MeetingFlash
Meetings, Productivity, AI tools
The short version: I built MeetingFlash — a tool that turns raw meeting notes into a complete "execution pack" (decisions, action items with owners, a follow-up email, a Slack message, the next agenda) in under 20 seconds. I'm a solo founder, and I'm not a traditional developer. I shipped the whole thing — Next.js frontend, Supabase auth + database, the Claude AI pipeline, live Stripe payments, a blog, and a full SEO setup — by working closely with AI coding agents.
Why I built it: Every meeting ends with the same dead 20 minutes — writing the recap, chasing who-owns-what, drafting the follow-up. The current AI meeting tools all transcribe (they capture words), but nobody struggles to capture words. They struggle to act on them. So I built the opposite: paste messy notes, get finished, ready-to-send work. No prompts.
How I actually built it (the part IH might care about):
I treated the AI agent like a very fast junior engineer and myself like a product owner. My job became: decide what to build, review what came back, and catch the things that "looked done but weren't."
I kept a single source-of-truth doc the agent reads before every change — architecture decisions, past mistakes, design rules. That one habit removed 80% of the "it rewrote the thing I asked it not to touch" pain.
The hardest bugs weren't code — they were the invisible ones (a sleeping free-tier database returning empty data with no error). Those took real debugging discipline, not just "ask the AI."
What's working: The product is live, polished, and people who try it get the "oh, that's the whole recap, done" moment. Stripe is in production. I just turned on email and a full SEO layer this month.
What's hard (the honest part): Distribution. I have a clean, working product and almost no traffic yet — classic builder trap. I over-invested in making it good and under-invested in making it found. I'm fixing that now: directories, content, and posts like this one.
Three things I learned:
Shipping is no longer the bottleneck — being found is. Build less, distribute more.
A non-coder can ship real software now, but you still need product taste and the discipline to verify, not just trust the output.
Talking to (and writing for) real people beats one more feature, every time.
If you want to see what it makes, you can try it free without an account: www.meetingflash.work
My question for you all: for those who shipped before having an audience — what actually moved the needle first on distribution? Directories, content, cold outreach, communities? I'd genuinely love to hear what worked vs. what was a waste of time.
1 Comment
I liked the distinction between capturing words and creating action. Most meetings don't fail because people forgot what was said—they fail because nothing happens afterward. Turning notes into actual execution feels like the more valuable problem to solve.