Founder Playbook · Starter Story
13 tactics from Colin Macintosh
How I Built It: $20K/Month AI App as a Non-Technical Founder
Watch the full episode“First, validate the idea by doing it for people in a way that either isn't for sale or doesn't scale.”
Validate by doing the work manually for free — in a way that doesn't scale
Don't ship a product to test the idea — do the work yourself for free, by hand, for the exact people you'd serve. If that doesn't pull demand out of them, no automated version will. Once it's pulling, then you can automate.
“The question that people should be asking themselves is, what do I do better than anyone else in the world? And then just help people, put good things into the world that are helpful, for free, for the purpose of being helpful.”
Ask 'what do I do better than anyone?' — then give it away free until demand monetizes itself
Pick the one thing you can do better than almost anyone, then give it away free until it's obvious the demand is real and large. The free help compounds into a trustworthy reputation, and eventually the audience itself will push you to monetize it.
“We launched it for free, we just wanted to see if people would use it… we were blown away by the usage statistics. In the first week I don't even know how many thousands of resumés for people, for totally for free.”
Launch the AI version for free first — let usage stats prove demand before pricing turns on
Charging on day one of a new AI product hides whether people actually want it. Launch fully free, watch what real usage does, then flip pricing on a week or two later once word-of-mouth has already started compounding.
“I sat down and wrote out a whole long resume guide of dos and don'ts, and that post on Reddit blew up so much that it is now the first result on Google for 'resumé advice Reddit'… over the last 6 years I've given away so much of my time.”
One 2018 Reddit post became the top Google result for 'resume advice' — and a 6-year content moat
One genuinely helpful Reddit post can rank in Google for years and seed a permanent audience. The asset isn't the post — it's the thousand conversations that follow, which compound trust and become the launchpad when you finally ship a product.
“The organic traffic — I've got multiple posts on Reddit: how to build a resume, the top 10 questions I get during a job hunt, how to interview, how to negotiate salary. That's a big SEO hack — just piggyback off of Reddit's incredible SEO.”
Piggyback on Reddit's massive SEO authority — write helpful posts, not promotional ones
Reddit ranks for almost every long-tail informational query. Posting genuinely useful, deep answers on the right subreddit borrows that domain authority, and the posts keep delivering organic traffic for years without any backlink work of your own.
“You can link to your products or to your company somewhere in there, but it should not be the front and center call to action — because then people will be suspicious of your intentions, and they may not react as strongly to what is valuable content.”
Bury the CTA — the product link belongs deep inside genuinely helpful content
An overt product CTA in Reddit-style helpful content kills conversion because readers smell the pitch. Keep the link deep inside the post and make the content stand on its own — the click-through rate is higher when the audience has already been helped.
“$99 for Lifetime access seems to be the sweet spot for people, and then we have a $29 one-week option for people that are just like, hey I just need this as I'm starting my job search. But we don't want to do monthly subscriptions because I find that predatory.”
$99 lifetime + $29 weekly beat monthly subscriptions for one-shot life-event products
For products tied to one-shot life events — job hunts, weddings, moves — monthly subscriptions feel exploitative because the customer never wants them past month one. A lifetime tier plus a short-window tier captures both budgets without the recurring drip.
“Dude, it's close to 100% margin product. I'd say we're probably operating around 90% net margins when it comes to the sale.”
90% net margins are achievable while OpenAI credits stay this cheap
AI wrappers running on current OpenAI pricing produce 90% net margins because inference is the only meaningful cost and it's currently subsidised. That window won't last — model providers are clearly running customer acquisition — so the time to capture margin is now, not in 18 months.
“People should not use AI for any of this — any of it should not.”
Do not use AI for the strategy work — brand identity and business plan should stay human
Brand identity, positioning, target market — these are documents that require you to think hard about what you actually believe. Outsourcing them to an LLM produces glossy nothing. Reserve AI for execution, not for the decisions only you can make.
“The cool thing about AI is, I don't have to code to teach the AI. A resume review would take me three hours to do, but now I can do 20, 30, 100, 1,000 resume reviews a day with this AI I built, for a fraction of the price.”
AI lets you scale expertise without learning to code — one expert review becomes 1,000 a day
The leverage in AI products is not faster code — it's bottling expert judgment that doesn't scale physically. Teach the model the rules a domain expert uses, and a service that consumed hours per customer becomes one that handles thousands a day at near-zero marginal cost.
“You're not going to ever convince some random software engineer to work on something with you that is your idea without paying them a lot of money, nor should they… It was working with Nate one-on-one for years.”
You won't convince a random engineer to build your idea — co-founders come from years of relationships
Cold-pitching a technical co-founder rarely works because the engineer carries all the build risk. Compounding relationships — staying in touch with old colleagues, helping strangers for years — is what produces the partner who'll equity-split because they already trust you.
“I split this three ways with these two other guys. I just give them the features and the notes and my vision. I probably spend 5 hours, 10 hours a week maximum on this project.”
Spend 5–10 hours a week max — design the side project to scale without you
A side project doesn't have to consume nights and weekends if you give the build to two partners on a clean equity split and keep your role to vision plus feature notes. The business compounds without you, which is the whole point of a side project.
“I would recommend that people move very quickly because right now you can get OpenAI credits dirt cheap, and I don't think it'll be like that forever. They're just doing user acquisition right now on their end.”
OpenAI credits are dirt cheap right now — this margin window won't last
AI infrastructure providers are subsidising inference to capture developers — the same playbook every cloud has run. Margins on AI wrappers are temporarily inflated, so the founders moving fastest into the gap capture the most. Building 'when the dust settles' means waiting for 30% margins instead of 90%.