This happened: www.anthropic.com/news/fable-m...
Who's Who of AI
Vikram Saraph
Why they matter
Directory member with public evidence across AI research, Compute & infrastructure.
- AI signals
- 3 past 30d
- Sources
- 3 distinct domains
- Discussions
- 13 past 30d
- Latest signal
- 13d ago
Software engineer, AI/ML researcher, and mathematician at Johns Hopkins APL.
Former New Englander, current Marylander.
Brown CS PhD and Notre Dame math alum.
Nerd of sorts (computers, math, language, puzzles, games, books, music).
Opinions are my own.
What they're sharing
Statement on the US government directive to suspend access to Fable 5 and Mythos 5 anthropic.com
Opinion | What A.I. Did to My College Class nytimes.com
Leiden Declaration on Artificial Intelligence and Mathematics leidendeclaration.ai
Claude Code Is Steganographically Marking Requests thereallo.dev
The Model You Audit Is Not the Model You Ship techpolicy.press
Flow Matching on General Geometries arxiv.org
Scalable spherical CNNs for scientific applications research.google
Exploring the periodicity of flight patterns arxiv.org
arxiv.org
arxiv.org
reddit.com
View on Bluesky →
The End of Mathematics — Daniel Litt daniellitt.com
View on Bluesky →
Articles & links
leidendeclaration.ai
AI Weekly's analysis
→
- The Leiden Declaration, released June 2, 2026, warns AI threatens proof integrity, attribution, and peer review in mathematics.
- Over 2,654 signatories including Fields Medal winner Terence Tao have endorsed the community-initiated declaration.
- The International Mathematical Union backs the declaration, which makes separate recommendations to researchers, publishers, policymakers, and AI developers.
Read full analysis →
This seems interesting: thereallo.dev/blog/claude-...
Vikram Saraph reposted
Almost every LLM that reaches a broad audience is quantized—compressed to run cheaply. USC's Emilio Ferrara argues this compression is treated as a safety nonevent, when it should be treated as a change to the deployed system. This has policy implications, he says.
AI Weekly's analysis
→
- Emilio Ferrara tested six open models including Qwen, Gemma and Llama across eight languages and found compressed versions volunteered stereotypes in roughly one in four answers.
- Nominal precision labels routinely understate how much a build has actually been compressed by a factor of three to 14, the article reports.
- Ferrara wants the EU AI Act and California SB 53 to require naming the deployed configuration and testing open-ended generation in multiple languages.
Read full analysis →
View on Bluesky →
also how flow matching generalizes to riemannian manifolds. straight-line interpolation generalizes to interpolation along geodesics, and you get those from a riemannian metric? arxiv.org/abs/2302.03660
learned of spherical CNNs while reading about flow matching: research.google/blog/scalabl...
Paper that mentions the p2p topology in contrast with other airlines arxiv.org/html/2606.00...
Reminded today that this paper exists: arxiv.org/pdf/1804.02767
Flow Matching for Generative Modeling arxiv.org/pdf/2210.02747
Vikram Saraph reposted
Another absolutely bonkers math result from a frontier LLM (announced in a tweet, no less): www.reddit.com/r/math/comme...
Vikram Saraph reposted
Yesterday I gave a talk at an OpenAI summit on the future of mathematics. I was asked to talk about the bad future, where humans become mathematically disempowered. Here's a blog post based on the talk: www.daniellitt.com/blog/2026/8/...
Their own posts
Recent commentary
Really good talk by Juliana Castro Varon of @nytimes.com at the Humans of Generative AI workshop at CVPR. All about how they explore usage of AI in journalism to improve user experience, without just using AI for the sake of using AI.
Their network
In Vikram Saraph's orbit
Center = Vikram Saraph. 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.