The hard part after launching a developer tool: first users
A small AI job needed a reliable way to pull clean data from resume PDFs. Asking an LLM to return JSON did not work well in real use. The field names changed between runs, the output was not checked properly, and broken cases were hard to repeat for .
The same kind of document problem also appears with invoices, contracts, KYC documents, and other files that need clean . A small developer tool was built over the last few months to let teams define the exact JSON shape they want first, then check the result before returning it. The tool launched this week with no users and no revenue yet.
The main open question is how a new developer tool or API can get its first few users, and whether a or a works better.
Key points
- Resume was unreliable when handled by an LLM alone.
- JSON output changed between calls and was hard to validate.
- The same problem can affect invoices, contracts, KYC documents, and other business files.
- The new tool checks output against a user-defined JSON shape before returning it.
- The founder is looking for practical ways to get the first users for a developer tool or API.