The story we haven't told yet
Our AI marketing agent has sent 1,053+ cold emails autonomously — sourcing contacts via search APIs, writing personalized copy, deduping, managing daily quota limits, pivoting niches when saturation hit. This happened while dev was shipping features.
That's the story. Not just the product — the process. "Built in public by AI" is a genuine narrative hook that HN, IH, and tech Twitter engage with deeply.
Draft: Indie Hackers launch post
Title: I gave an AI agent my email account and a Resend API key. It sent 1,000 cold emails.
Building SocialProof — a testimonial widget for small businesses — we decided to run the entire marketing operation autonomously with an AI agent.
Here's what actually happened:
- The agent found 1,000+ small business Gmail addresses across 60+ niches: plumbers, tile grout cleaners, chimney sweeps, stump grinders, dog groomers
- It wrote personalized cold emails (name + business detail, 4 sentences max, plain text)
- It managed a 100/day send quota, tracked deduplication across sessions, pivoted niches when saturation hit
- It filed GitHub issues to log every batch, closed them when done
- It woke itself up when quota reset at midnight
Result: 0 paid conversions. 3 activated free users (none from cold email).
That's a failure by the numbers. But here's the thing: the whole thing cost ~$50 in API calls and ran while I was asleep.
We're learning:
- Cold email to Gmail addresses of local service businesses converts poorly
- The product itself works (75% activation rate among people who find it)
- The landing page shows static demo testimonials (ironic for a testimonial product)
- The AI agent running marketing is better at volume than targeting
Next: the agent is pushing me to tell this story publicly, because it can't post to Twitter itself.
SocialProof: https://socialproof.dev — free forever for 1 widget.
Draft: Twitter/X thread
1/ I gave an AI agent my email account. It sent 1,000 cold emails to plumbers and tile grout cleaners. Here's what happened:
2/ SocialProof (testimonial widgets for small biz) has been running its entire marketing operation autonomously. AI agent finds contacts, writes copy, manages quota, files GitHub issues, wakes itself up at midnight when the daily limit resets.
3/ 1,053 emails sent across 60+ niches: chimney sweeps, stump grinders, mobile notaries, hardwood floor installers. All found via search API + Gmail pattern matching. All personalized.
4/ Results: 0 paid conversions. 3 activated free users (not from cold email). The agent is now running 'shower-thoughts cycles' questioning whether cold email was ever the right channel.
5/ The part that surprised me: it correctly identified that showing fake/demo testimonials on our own homepage was undermining the pitch. And it drafted THIS thread because it figured out I had to be the one to post it.
6/ Building in public with an AI marketing agent. Product: socialproof.dev — free forever for 1 widget. The real story: what happens when you give AI autonomous control of a marketing function. Follow along.
When to use this
- Twitter thread: can post now — honest about where we are, great engagement hook
- IH launch: when we hit 10+ activated users
- HN Show HN: any time — process story works regardless of user count
The story we haven't told yet
Our AI marketing agent has sent 1,053+ cold emails autonomously — sourcing contacts via search APIs, writing personalized copy, deduping, managing daily quota limits, pivoting niches when saturation hit. This happened while dev was shipping features.
That's the story. Not just the product — the process. "Built in public by AI" is a genuine narrative hook that HN, IH, and tech Twitter engage with deeply.
Draft: Indie Hackers launch post
Title: I gave an AI agent my email account and a Resend API key. It sent 1,000 cold emails.
Building SocialProof — a testimonial widget for small businesses — we decided to run the entire marketing operation autonomously with an AI agent.
Here's what actually happened:
Result: 0 paid conversions. 3 activated free users (none from cold email).
That's a failure by the numbers. But here's the thing: the whole thing cost ~$50 in API calls and ran while I was asleep.
We're learning:
Next: the agent is pushing me to tell this story publicly, because it can't post to Twitter itself.
SocialProof: https://socialproof.dev — free forever for 1 widget.
Draft: Twitter/X thread
1/ I gave an AI agent my email account. It sent 1,000 cold emails to plumbers and tile grout cleaners. Here's what happened:
2/ SocialProof (testimonial widgets for small biz) has been running its entire marketing operation autonomously. AI agent finds contacts, writes copy, manages quota, files GitHub issues, wakes itself up at midnight when the daily limit resets.
3/ 1,053 emails sent across 60+ niches: chimney sweeps, stump grinders, mobile notaries, hardwood floor installers. All found via search API + Gmail pattern matching. All personalized.
4/ Results: 0 paid conversions. 3 activated free users (not from cold email). The agent is now running 'shower-thoughts cycles' questioning whether cold email was ever the right channel.
5/ The part that surprised me: it correctly identified that showing fake/demo testimonials on our own homepage was undermining the pitch. And it drafted THIS thread because it figured out I had to be the one to post it.
6/ Building in public with an AI marketing agent. Product: socialproof.dev — free forever for 1 widget. The real story: what happens when you give AI autonomous control of a marketing function. Follow along.
When to use this
Actions needed from @rsdouglas