We’re past 40 people now, and a real line in the budget goes to development. Courses, programs, the occasional coach. For a long time it seemed obvious which of these counted. The big, structured things were serious development, a nanodegree or a part-time master’s. Coaching and a good book were the lightweight option. The more involved the program, the more it looked like the real thing.

The engineers and designers never ranked things that way. Given the choice, they’d take a short, specific course on the exact thing in front of them and leave the comprehensive program untouched. The coaches said something close to the same. Coaching is the bigger lever for building people, even for individual contributors, and the large programs mostly aren’t needed. For a while it was easy to write both off as recency bias, people overrating whatever they’d touched last.

Two things made that harder to hold. The coaches arguing against structured programs had almost all come up through structured programs themselves, standing on a foundation they’d stopped noticing. And there was a case closer to home. A couple of years ago a long, broad, expensive program left me with one thing that stuck, the Balanced Scorecard. Not just that it exists, but a feel for how an instrument like that gets used in practice. That’s not something you go looking for on your own. You can’t search for what you don’t know is there.

That was the part the ranking missed, and the part the team had been right about all along. A big program is good at showing you the territory, what exists and how the pieces fit together. It’s worse at the thing they actually needed, which is getting good at one specific skill. That part is a muscle, and muscle comes from doing the work over and over with quick feedback. A two-year program gives you almost no reps, and the feedback comes back too late to change anything. Two different needs, treated as one, with the default quietly set to the version that suited me.

What I’m taking from this: a lot of it was a founder’s time horizon, applied without noticing. Founders think in years, because that’s the job, building something that compounds over a decade. From that seat a slow, broad program is obviously worth it, because the map pays off for ten years. But the people on the team are measured on what ships this quarter. A map they can’t use for eighteen months isn’t an investment to them, it’s dead weight with a tuition bill. That wasn’t really an evaluation of learning. It was one time horizon, quietly handed to everyone else.

The more uncomfortable part is that long-term thinking is a bet that the world stays still long enough for the knowledge to pay off. In a stable field that’s a safe bet. In a startup the ground moves faster than anything else, and the slow, broad investment can be stale before it ever comes back. The same instinct that pays off on strategy can quietly work against you on skills.

Open question I’m still sitting with: how often is the big program even needed now? The thing it really gave was the map, knowing the instrument existed and roughly how to use it, and that’s exactly the part AI tools are getting good at handing you the moment a problem shows up. If the map keeps getting cheaper to borrow, maybe the only part worth building the slow way is the problem-solving itself, and the case for the big program gets thinner every month, for founders as much as for their teams.