I let an open model delete 1,242 of my saved articles. On purpose. I feel lighter.

Description and use

Unpile is a local agent that automatically triages a read‑later backlog, deleting almost all items and keeping only those the user will actually read, freeing time and reducing guilt.

Problem

A large read‑later backlog (1,247 items) creates guilt and wasted time because most items are never read.

Outcome

Unpile removed 1,242 of 1,247 saved articles, keeping only the ones the author would actually read, and produced a calm two‑line digest, freeing 5.5 hours of reading time.

Workflow

• Export read‑later items to a CSV or similar format
• Run Unpile locally on the laptop with Wi‑Fi off
• Unpile learns the author’s reading history with TabPFN‑v2
• For each item it outputs a keep/skim/drop decision with probability and reason
• It writes a two‑line morning digest of keepers
• It silences all other items so the author wakes up to a concise list

Evidence

Unpile is a local triage agent for your read‑later backlog.

Source

It reads 14 saved items, keeps the 5 worth my morning, flags 1 maybe, and silences 8, handing back 5.5 hours of reading time I was lying to myself about.

Source

Unpile is a 100% local agent that triages your read‑later backlog with TabPFN‑v2 and Gemma 3, keeping only the few articles worth your morning.

Source