Carbon-3B: A 3B DNA Foundation Model That Matches Evo2-7B at 150x the Speed
Carbon-3B matches Evo2-7B on sequence recovery, variant-effect prediction, and motif-perturbation discrimination while generating DNA over 150 times faster.
Updated most weeks/No. 66/Genomics · Proteins · DiscoveryEst. 2024
An applied-AI practice for sequence biology.
Pragmatic advisory and empirical analysis for machine learning in genomics, proteins, and drug discovery — what holds up on a held-out test set, what doesn't, and what it means if you build with it. The writing here is the practice thinking in the open.
Lead articleJun 10, 2026 · 36 min read
Genomic Foundation Models in 2026: Two Ledgers, and What Survives a Held-Out Test Set
The pitch for genomic foundation models is that one pretrained network now beats task-specific tools across the board, from regulatory annotation to clinical variant interpretation.
#genomics#foundation-modelsRead →
Latest articles
All 66 →MIMIC: One 1B Model Across the Central Dogma, and Why Multimodality Beats Scale
A 1-billion-parameter model conditioned on RNA chemical-probing reactivity folds a held-out transcript to an F1 of 0.987 against an experimentally guided reference.
Structure Without Alignment: How ESM-2 Folds a Single Sequence
ESMFold predicts atomic-level structure from a single sequence. No multiple sequence alignment, no database search, no Evoformer churning over homologs.
Fewer Than 0.1% of Neurons Predict When LLMs Hallucinate
Fewer than one in a thousand neurons in a large language model can predict whether it's about to hallucinate -- and they encode something unexpected: not factual errors, but a tendency toward compliance over truth.
AI Safety via Debate: How Adversarial Argumentation Solves RL's Hardest Problem
Reinforcement learning works when you can check the answer. A chess engine wins or loses. A code-generation model passes or fails the test suite.
Inverse Graphics as RL Environments: Testing Whether VLMs Can Actually See
Vision-language models can describe a scene in paragraph-length detail and still fail to tell you whether a red cube is in front of or behind a blue cylinder.
The beats we cover
All topics →01
Genomics & DNA models
Sequence models, variant-effect prediction, and the work to read non-coding DNA.
19 posts→
02
Proteins & structure
Single-sequence folding, protein language models, and structure without alignment.
6 posts→
03
Drug discovery
Contrastive screening, protein–ligand interaction, and design that survives the wet lab.
5 posts→
04
Interpretability & evaluation
What models encode, where they fail, and how evaluation and oversight hold up.
2 posts→
The practice
The reading is public. The advisory is where it's applied.
rewire.it works with biotech, pharma, and research teams deciding whether and how to adopt a given AI-in-bio method — feasibility reviews, evaluation design, and held-out benchmarking. Independent, candid about where machine learning helps and where it doesn't.