenge some of those fundamental assumptions is pretty wild! github.com/openai/math/blob/…
4
3
1
126
7,342
We’re going to see a lot of teams rebuilding their codebase from scratch to be agent native.
Talked to multiple founders yesterday who have been building their products over the last 3 years. Started manual code. They feel behind and slower than companies that started this year with all code written by agents. I have never seen such an order of magnitude difference.
3
2
25
2,606
Arnaud Benard retweeted
evals are the cornerstone for both selling data to frontier labs and building your own self-improving models. building high-end evals is extremely valuable and important
25
22
3
332
16,846
Sad to see France struggling like this. Meaningful reforms seem increasingly difficult, politicians can’t seem to agree on much and everyday life feels harder for many. The growing concerns around education and inequality are hard to ignore. It’s difficult to see how the country breaks out of this cycle.
🇫🇷 Footage from central Paris today amid intense rioting across France.
8
1,128
My favorite interview question is: "Explain to me the AI workflows you're using on a weekly basis". I can easily spot who are AI power users and have the drive to experiment.
🗞️ if your team is too busy doing their 'normal job' to experiment with AI, you're preparing them to be replaced ghuntley.com/replaced/ In this post, I guess this's the last time I'll say it. AI use is no longer optional if you wish to be employed. What follows is advice for current employees and hiring managers on how to hire in 2026 to source AI-first candidates.
7
1,038
Arnaud Benard retweeted
We have a Stitch MCP. We have a Stitch SDK. But, we don't have a Stitch CLI. Well, not until now. Introducing the @google/stitch CLI: 🔷 Connect to your local coding agents 🔷 Generate screens and design systems 🔷 Send a local dev server snapshot to Stitch Do it all without leaving the terminal or better yet, ask your favorite harness like @antigravity 😎 Learn more 👇
47
161
44
1,406
375,804
Arnaud Benard retweeted
Today we’re introducing Gemini 4 Argon. It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.
1,726
4,795
2,790
43,746
10,015,745
We’ve selected the top winners of the Google Stitch Challenge (coming soon 👀). Next week, we’re going to break down the leading techniques and patterns to help you #BuildWithStitch. Drop a 🧵 if you participated, and show others what you built! 👇 #StitchChallenge
11
4
54
8,665
The Stitch Challenge is coming! $10K in total prizes. May 27th. Fueled by the good folks @contra
32
72
6
710
58,045
Big day for Stitch, we're streaming first for faster feedback and > 100M design generated!
Your vibe design partner just got more collaborative! To celebrate I/O, here are 5 major upgrades (and one big bonus!): 🌊 Streaming 📄 Start with your existing design ⏩ In-place AI Edits 🎥 Motion on HTML-native Canvas 🚀 More Import / Export Options (.fig, Netlify, Loveable, Bolt) 🌟 BONUS: Import and sync your codebase Rolling out now. Details and product walkthrough video in 🧵
5
649
I want @AcquiredFM to do an episode on Swatch. They single-handedly saved the entire Swiss watchmaking industry in the 80s and now sell budget versions of 6-figure watches. Wild!
A bunch of Swatch boutiques stayed closed and cancelled the Audemars Piguet Royal Pop launch. Lines out of control at boutiques around the world (see Dubai, Milan, Tokyo below). Swatch official accounts posted its “not a limited edition” so that people can chill out. Wild.
2
6
937
Team has been cooking for IO, can't wait to share what we've been up to.
Very excited for I/O next week! 👀 While we count down the days, I wanted to share one of my favorite Stitch prompts that we just made even better: 📝 Copy/paste your messy resume or portfolio data 🖼️ Attach a profile picture 🎨 Ask Stitch: "Make me a resume website in an 80s retro / nature / steampunk style" Stitch builds the site and automatically crops and styles your photo to perfectly match the aesthetic. Export the HTML, drop it on any host, and you have a standout portfolio in seconds. Try it out and show us what you get! 👇
4
1,502
Understandable on a technical level. As a business decision, you would want your model API to support any type of coding harness.
This quoted post is unavailable.
1
4
820
We would love feedback from the community on DESIGN md!
Today, we’re open-sourcing the draft specification for DESIGN.md, so it can be used across any tool or platform. We’re also adding new capabilities. DESIGN.md lets you easily export and import your design rules from project to project. Instead of guessing intent, agents know exactly what a color is for and can even validate their choices against WCAG accessibility rules. Watch David East break down this shared visual language in action👇. New capabilities and links in 🧵
1
1
11
1,722
Arnaud Benard retweeted
Replying to @Steve_Yegge
Much love for your work, Steve. On behalf of @Google, this post doesn't match the state of agentic coding at our company. Over 40K SWEs use agentic coding weekly here. Googlers have access to our own versions of @antigravity, @geminicli, custom models, skills, CLIs and MCPs for our daily work. Orchestrators, agent loops, virtual SWE teams and many other systems are actively available to folks. We'll be writing more about our stacks this year. For personal projects, many engineers actively try the same models and tools the community does - much love to Claude Code, Open Code, Conductor and others - which helps us also learn about opportunities to improve (folks can even use @AnthropicAI's models on Vertex). Many, many of us are here on @X and build our own tools in addition to make sure we're staying sharp. With so many friends working at other frontier labs and start-ups to give ourselves a baseline, Google is anything but average.
37
36
21
1,004
173,943