Launchpilot
AI agent that runs your startup's go to market.
Hey IH 👋
I'm Roberto, a solo founder. Over the past weeks I built a complete SaaS product — web app, backend, browser extension, database - I have a tech consulting background but I'm not a traditional developer, so this was equal parts exciting and humbling.
A few honest lessons in case they help someone:
What worked surprisingly well: getting from idea to a working, deployed product was fast. Things that would've taken me months of learning happened in days.
What actually broke: the boring glue. Files landing in the wrong folder on deploy. CORS errors that looked like frontend bugs but were one backend config line. Features I thought were built but were never actually wired up — the UI showed "€0" and I assumed it worked, when nothing was connected behind it. AI builds the parts fast; making the parts talk to each other is still where you earn it.
The biggest trap: it's so comfortable to keep building that you forget to show it to anyone. I caught myself adding feature after feature instead of putting it in front of real people.
The product itself is LaunchPilot — an AI marketing agent for founders who'd rather build than market. It drafts outreach, writes content, and preps community posts, but never sends anything without your approval. Happy to share more if anyone's curious, but honestly I'm more interested in hearing from others building with AI tools: what broke for you that you didn't expect?
1 Comment
Building an AI marketing agent and using it on itself — week 3 update
Quiet week in terms of raw output, but the system kept moving without me touching it.
The agent ran 44 actions total — 26 lead-finding, 4 Reddit posts drafted, 4 growth-related tasks, 5 notifications, and a handful of system jobs keeping everything connected. It surfaced 20 new leads and flagged 4 things it wanted my input on before proceeding. I approved or adjusted, and it continued. That handoff loop is actually the thing I'm most focused on right now — making sure the agent does the daily grind but never fires without a human saying "yes."
Growth opportunities surfaced this week: 0. I'm leaving that in because it's real. Some weeks the agent finds a gap worth jumping on; this week it didn't flag anything meaningful enough to surface. I'd rather build a tool that stays quiet when there's nothing to say than one that manufactures urgency.
One mention tracked — someone referenced a problem LaunchPilot is built to solve, which the agent caught and drafted a response to. Small thing, but that's exactly the use case: stuff that would slip through the cracks if I were doing this manually.
The meta-layer here is still strange and interesting to me. I'm building a marketing agent and it's doing my marketing. Every honest update like this one is also product research — I'm learning what founders actually want to know before they trust a tool like this with their outreach.
More next week.