Solo dev ships reright: agent-text humanizer that corrupts data in only 1% of cases vs 30-46% for rivals
paulofilip3 · reddit · 2026-11-07
Developer paulofilip3 built reright, a filter step agents call before sending any text, to fix two pain points: agent-written copy sounds robotic, and humanizers often silently change names, numbers, or meaning — a wrong amount in a client email is worse than stiff tone.
How it works:
- The agent passes its draft through reright and gets back a version matching your voice
- reright verifies numbers, names, links, and code still match, plus language and length; if a check fails, the agent gets its original back and compares meaning itself, keeping the original if anything shifted
- Runs on Claude Code (MCP server + a hook blocking unfiltered commits, PR text, and Gmail sends) and on Claude/ChatGPT via custom connector; it learns your voice from your corrections
Self-tested over three months on held-out AI Humanizer Benchmark texts: reright changed a number, name, or quote in 1% of texts vs 30%-46% for nine other tools. It ranks 3rd of 10 on open AI detectors. Free plan: 3,000 words/month.
Related event: Indie Dev Launches reright, a Low-Distortion Humanizer for Agent Text(3 posts)→
More from coding & agent
- Developer Relays Messages Between Claude Sessions Across Machines via Claude Artifacts — PawelHuryn · 2026-11-07
- New Favorite Benchmark: Can AI Debug Your Z-Wave-Saturated Smart Home? — DynamicWebPaige · 2026-11-07
- Veil 2.0 released: verify distributed protocols in Lean 4 with 2–20x faster model checking — tianyin_xu · 2026-11-07
- Indie dev: desktop apps have $0 operating costs and I'm building more of them than web apps — CtrlAltDwayne · 2026-11-07
- Dev mocks apps that shipped 'garbage for a decade' until LLMs could code — ludwigABAP · 2026-11-07
- Swarms launches open-source Skills to connect coding agents to its marketplace — KyeGomezB · 2026-11-07