I let our AI agent publish 3 blog posts. They sounded like AI, so I sent them back.

Real task from building TODO for AI. The agent posted 3 comparison articles to DEV on its own. Then I read them: same template, same "we're the best" ending. One line of feedback and it rewrote the live posts itself. No traffic numbers yet, will share when there are.

Transcript

1. I let our agent publish 3 blog posts. They sounded like AI. So I sent them back. Marcell, building TODO for AI 1
2. the ask We had comparison pages on our site. I wanted them on DEV too. Told the agent: take these, adapt them, post them. 2
3. that part just worked Used our DEV account through the API. Posted. Opened every URL. Checked the canonical links point back to us. I approved the first one, said "do two more". Done. 3
4. then I actually read them table. paragraph per tool. "and that's why we're great." same shape, three times. you've read this post a thousand times. 4
5. "how could we make this less AI feel?" that was my whole feedback. 5
6. what changed killed "best" and "strongest". we never tested those tools hands-on, so why say it open with the actual problem the reader has say it straight: these are notes, not benchmarks the free tier limits people actually hit no made-up customer stories 6
7. it fixed the live posts itself read each one first. edited. checked it saved, same URL, canonical still there. pushed the markdown to git. I didn't touch the DEV editor once. 7
8. does it work? no idea yet. 3 posts live. zero data on traffic, rankings, signups. I'll post numbers when there are numbers. next time: review one post before asking for three. 8
9. go look the Lindy post on DEV the original on todofor.ai yes, it's our product. yes, the agent helped with this deck too. 9