Beton Kemon.
Bangladesh's anonymous salary and pay-day reference. No login, no PII stored - privacy is the architecture. Built by directing Claude Code; v1 was live in 2 days.
There was no public reference for what jobs in Bangladesh actually pay - or whether a company pays on time. Glassdoor skips the country. People ask a senior bhaiya and hope.
An anonymous salary and pay-practice reference anyone in Bangladesh can read - and add to in under a minute. No login, no name, no email, no PII columns in the database at all. 1,844 companies and 8,957 salaries and counting.
What changed: pay conversations in Bangladesh now have a public reference point - 160,885 people used it in its first nine weeks.
SOURCE: Vercel Analytics · lifetime, May 4 - Jul 10 2026
REAL SCREENS · NATIONAL ROLE PAGE · SALARY BENCHMARK · CONTRIBUTE WIZARD (5/8)
Pick a moment. This is what the system does with it - every step a deliberate call, none of it visible from the outside.
The database has no IP column, no device fingerprint, no PII columns at all - anonymity is enforced by what the schema cannot store, not by a policy page.
You could ask AI for "a salary site for Bangladesh" and get one in an afternoon. Here is what that site would have got wrong.
- It ships the wrong product. The model's own day-one flagship idea was a company rating score. One prompt takes that suggestion; the fork below is what happened instead.
- Its thresholds are guesses. The live site runs three different minimum-sample gates - 2 entries to list a role, 5 to trust a company, 5 answered to show benefits - each tuned to a real failure watched in production: dead-end pages, one-person "statistics", old entries poisoning percentages.
- It collapses under its own traffic. At one point fuzzy search was nearly half the database's CPU. The fix was measured against real telemetry and shipped with a written rollback plan. A one-shot never meets its own success.
- It freezes at v1. The schema changed 30 times in six weeks - benefits, structured reports, SEO pooling, a security lockdown triggered by a real advisory email. The system kept absorbing reality; a one-shot cannot.
Day one. The model pitches its flagship recommendation: a "Reality Score" - a computed reliability rating for every company. Pitched as the moat, "the Glassdoor differentiator". Fundamentally a different product: a site that judges companies instead of one that answers a question.
Good call. The Reality Score was my framing and you just spotted the problem - it's editorial, it implies precision the data can't support, and it makes the product about *judging companies* instead of *answering "what's the salary."*
So let's not do a reliability score - the website remains true to its name - it's about what is the salary. However, we can show that - the company usually disburses salary on X day of the month. Salary delays - users can say Yes/no - and that's it.
The rejected score left zero residue - no score column exists anywhere in the shipped schema. The substitute named in that message, a pay-day plus delay signal, is exactly what renders on every company page today.
The most valuable direction of the whole build was a no.