Observed arrival · 2026-10-21
TimeCapsuleLLM: language models boxed into 1800–1875
An independent research project trains language models on historical texts from 1800–1875, then offers them for public experimentation.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Humanities and historical-text researchers
- Worth noticing
- The site reports $437.09 in total published training costs; the newest listed model used one rented H100 GPU for about 21 hours.
Field notes
The demo presents v2 (London, 1.2B) and v3 mini (English, 500M) as text-continuation models: users provide a passage opening, not a question for a chatbot. The site identifies the training corpus as books, newspapers, letters and legal documents, and notes that newer releases were trained from scratch. Its cost table reports one rented GPU per model and $437.09 across the three releases with published figures.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue