Researcher studying nonequilibrium thermodynamics, info theory, origin of life, complexity. Currently at U Pompeu Fabra in Barcelona. @artemyte.bsky.social

Joined July 2014
Artemy Kolchinsky retweeted
Anthropic’s internal research roadmap from July 2021 came out through the Bartz copyright case. Jared Kaplan wrote it, with comments from Dario, Tom Brown, Amanda Askell, Chris Olah and others. The long-term goal: an AI assistant that writes functions, finds bugs, designs APIs, and eventually works like a software engineer colleague. This was over a year before ChatGPT, and the whole roadmap is built around that goal: 1T+ models, RL, TPUs and Trainium, data from code, diffs and PRs, scaling laws, test-time compute, long context, memory, and RL for tool use (in 2021!). Five years later the goal is basically here, but what the comments worried about (sandbox safety, alignment keeping up with capabilities) still isn’t fully solved in 2026.
2
1
2
7
1,412
Artemy Kolchinsky retweeted
Come and study technical AI alignment, with an emphasis on applied mathematics: New fully funded cohorts! We just announced 3 new Fellowship cohorts taking place in the Bay Area starting October, November, and December. And just 10 days left to apply to our October Intensive in London. One application takes 10 minutes and covers both programs: iliad.ac/apply?r=X
8
4
77
16,819
Artemy Kolchinsky retweeted
How do living organisms use information to navigate up sensory gradients? New preprint from @kevinschen13, with Damon Clark and @EmonetLab, shows that navigation works best with a feedback loop: information flows both forward from signal to action, and back from action to signal.
2
13
76
3,529
What is the fundamental cost of stochastic oscillators? In a recent PRE, I provide an elementary derivation of a fundamental "dissipation-coherence bound". Builds on work on thermodynamics of limit-cycle oscillators by Santolin-Falasco and Nagayama-Ito journals.aps.org/pre/abstrac…
2
16
90
5,217
Artemy Kolchinsky retweeted
Only 4 days remaining to apply! Takes 5-10 minutes and one application covers both programs. September 2026 Iliad Intensive A 4-week taught introduction to technical AI alignment, emphasizing applied mathematics and broad coverage of the field Location: LISA, London, UK Duration: Sept 7—Oct 2 inclusive Financial Support: $5,000 (USD) travel-and-housing allowance Application Deadline: July 27 EoD AoE, apply here: iliad.ac/apply?r=direct) Fall 2026 Iliad Fellowship A 3-month mentored AI safety research fellowship, with a focus on applied mathematics Location: LISA, London, UK Duration: Sept 7—Dec 4 inclusive Financial Support: $6,000 (USD) monthly travel-and-housing allowance Application Deadline: July 27 EoD AoE, apply here: iliad.ac/apply?r=direct
The hardest problems in AI safety are mathematical. Come work on them. Fully funded. Application link in the comments.
5
54
5,489
Artemy Kolchinsky retweeted
Request for Proposals — we're soliciting work in two directions: raising the epistemic bar of AI safety, and theory-driven research on ReLU networks. Mathematicians especially welcome. docs.google.com/forms/d/e/1F…
1
1
7
630
Artemy Kolchinsky retweeted
The hardest problems in AI safety are mathematical. Come work on them. Fully funded. Application link in the comments.
8
36
6
375
62,570
Open postdoc position. Interested in complex systems, dynamical systems, and computational mechanics? With Alexandra Jurgens at INRIA (Bordeaux) recrutement.inria.fr/public/…
1
8
20
4,195
I just learned Peter Schuster passed away recently. Peter Schuster was a theoretician who did incredible work on the origin of life, evolutionary dynamics, genotype-phenotype maps, and other topics in theoretical biology. A huge loss.
3
4
1
31
2,375
Artemy Kolchinsky retweeted
R.I.P. Peter Schuster 🪦 Your wonderful contributions to science will live on... 🧬 chemie.univie.ac.at/wissensa…
1
1
170
Artemy Kolchinsky retweeted
You're not even allowed to ask Fable about basic biology questions, let alone anything that could potentially be dangerous.
388
436
242
11,360
1,086,748
Artemy Kolchinsky retweeted
I'm so very sad to hear of the passing of Inman Harvey. He was my PhD examiner in 1999 (he passed it!), and a mentor, inspiration and friend not just to me but to a whole generation (and more) of Artificial Life researchers. He will be very sorely missed, but his legacy lives on.
2
4
662
Artemy Kolchinsky retweeted
D3の永山くんの新作です。Artemyさんとの共著です。 Ornstein–Uhlenbeck過程における、散逸と振動の(dissipation-coherence)トレードオフおよび、行列の非正規性に基づいた緩和の加速の双方を一度に表現可能な、振動-非正規性に基づく定常エントロピー生成率の新分解法です。 arxiv.org/abs/2606.07263
1
12
61
8,507
Artemy Kolchinsky retweeted
Our new work on arXiv🔥 We introduce the geometric complexity for reset maps and uncover a dynamics-independent trade-off relation between complexity and error, which can be regarded as a generalized geometric third law of thermodynamics. arxiv.org/abs/2604.27858
2
36
178
13,495
How does phase separation, a key mechanism of equilibrium self-organization, interact with nonlinear nonequilibrium dynamics? In our paper with Jonathan Bauermann and Giacomo Bartolucci, we look at phase separation in chemical oscillators journals.aps.org/prl/abstrac…
1
24
135
7,075
We find that phase separation can slow down, speed up, damped or amplify oscillations, depending on alignment b/w interaction energies & nonlinear kinetics. In spatial systems with phase separation + reaction + diffusion, rich traveling patterns of concentration & phase emerge
1
1
7
449
This was a very fun collaboration with Giacomo and Jonathan, and my first chance to dive deep into the thermodynamics and dynamics of phase separation.
1
6
337
Open postdoc with the amazing Sarah Marzen (Scripps College, Claremont, California). Intersection of information theory, biophysics, neuroscience, and machine learning. Please RT theclaremontcolleges.wd1.myw…
1
3
687