5 min read

Does pretraining pay off for single-cell biology?

Does pretraining pay off for single-cell biology?
Nº 01 · The Lede bioRxiv Cell biology · Funding

Simple models beat big pretrained single-cell AI models

Simple models beat big pretrained single-cell AI models
Fig. IbioRxiv · Filed 10 Nov 2026.

Simple models outperform single-cell foundation models on downstream benchmarks, according to a new bioRxiv preprint. Foundation models here are large AI models pretrained broadly on single-cell data and then reused for many tasks, rather than built for any one of them. Downstream benchmarks are shared tests of whether that pretraining helps once a model is put to a specific job. The result bears on the main promise of these models, which is that broad pretraining yields a representation of cells that simpler methods can't match. Before judging any one model, it's worth checking how wide the margin is and which tasks it holds on.

Read the source →

TrialDesignBench grades AI on phase I trial design
Fig. IIarXiv · Filed 10 Nov 2026.
Nº 02 arXiv Field report

TrialDesignBench grades AI on phase I trial design

TrialDesignBench grades language models on phase I trial design, scoring their proposals against established clinical knowledge, according to a new arXiv preprint. A benchmark in this sense is a shared test, a fixed set of problems with reference answers, so different models are compared on equal footing. Phase I is where first-in-human decisions carry the most direct safety stakes. Anyone claiming a model can help draft a trial protocol now has a published standard to measure it against.

Read more →
Stanford and MIT labs run genome-wide studies with Claude
Fig. IIIAnthropic · Filed 10 Nov 2026.
Nº 03 Anthropic Computational biology

Stanford and MIT labs run genome-wide studies with Claude

Stanford and MIT labs built systems on Claude, Anthropic's AI model, that run genome-wide studies in minutes and automate the interpretation of gene clusters, according to an Anthropic research post. The post is the company's own account, so the speed figures come from Anthropic, not from an independent evaluation. Both are working systems built inside university research groups, not product demos put together by the vendor.

Read more →
Also Filed · Three Briefs from the queue
Nº 04 arXiv Field report

SciMARS lets AI research systems improve themselves

SciMARS lets AI agents (software that runs multi-step work on its own) improve their own methods round by round as they carry out closed-loop scientific discovery, according to an arXiv preprint. The name stands for Multi-Agent Recursive Self-Improvement.

Read →
Nº 05 bioRxiv Field report

GeneEmbedLLM predicts gene links with language models

GeneEmbedLLM uses language models to represent gene function as embeddings, lists of numbers in which related genes sit close together, and then uses those representations to predict associations between genes, according to a bioRxiv preprint.

Read →
Nº 06 Anthropic Field report

Anthropic explains how Claude sorts what to remember

Anthropic detailed how Claude decides what to remember before it knows what will be useful, and described a process it calls "dreaming" that reorganizes stored memories and corrects mistakes. The post presents memory as what lets Claude learn from experience.

Read →

Reply with your discoveries. A human reads them. Forward freely.

Agentic Discovery  ·  Nº 139  ·  10 Nov 2026

Editor's Note

Today the big single-cell models lose to simpler ones, and phase I trial design gets its own AI exam.

 

Nº 01 · The Lede  —  bioRxiv  —  Cell biology · Funding

Simple models beat big pretrained single-cell AI models

Simple models beat big pretrained single-cell AI models

Fig. I  bioRxiv · Filed 10 Nov 2026.

Simple models outperform single-cell foundation models on downstream benchmarks, according to a new bioRxiv preprint. Foundation models here are large AI models pretrained broadly on single-cell data and then reused for many tasks, rather than built for any one of them. Downstream benchmarks are shared tests of whether that pretraining helps once a model is put to a specific job. The result bears on the main promise of these models, which is that broad pretraining yields a representation of cells that simpler methods can't match. Before judging any one model, it's worth checking how wide the margin is and which tasks it holds on.

Read the source →

Why it matters

The case for building large pretrained single-cell models rests on beating simpler methods at downstream tasks, and on the benchmarks it ran, this preprint reports the opposite.

 

Nº 02  —  arXiv  —  Field report

TrialDesignBench grades AI on phase I trial design

Fig. II  arXiv · Filed 10 Nov 2026.

TrialDesignBench grades AI on phase I trial design

TrialDesignBench grades language models on phase I trial design, scoring their proposals against established clinical knowledge, according to a new arXiv preprint. A benchmark in this sense is a shared test, a fixed set of problems with reference answers, so different models are compared on equal footing. Phase I is where first-in-human decisions carry the most direct safety stakes. Anyone claiming a model can help draft a trial protocol now has a published standard to measure it against.

Read more →

 

Nº 03  —  Anthropic  —  Computational biology

Stanford and MIT labs run genome-wide studies with Claude

Fig. III  Anthropic · Filed 10 Nov 2026.

Stanford and MIT labs run genome-wide studies with Claude

Stanford and MIT labs built systems on Claude, Anthropic's AI model, that run genome-wide studies in minutes and automate the interpretation of gene clusters, according to an Anthropic research post. The post is the company's own account, so the speed figures come from Anthropic, not from an independent evaluation. Both are working systems built inside university research groups, not product demos put together by the vendor.

Read more →

 

Also Filed  ·  Three Briefs from the queue

Nº 04  —  arXiv  —  Field report

SciMARS lets AI research systems improve themselves

SciMARS lets AI agents (software that runs multi-step work on its own) improve their own methods round by round as they carry out closed-loop scientific discovery, according to an arXiv preprint. The name stands for Multi-Agent Recursive Self-Improvement.

Read →

Nº 05  —  bioRxiv  —  Field report

GeneEmbedLLM predicts gene links with language models

GeneEmbedLLM uses language models to represent gene function as embeddings, lists of numbers in which related genes sit close together, and then uses those representations to predict associations between genes, according to a bioRxiv preprint.

Read →

Nº 06  —  Anthropic  —  Field report

Anthropic explains how Claude sorts what to remember

Anthropic detailed how Claude decides what to remember before it knows what will be useful, and described a process it calls "dreaming" that reorganizes stored memories and corrects mistakes. The post presents memory as what lets Claude learn from experience.

Read →

 

· · ·

Reply with your discoveries. A human reads them. Forward freely.