Nov 5, 2026
Three things stuck with me today out of 17 new cards and 56 signals. All the interesting stuff came from Product Hunt and competitor gaps — zero from any other platform, which tells you the radar is basically reading two rooms right now.
The top scorer is Chrome extension that drafts industry-specific customer support replies at 73, monetized at $12–29 per seat/month with a free tier capped at 50 drafts a month. My honest reaction: the pain is real but the wedge is soft. "Industry-specific" is doing a lot of work in that title — a legal firm's support replies and a Shopify store's support replies are different products wearing the same skin. And a Chrome extension is the worst possible surface for support work, because half the messages live in Gmail, a quarter in a helpdesk, and the rest in DMs. You'd end up maintaining six integrations behind a browser popup. Per-seat pricing also means you need actual teams, which means a sales motion, which means this is not a weekend build. I'd pass, but I'd steal the pricing page.
Post-launch AI model versioning, evals and fallbacks at 70 is the one that genuinely excited me, and also the one I'd be most scared to build. Every person shipping an AI feature right now has the exact same story: the model provider ships an update, your output quietly changes, and you find out from a customer. "Prompt versioning plus a fallback chain plus eval runs in CI" is a real product, and it's a product that gets more valuable every month. The catch is that your buyer is a developer, and developers are the cheapest, most DIY-prone customer on earth. You'd have to price it like a CI tool, not a SaaS — usage-based, probably $50–200/mo per team, and you'd still lose half your trials to someone wiring up promptfoo on a Friday afternoon. If I built it, I'd go narrow: Postgres-backed prompt registry, GitHub Action that runs evals on every PR, and an alert when your eval score drops more than 10%. That's a week of work for an MVP and maybe six months of grindy enterprise-ish sales. Tempting. Not this month.
Then there's Freelancers struggling to chase late clients and reconcile payments at 67, freemium around $8–12/mo with auto-reminders and reconciliation behind the paywall. This is the least sexy and, I think, the most likely to make money. Freelancers pay for things that make them money directly, and "I got paid 11 days faster" is a number they can feel. The problem is competition from every invoicing tool that already does reminders as a free feature — the "can never tell who actually paid" reconciliation angle is the only defensible part, and that's a bank-feed/Stripe integration problem, not a UI problem. I'd build it anyway. Next.js, Supabase, Stripe Connect, plain Gmail/SMTP sends for the reminders, and skip the fancy dashboard — send a weekly "you're owed $4,200 across 6 clients" email and you've basically shipped v1.
The email HTML converter at 65 is the one I'd quietly skip. It's a genuinely painful problem — Outlook rendering is a war crime — but it's a $49 one-time purchase with a free tier, and one-time purchases don't compound. You'd be rebuilding it every time Google tweaks Gmail.
What I keep coming back to: the highest score today is the softest business, and the lowest-scoring idea I looked at is the one with the clearest path to $5k MRR. Scores measure how loud the complaint is, not how much someone will pull out a card. I'd rather have a boring 67 with a wallet attached.