I let an open model delete 1,242 of my saved articles. On purpose. I feel lighter.
Unpile is a 100% local read-later triage agent. TabPFN decides which saved articles you will actually read, Gemma writes the digest, offline and free.
First seen by Oriel on .
AI summary
Unpile is a local, offline agent that uses TabPFN-v2 and Gemma 3 to triage saved articles, keep likely reads, generate a digest, and silence the rest.
- Unpile accepts Pocket, Instapaper, browser bookmark, or CSV exports and classifies each item as keep, skim, or drop with probabilities and explanations.
- TabPFN-v2 predicts reading likelihood in one CPU forward pass using personal history, while Gemma 3 writes a local two-line digest through Ollama.
- A held-out evaluation reported 0.60 precision, 1.66x lift, 58% silenced, and a 36% raw-pile read rate.
- The system trades recall for a calmer backlog: 44% of items the author would have read were wrongly dropped, adjustable through UNPILE_KEEP_T.
- The project uses open-weight models, a synthetic-but-documented dataset, FastAPI, a single-file UI, and deterministic fallback summaries when Ollama is unavailable.
#tabpfn #gemma #local ai #read-later triage
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Reactions
Total public reactions reported by the source; distinct from points and likes.
- DEV Community: 41, latest observation ; +18 over observed period since
Comments
Comment count reported by the source.
- DEV Community: 45, latest observation ; +31 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| DEV Community | 4904556 | https://dev.to/sarvar_04/i-let-an-open-model-delete-1242-of-my-saved-articles-on-purpose-i-feel-lighter-l7f Read discussion |