ISSUE 01
LIVENext.jsTypeScriptSupabaseStripe
CONTEXT
By 2026 the first readers of any resume are LLMs: ATS parsers, screening agents, summarize-this-candidate copilots.Most resume tools still optimize for the human reader and treat the AI layer as an afterthought.
BUILD
A multi-tenant Next.js web app.You add your career once; for every job you paste, a staged LLM pipeline weights each requirement (must-have vs nice-to-have, plus knockout checks), scores your honest fit, then rewrites your bullets from your real experience in a chosen tone.Nothing is fabricated: grounding is enforced in code with a post-generation detector, and refusal is a first-class output when a rewrite would need a fact you never gave.A skeptical-recruiter critic then reads the tailored resume bullet by bullet and flags what to cut, swap, or reword.Built on Supabase/Postgres, NextAuth, Stripe, Zod, and live SSE streaming, with an ATS-clean single-column Typst PDF export.
RESULT
Live in production at secondround.app.Building is free; Pro unlocks PDF and plain-text export.
LINKS