Xander Dunn
@xanderai
Followers
2K
Following
8K
Media
279
Statuses
1K
Building AI scientists @periodiclabs Previously: Building LLMs for code gen, Deep RL for industrial robotics, @Apple @UChicago
Joined February 2014
The team @periodiclabs is absolutely cracked. Favorite team I've ever worked with. We're building AI scientists, and our progress is breakneck - already in the physical lab! The world doesn't need another social media app. We're turning bits into atoms.
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
20
48
850
"There are no winners on the losing side." One big cultural difference I've observed working at big tech for several years vs being at @periodiclabs is the shift from "individual impact / promo" to "we all win together". Highly recommend folks in big techs to consider
9
21
260
Do not listen to these people: "My favorite language is Rust, we should use Rust." "I hate Rust, we shouldn't use Rust." DO listen to this person: "My favorite language is Rust, we shouldn't use Rust."
0
0
4
Left: What the LLM wrote. Right: What it should've written. You almost always want a super simple happy path that crashes when the world is broken rather than outrageously elaborate attempts to never ever crash with god-knows-what consequences. It's slop.
1
0
8
Silicon Valley License Plates Part 2 We've got some chronically online individuals here. The Cybertronk driver is a Windows user š¤
1
0
2
Unfortunately I can't put much stock in this stat, devs approve/merge a llooootttt of bad code. Very soon we will have coding agents that are post trained specifically on the task of de-LLM-slopifying code bases.
0
0
7
Puts away the LLMs for a few hours and writes code that is 1 order of magnitude fewer lines of code and 2 orders of magnitude more performant than anything any of the LLMs could come up with.
0
0
7
@LiamFedus @ekindogus @periodiclabs We're not going to reason our way to Dyson spheres. Excited for you and all the best to the team!
12
23
669
@LiamFedus @ekindogus @periodiclabs This is the most exciting mission Iāve seen for a new company for a long time
4
3
94
.@slaterstich's @BainCapVC article on Periodic is incredibly well written.
Very excited (millions of eV) to be working with @LiamFedus and @ekindogus and the rest of the @periodiclabs team. In the long arc of scientific history, there are a few key moments when things really accelerated ā and Periodic is working on the next one. Below are some
1
15
122
"One of our goals is to discover superconductors that work at higher temperatures than today's materials" I'm happy to see @LiamFedus and team working toward breakthroughs like these! With AI starting to meaningfully contribute to scientific discover, I think it's the right time
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
15
33
619
Can't understate the talent level I'm surrounded by at @PeriodicLabs The team is largely a mix of: - researchers from OpenAI, GDM, FAIR/MSL - PhDs from Stanford, MIT, Harvard, etc - leaders in open-source ML - world-class physics professors Humbled to learn from them every day.
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
13
18
498
Exciting mission with a great team! With the progress of AI, now is the right time to start approaching these problems!
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
2
8
125
congrats, liam!
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
131
95
2K
Excited to finally announce what I've been up to since leaving OpenAI: autonomous science at @PeriodicLabs Why? Because I Take the Bitter Lesson Seriously To accelerate AI, we must enable it to hill-climb compute & energy through experimentally verifiable science š§µ
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
77
62
1K
Join us: https://t.co/Zdr4bGXkZp Frontier compute. Frontier models. Frontier talent. Frontier funding. DM me directly if you're interested in building infrastructure for frontier GPU clusters!
jobs.ashbyhq.com
Periodic Labs Jobs
2
2
19
Your longevity routine isn't working until chatGPT labels you Gen Z. š¤ Also great how I'm both "traditional" and "outside the box" š¤
0
0
0
If the new CEO / head of @github is looking for some low hanging fruit to make their highest paying customers happy, this is it: https://t.co/YUjkFvHolt
0
0
0
"gpt-5-codex high" on codex cli is great. Much faster than codex web. Only issue, sometimes on Auto approval it will still ask me to manually approve every code edit. Super annoying. Curiously, I've seen it output a couple Chinese tokens in its English summaries!
0
0
2