Who's Who of AI

Ethan

171 trust practitioner @ethanrosenthal.com · 6,081 followers
AI business Culture, work & education
Why they matter

Practitioner with public evidence across AI business, Culture, work & education.

AI signals
2
past 30d
Sources
1
distinct domains
Discussions
4
past 30d
Latest signal
17d ago
View every signal from Ethan →
data/ml/ai @ runway in 🗽 formerly spent time in startups, biggish tech, and physics labs www.ethanrosenthal.com
What they're sharing

Articles & links

↻ Ethan reposted
Kyle Cranmer @kylecranmer.bsky.social

Wow, I just found out that our team has been scooped by Anthropic www.anthropic.com/research/yes... See the addendum of the post “How does it feel to be scooped by a machine?” written by my collaborator Lance Dixon (who was under NDA while checking the result)

Claude computes a nine-loop amplitude in N=4 super-Yang-Mills anthropic.com
AI Weekly's analysis →
  • Anthropic's Claude and Song He's CAS team using GPT-6 both reached the nine-loop result in the same week, independently confirming simultaneous AI convergence.
  • The full computation cost $1,000 to $2,000 on 96 CPUs over one week, putting frontier-level theoretical physics within reach of individual researchers.
  • Matt von Hippel, who issued the challenge, concluded many apparently unreachable research goals simply require more compute than experts had previously considered worthwhile.
Read full analysis →
 View on Bluesky →
↻ Ethan reposted
Kyle Cranmer @kylecranmer.bsky.social

Wow, I just found out that our team has been scooped by Anthropic www.anthropic.com/research/yes... See the addendum of the post “How does it feel to be scooped by a machine?” written by my collaborator Lance Dixon (who was under NDA while checking the result)

anthropic.com View on Bluesky →

Nice new DataComp paper for VLMs, especially the part about scaling laws for data mixtures. arxiv.org/abs/2606.28551

DataComp-VLM: Improved Open Datasets for Vision-Language Models arxiv.org
AI Weekly's analysis →
  • DCVLM assembles 160 datasets and 6T multimodal tokens across image-caption pairs, interleaved documents, text-only, and instruction-tuning data as an open VLM testbed.
  • The testbed spans 1B to 8B parameter models and 6.25B to 200B token budgets, evaluated on up to 52 downstream benchmarks across 9 domains.
  • The authors' experiments find data mixing, not filtering, is key to training quality, with instruction-heavy mixtures scaling better than caption-heavy ones.
Read full analysis →
View on Bluesky · ♥ 0 ↻ 0 ↩ 0 · 52d ago
Their own posts

Recent commentary

It’s oddly vindicating(?) that our Agentic Age rewards those who invested in software engineering best practices. Tests, CI/CD, platform APIs, infrastructure as code. They’re all very well-aligned with agent utility, which is maybe not surprising; these are all done in the name of automation

View on Bluesky · ♥ 10 ↻ 1 ↩ 1 · 91d ago

You could retrain the exact same diffusion model on the exact same videos each year. As long as you reception those videos with each year’s SOTA LLM, the diffusion model will get better and better.

View on Bluesky · ♥ 2 ↻ 0 ↩ 1 · 16d ago
Their network

In Ethan's orbit

Center = Ethan. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.

Are you Ethan? Show it.

Add the Who’s Who of AI badge to your site or bio. It links back to this profile.

Listed in AI Weekly's Who's Who of AI

Markdown: [![Listed in AI Weekly's Who's Who of AI](https://aiweekly.co/modules/custom/aiweekly_whoswho/images/whoswho-badge.svg)](https://aiweekly.co/whos-who/person/ethanrosenthal-com)