When Machines Design, Prove, and Discover
This issue traces a shared arc: AI systems are moving from assistants to autonomous agents that design molecules, and formal verification is emerging as the trust layer that keeps their output honest.
The Eigenius community
Verified Minds
A weekly brief on the frontier of reliable, agentic AI. We cut through the noise to bring you the latest papers, industry news, and open-source tools across four pillars: autonomous AI agents, neurosymbolic logic, formal verification, and applied life sciences. Sourced, summarized, and strictly hype-free.
What's here
Read the digest — and browse the field.
Newsletter
One issue a week, curated from across the field.
- New papers, tools, and writing — summarized
- Every item linked to its primary source
- Archived here in full, readable without email
Longer-form pieces that stand on their own.
- Deep dives on concepts and open problems
- Perspective and analysis, not just links
- Written to be readable months later
Shorter, timelier notes from the project.
- Project updates and announcements
- Meetups, talks, and community calls
- Quick notes between newsletter issues
Latest issue
From the newsletter.
Every issue is archived on this site and also available as an RSS feed.
Recent writing
Latest articles and posts.
-
Verification Should Be a Spell-Checker
Peer review spends its scarcest resource — expert attention — on the one class of error that needs none of it: checking that citations resolve and statistics match. Those checks should feel like a spell-checker, absent rather than adjudicated — which means changing the artifact so claims carry their own verifiable derivations.
-
Testing Tells You How Hard to Look. Verification Can Make Some Failures Impossible.
A response to Corpas, Guio & Fatumo on agentic genomics: their tiered validation framework calibrates how hard to look for silent failures, but several of the failure modes it catalogs are structural — and typed contracts can make those impossible to express rather than merely improbable.
-
Hello, community site
The Eigenius community site is live: a weekly, sourced digest of new work in neurosymbolic AI, AI-driven science, formal methods, and AI in the life sciences — plus articles, a blog, and a browsable index of the field.
The archive, indexed
Browse the field.
Every issue is drawn from a growing, sourced archive — which doubles as a typed index of the field, the same shape Eigenius is built around. Explore it by research area.