How to use Cursor to build fantasy football analytics
The author reports a full-stack NFL fantasy analytics platform, 58.4% in-range accuracy across 81K fantasy predictions and 4.54 MAE.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Help build [NFL fantasy analytics project] using [available data and infrastructure]. Work on [one bounded task]. Record scope, file paths, verification steps and rollback in [handoff document]. Confirm each claimed change in the code and verify completion before beginning another task.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Incarcer
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports a full-stack NFL fantasy analytics platform, 58.4% in-range accuracy across 81K fantasy predictions and 4.54 MAE.
- Human involvement
- The author checked code presence, audited claimed fixes and required verification before proceeding.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Experimented with AI and developed an NFL analytics project.
- Stored project documentation in Notion and code in GitHub.
- Established formal session handoffs with scope and verification details.
A novice developer described building a full-stack analytics platform for NFL fantasy football projections after starting to experiment with AI. They used a combination of Cursor, Windsurf, VSCode and Antigravity, with Notion holding project documentation and GitHub storing the work. The author said agents recommended the infrastructure piece by piece, rather than the author deliberately choosing an unusual Cloudflare stack.
The workflow relied on persistent documentation. Every session had a formal handoff covering scope, file paths, verification steps and a rollback plan. The author also introduced grep-confirm protocols after finding that agents sometimes claimed changes were complete without corresponding code. An audit found three such items. Explicit boundaries limited scope, and each task had to be verified before the next began.
The resulting platform combined a Next.js frontend, a Hono API, multiple databases and scheduled ingestion and calibration work. Its walk-forward system used seven NFL seasons, attached weekly actuals and updated position-specific bias corrections. The author reported 58.4% in-range accuracy across 81K fantasy predictions and 4.54 MAE, while still describing themselves as unable to write code.
Who it's for: novice builders who can direct tasks, maintain documentation and verify agent changes.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · Incarcer
Frequently asked questions
How did the author prevent agents from losing context?
Each session used a formal handoff specification covering scope, file paths, verification steps and a rollback plan. Notion stored the project documentation.
How were claimed fixes checked?
The author introduced grep-confirm protocols. An audit had found three items marked complete with no corresponding code presence.
More details
Remaining steps
- Added grep-confirm checks after discovering missing claimed fixes.
- Locked task scope and verified each task before starting another.
About this account
Agents recommended infrastructure and helped build the code across multiple development environments.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47146087. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: Incarcer on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Cursor · Windsurf · VSCode · Antigravity · Notion · GitHub · Cloudflare · Neon
Fantasy football · Documentation · Cloudflare
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.