Personal Intelligence Digest
A daily agentic pipeline that generates its own queries from a knowledge base and delivers to Telegram.
Why there is no measurement
No measurement, and the repo is not public yet
No measurement
Why not
Nothing here is measured. The obvious metric would be whether the generated queries beat a hand-written keyword list, and that comparison has not been run — there is no held-out set and no rubric for what a good digest item is. The repository is also still private, so a reader could not check a number even if I produced one. This is an honest small project, and inflating it would cost more than leaving it plain.
Problem
A news digest built on a fixed list of keywords tells you about the things you already knew to ask for. What I wanted was the opposite: a digest whose questions move as my interests move, without me maintaining a query list by hand.
Approach
Rather than a static feed list, the pipeline generates its search queries from an accumulated knowledge base, so what it asks about each day is derived from what it has already collected. Results are filtered and summarised, then pushed to Telegram on a schedule. It is deliberately one small module — the interesting part is the query-generation loop, not the plumbing.
The idea
Most personal news automation is a keyword list with a scheduler attached. It works, and it has one structural flaw: it can only surface what you already thought to name. The list ages, you stop trimming it, and eventually it delivers yesterday’s interests reliably.
This inverts the dependency. The queries are generated from a knowledge base that the pipeline itself builds up, so the questions asked on any given day are a function of what has already accumulated rather than of a list I last edited months ago.
Current state
It is one module of Python on a schedule, delivering to Telegram. It has been useful to me and it has not been evaluated. Those are both true, and the second one is why there is no number at the top of this page.
Open-sourcing is in progress; the repository link will appear here once it is public and I have audited the claims on this page against the code.