Tokenta Signal SIGN IN TERMINAL >
2026.11.07 13:37 UTC
● ISSUE 256 · MICROSOFT × ORACLE × AI CLOUD · Reuters report of Bloomberg News (Duke FM syndication)● ISSUE 255 · TERAWULF × Q3 2026 × WULF · TeraWulf Q3 2026 results (GlobeNewswire release)● ISSUE 254 · PSI × QUANTUM · PSI Aaronson-Ambainis proof announcement (PSI)● ISSUE 253 · IBM × QUANTUM · IBM Quantum System Two Chicago unveiling (PR Newswire)● ISSUE 252 · OKLO × Q3 2026 · Oklo Q3 2026 Form 10-Q (SEC)● ISSUE 251 · VISTRA × Q3 2026 × VST · Vistra Q3 2026 results (PR Newswire)
● ISSUE 256 · MICROSOFT × ORACLE × AI CLOUD · Reuters report of Bloomberg News (Duke FM syndication)● ISSUE 255 · TERAWULF × Q3 2026 × WULF · TeraWulf Q3 2026 results (GlobeNewswire release)● ISSUE 254 · PSI × QUANTUM · PSI Aaronson-Ambainis proof announcement (PSI)● ISSUE 253 · IBM × QUANTUM · IBM Quantum System Two Chicago unveiling (PR Newswire)● ISSUE 252 · OKLO × Q3 2026 · Oklo Q3 2026 Form 10-Q (SEC)● ISSUE 251 · VISTRA × Q3 2026 × VST · Vistra Q3 2026 results (PR Newswire)
AI FRONTIER · SIGNAL REPORT

Claude Mapped a 96-Protein Vault. The Mystery Waited 40 Years

Anthropic said on November 4 that Claude helped its biology team resolve the first three-dimensional structure of a prokaryotic vault particle: phage protein L05 self-assembles in 96 copies into a hollow nanobarrel, at 3.2 Å cryo-EM resolution. The paper is a preprint, not yet peer-reviewed.

ISSUE 163 · 2026-11-04 · SOURCE / Anthropic research report · STATUS / VERIFIED · TOKENTA SIGNAL RESEARCH
ANTHROPIC × BIOLOGYREPORTED + ANALYSIS
IMAGE: Anthropic research report — FROM THE ORIGINAL REPORT ↗
96
PROTEIN COPIES
Copies in the L05 particle; mammalian vaults have 78.
3.2 Å
CRYO-EM
Cryo-EM resolution.
15 billion
SEQUENCES
Predicted protein sequences searched.
KEY TAKEAWAYS

WHAT HAPPENEDWhat Happened

Anthropic said on November 4 that Claude helped its biology team resolve the first three-dimensional structure of a prokaryotic vault particle. Phage protein L05 self-assembles in 96 copies into a hollow nanobarrel, at 3.2 Å cryo-EM resolution.

Finding it was a needle in a haystack. Claude Science searched about 15 billion predicted protein sequences and found vault homologs in 29 bacterial and 4 archaeal phyla, in under 1% of species with sequenced genomes. Of 44 candidate proteins tested in cell-free expression, 24 were moved into E. coli, 15 produced soluble protein, 7 formed vault-like particles, and cryo-EM confirmed 4.

The structural numbers are clean. The L05 particle is built from 96 identical proteins, 48 per half, about 54 by 45 nanometers, about 9.5 megadaltons. Mammalian vaults have 78 copies; the L05 interior holds about 25% more volume.

FISHING IN 15 BILLIONFishing in 15 Billion Sequences

A search across 15 billion sequences, with a hit rate under 1%. Without Claude's retrieval power, the starting point would not exist.

From 44 candidates down to 4 confirmed, the experimental pipeline itself was partly designed by Claude. Retrieval, purification, modeling — machines took over most of the manual work.

96 VS 7896 vs 78, 25% More Volume

Mammalian vaults have 78 copies; L05 has 96. Shorter and wider, yet with about 25% more interior volume. Doubling the repeat segments could add about 20% more.

The structural difference points to functional possibilities. Anthropic sees such structures as potential miniature containers for delivering enzymes and other molecules, though the vault's natural function remains unknown.

THE MISTAKES COUNTThe Mistakes Count

Purification shrank from a routine 3 days to about 90 minutes, the first complete model came within 34 hours, and the whole project took under 4 months. The speed is real.

Claude made mistakes too. One media-additive suggestion delayed the project by two weeks, and one cryo-EM processing run produced a mirrored image. The paper is a preprint, not peer-reviewed, and the particle was only confirmed to form in E. coli.

THE SHIFTWhat Changed

FROM UNKNOWN TO MAPPED
From 40 years unknown to a 96-copy structure
The first prokaryotic vault structure in 3D, found in under 4 months of search and experiment.
FROM DAYS TO MINUTES
From 3 days of manual purification to 90 minutes
Purification and modeling sped up; the manual work of research was taken over first.

SIGNAL JUDGMENT — ANALYSISMachines Do the Searching. Humans Judge Where It Can Go Wrong

Tokenta Signal's judgment: this preprint's value is not the vault itself, but the new division of labor it shows for AI in research. Retrieval, purification, and modeling go to the machine; humans judge where it can go wrong.

WHAT REMAINS UNCERTAINWhat Remains Uncertain

The boundaries of this piece: the paper is a preprint, not peer-reviewed and not independently reproduced. The particle was only confirmed to form in E. coli, not in its natural host or during phage infection. Claude made mistakes too — one media-additive suggestion delayed the project by two weeks, and one cryo-EM processing run produced a mirrored image. The identity of the small-molecule pocket inside each L05 copy is unconfirmed; coenzyme A is only a candidate. This does not constitute investment advice.

EVIDENCEEvidence

Anthropic's research report confirms that Claude helped its biology team resolve the first three-dimensional structure of a prokaryotic vault particle: phage protein L05 self-assembles in 96 copies, at 3.2 Å cryo-EM resolution. The paper is a preprint.

ORIGINAL: Investigating vault particles with Claude ↗
anthropic.com · 2026-11-04 · Anthropic research report

PRIMARY SOURCESPrimary Sources

RELATEDRelated Intelligence

CORRECTIONS: if a fact on this page is wrong, we keep a correction record on the page. Contact team@tokenta.space