The irritation
Three weeks before the 2024 NFL season, I decided to test whether leagues would pay for a personalized ESPN-style recap full of their own matchups, standings, storylines, and trash talk.
A retired SaaS product that turned Sleeper league data into weekly AI-written fantasy-football newsletters, emails, and PDFs.

Built because…
Three weeks before the 2024 NFL season, I decided to test whether leagues would pay for a personalized ESPN-style recap full of their own matchups, standings, storylines, and trash talk.
Commissioners pasted a Sleeper league ID; Django and Python pulled rosters, scores, standings, and player data; OpenAI generated configurable recaps, rankings, projections, and roast pieces; then the system delivered email newsletters and PDFs on a schedule.
The promised automation became weekly editorial labor: correcting stat errors and hallucinations, changing prompts, writing new article scripts, and trying not to break older formats. The Sleeper-specific data model also made ESPN and Yahoo expansion expensive.
Under the casing
Scope of work
No victory-lap math
League Lingo shipped on time and had a handful of paying users, but the quality did not justify the original price or the maintenance burden. I reduced the price, stopped the service around Week 9, and published the failure honestly.
AI does not remove the editorial product. Accuracy, variety, source structure, and scalable data modeling mattered more than the novelty of generated copy.
In the playable workshop, League Lingo lives in the Archive Yard as a fantasy pennant marker. You can inspect it without unlocking anything.