Skip to content
Longterm Wiki
Index
Personnel·MwZ6VCPEIg·Record

Nelson Elhage @ Anthropic

Verdictunverifiable85%
2 checks · 5/11/2026
Headline unverifiable — 1 high-relevance source unverifiable, 1 high-relevance source partial.

1 → unverifiable; dissent: 1 → partial

Our claim

entire record
Role
Research Scientist
Role Type
career
Start Date
2021
Is Founder
No
Notes
Key contributor to mechanistic interpretability research; co-authored 'Toy Models of Superposition' and other foundational interpretability papers Per nelhage.com, currently on Anthropic pretraining team; previously worked on reverse-engineering large language models (interpretability/circuits).

Source evidence

2 src · 2 checks
unverifiable85%Haiku 4.5 · 5/11/2026

NoteRe-check. Previous verdict: partial. Verdict changed.

partial85%Haiku 4.5 · 5/1/2026

NoteThe source confirms Nelson Elhage works at Anthropic in a research/engineering capacity. However, the source describes his role as 'engineer and researcher' rather than the specific title 'Research Scientist' claimed in the record. The source does not provide a start date, so the 2021 startDate cannot be verified from this text. The person and organization are confirmed, but the specific role title and start date are unverifiable from the provided source.

Case № MwZ6VCPEIgFiled 5/11/2026Confidence 85%