The 6th rejection email of the week arrived on a Tuesday afternoon.
Automated sender. Generic subject line. No feedback. No name. Just "we've decided to move forward with other candidates" for a role I'd spent three rounds and a take-home project on.
I'd already built the take-home. I figured I'd build something else.
What I built
Unfortunately is a satirical reverse-interview platform. You pick a company from a growing directory, chat with an AI representative of that company in a live interview, score the employer across six categories, and generate a shareable rejection letter.
The whole product is satire. Company profiles are community-generated simulations, not official employer data. The goal isn't to attack specific companies — it's to flip the power dynamic for a few minutes and make hiring feel slightly less dehumanizing.
The interesting technical problem: AI personas that feel different
The part I spent the most time on was making each company's AI rep feel distinct.
Meta's rep references PSC ratings and org renames. Stripe's rep talks about rigor and transparency. Goldman's rep doesn't apologize for weekend work. The goal was to make the "company representative" feel like a plausible caricature of that company's actual culture, not just a generic chatbot wearing a logo.
This required building per-company dossiers that the AI draws from — risk factors, known archetypes, common community complaints, and potential benefits — so that the AI answers questions in a way that's consistent with the company's real public persona, but filtered through a satirical lens.
Getting the tone right was harder than getting the information right. Too mean and it's just noise. Too neutral and it loses the humor entirely.
What's working
- The concept lands immediately for anyone who has job-hunted recently. No explanation needed.
- The shareable rejection letter format is the main viral hook — people can edit the letter and export it as PNG or PDF before sharing.
- The employer nomination system lets users submit new companies to expand the pool.
What isn't working yet
Honestly:
- The company directory is thin (13 companies). All big tech and finance. Most real-world frustration happens at mid-size and regional employers that aren't in the pool yet.
- Community data is sparse. The platform is in public beta and the numbers are low enough that the live stats on the homepage might actually deter new visitors.
- Some AI-generated rejection letter reasons come out garbled or generic. Quality of the generated content is inconsistent.
- The free tier caps at 3 interviews per day due to compute cost. That's a friction point I haven't solved elegantly yet.
The build
Next.js frontend, Postgres, Vercel deployment. The AI interview engine is the core of it. I've been iterating on the prompt architecture to balance satire with coherence.
Launching on Product Hunt today at unfortunately.life.
Disclosure: I used an AI browser assistant to help coordinate community outreach for this launch. Disclosing that here directly.
Happy to go deep on the persona architecture or the prompt design if anyone's curious.
Top comments (0)