Direction Is the Only Scarce Variable

One day in June I ported GrowthMentor's entire marketing site into our Rails app. Homepage, membership, events, the blog, 99 customer stories. Started in the morning, rendering by 4am.

I'm not an engineer. Claude Code does the typing, I direct.

For years I thought about growth as Speed × Quality × Direction. Multiplicative, so a zero in any one kills the whole thing. You could be fast but sloppy. High quality but slow. Fast and good but pointed at the wrong mountain. The game was balancing all three.

Then AI happened.

Speed went first. I rebuilt GrowthMentor's AI matching system, the core of the product, in one week. The kind of build that used to be a two-week sprint with a dev team is done before dinner. Add the month up honestly and it's about a year of development, shipped in June.

Quality followed. The output is genuinely good, and where it isn't, iteration costs almost nothing. You ship, you see what breaks, you fix it in minutes. The feedback loop went from weeks to hours.

So what's left?

Direction.

And here's the part nobody talks about, direction was always the hardest variable. It just used to be hidden behind the other two. When speed and quality were expensive, you never got to the direction problem. You were too busy arguing about sprint capacity and whether the button should be blue or green. Now that they're nearly free, direction is exposed. Every day, all day.

A month of verdicts

I keep a log of every working day, and reading June back, the building barely registers. The entries are all verdicts.

Our old Y Combinator blog hub had been "improved" at some point by an AI rewrite, fake avatars, a fabricated "patterns across 15 founders" analysis that none of those founders ever said. Restoring the real version and porting the 14 founder interviews it links to was mechanical, an afternoon of work. The part that mattered was the verdict to throw the rewrite out.

Same week, I pulled twelve months of Search Console data before deciding which old posts were worth saving. One post had 29,500 impressions and looked like the prize. Wrong intent, zero clicks, worthless. The asset worth rescuing was a hub stranded on page 3 that most tools would tell you to delete. The data pull took minutes. Knowing which number to distrust was the work.

And the night before that site port, at 11:30pm, I talked myself out of starting. Nothing was forcing it, and code written exhausted at midnight is code you redo with fresh eyes. I stopped, slept, started in the morning, and the whole thing took a day.

Every one of those decisions is small. None of them can be delegated to the model doing the work, because the model is the thing being judged.

Smart, fast, pointed wrong

The founders I talk to are living the same thing. Over 62,000 sessions have run through GrowthMentor since 2018, and I've sat in thousands of them myself. The ones who struggle are smart and fast. They're optimizing a funnel that shouldn't exist, building a feature nobody asked for, chasing a market that doesn't want what they're selling. Genuine speed problems are rare, I could count them on one hand.

The best mentors I've ever seen don't hand you a playbook. They watch you describe your situation for five minutes and then ask why you're doing that at all. And you realize you don't have a good answer. That single question, from someone with the right pattern recognition, saves you six months. No AI does that yet. Maybe ever.

Why mentorship is worth more now

Direction doesn't brute-force. There's no A/B test for which mountain to climb, no prompt that produces taste. It comes from reps, from having been wrong enough times that you recognize wrong early, while turning around is still cheap. AI is a $500K rally car that anyone can rent for the day. The race goes to whoever already knows the road.

The tactical layer, how to set up Google Ads, what a cold email should say, is a commodity and AI eats it for breakfast. It should. What's left are the questions AI answers confidently and generically because it doesn't know you. Pivot to enterprise or stay SMB, is this feature for the market or for avoiding the harder problem, are you even on the right mountain.

The growth equation simplified itself in the background while everyone was busy benchmarking models. Speed and quality fell out. Direction is what's left.