curatedai.net
Start here Directory Categories Radar Models Leaderboard Guides Prompts Skills Daily Ten Builds Newsletter
Light Dark

The systems behind CuratedAI

Two systems run everything on this site, both built and maintained end to end. This is how each one actually works, stage by stage. For the long-form writeup of both, see /systems.

2
systems, both built and run end to end
~81,500
lines of Python across both
12
deliverables written from one daily curation
266
published days the workforce has produced

AI Shorts Clipper

PRIVATE · PYTHON · ~7,500 LINES

The node marked Shorts & reels in the schematic above is not a person with an editor. It is this: a pipeline that takes a YouTube URL and returns finished vertical clips with burnt-in word-level captions, a written hook, a description and a publish slot. It runs unattended.

8 hooks, one kept3 a day, then it stopstranscriptINkeyword score01model re-rank02cut segments03transcribe04burn captions05hook tournament06render 9:1607quota gate08scheduledOUT
  • The hook is a tournament, not a prompt. Eight candidates are generated, validated, judged, run off against each other and trimmed to twelve words. Any clip that falls back to its own title gets a second pass automatically, because the fallback is the failure case and it should not ship quietly.
  • Publishing is quota arithmetic. The YouTube API allows 10,000 units a day and an upload costs 1,600, so the scheduler posts three, spends 4,800, writes its state and stops. It picks up where it left off the next morning.
  • A y/N prompt is not a safety gate. Anything that rewrites already scheduled or already live videos passes through a separate confirmation that a reflexive return key, a piped yes, or a --yes flag cannot satisfy. Everything reachable from a terminal eventually gets run by accident.
three passes, no shared pickstwelve deliverables61 sourcesINscore01rank x30230-day dedupe03reconcile041 per company05write x1206render cards07validate08briefingOUT

Content Creation Workforce

PRIVATE · PYTHON · ~74,000 LINES

The machine behind the daily briefing. It reads 61 sources every morning, ranks a few hundred stories three separate ways, reconciles the three into one Top 10, and then fans that single curation out into twelve finished deliverables: the briefing itself, a narration script, social copy for five platforms, thumbnail and card artwork. 266 published days come out of it.

  • It ranks the day three times and makes the versions disagree. One pass by raw AI score, one by a composite, one tuned for spread, each forbidden from reusing the first pass’s picks. Thirty distinct stories surface, and the final ten are reconciled out of that, capped at one story per company so a single funding round cannot take the day.
  • Yesterday’s news is the hardest thing to filter. A thirty-day lookup blocks exact URLs and titles, then the same company plus the same key terms, then the same company plus five overlapping words. “Cursor Raises $2.3B” on Monday kills “Cursor $2.3B Valuation” on Thursday, which no exact-match check would ever catch.
  • Nothing ships unvalidated. A final pass checks every deliverable for completeness, duplicate titles and dead links before the day is called done. Twelve outputs generated from one source is twelve chances to publish something malformed at four in the morning.