Borrowing a smart method without pretending you have the firepower behind it

By fast2future · June 5, 2026 · Build-in-public series

There’s a move almost everyone makes when they admire a more advanced operation: they adopt its language. The frameworks, the impressive vocabulary, the way the smart people talk. It feels like progress. It usually isn’t.

We hit this directly. We’d been studying a frontier AI lab whose approach we genuinely admire — a recursive self-improvement loop: identify your own limitations, write your own success tests, propose and validate small experiments, keep the winners, repeat. It’s a beautiful method. And we wanted in. The question we had to answer honestly was: what do we actually adopt — the method, or just the words that make us sound like them?

Here’s what we learned pulling those two apart.

Adopting the vocabulary is not adopting the capability

This was the sentence that kept us honest: you can adopt a method’s vocabulary without adopting its capability — and confusing the two is how you fool yourself.

The operation we admired is vastly better resourced than we are — a serious team with serious funding. We have one operator and an AI crew. We could absolutely start describing our work in their language — “recursively self-improving,” “open-ended,” “adversarial co-evolution.” The words are free. But saying them wouldn’t give us their compute, their team, or their depth. It would just give us a more impressive way to describe a small operation — which is a quiet form of lying, mostly to yourself.

So we set a rule: borrow the method’s sound parts, never the vocabulary-as-substance. We’re allowed to run their loop. We’re not allowed to talk like we’re their size.

Judge the tactic, not the valuation

Here’s the part that made the borrowing safe to do. A method’s worth has nothing to do with who’s funding it.

The loop we admired wasn’t invented by the well-funded lab — its core ideas (open-endedness, growing a diverse library of good solutions instead of hunting one perfect answer) draw on open-endedness research that’s been developing for years, well before the company existed. That mattered enormously. It meant we could adopt the tactics because they’re independently sound, not because a big funding round blessed them.

The discipline: judge a tactic on its own merit, not on the size of the operation using it. A good method used by a giant is still a good method for you. A bad method used by a giant is still a bad method. The valuation tells you nothing about whether the tactic works at your scale. Strip the prestige away and ask only: is this independently sound, and does it work with what I actually have?

What we actually took

Pulling the prestige off left three tactics we could genuinely use today, at no cost, that sharpen loops we already run:

1. Self-benchmarking. Before acting, the loop writes its own test: “how would we know this worked?” A system that grades itself improves faster than one waiting to be graded. Cheap, and it forces honesty about whether a thing actually moved the needle.

2. Quality-diversity. Stop hunting the one perfect article or offer. Cultivate a diverse library of good experiments and let the winners emerge from real results. (Honest caveat: this one only fully works once you can measure which won — and our analytics aren’t live yet. So we’ve adopted the structure and flagged the dependency rather than pretending it’s running at full power.)

3. Propose → test → validate → keep. Small, cheap experiments, each validated against a real check before it earns a permanent spot. Nothing enters the library on vibes.

None of those needed new money or infrastructure. They’re sharper versions of things we were already doing.

The honest anchor: a method needs a destination

There’s a seductive quality to language like “recursively self-improving, open-ended, forever-innovating.” It can pull you into improving for its own sake — building an ever-more-elaborate machine that isn’t pointed at anything a real person needs.

So we tied every adopted tactic back to a single question: does this help a real person and move us toward our actual goal? Open-endedness without a north star is just drift dressed up in impressive words. The method is the engine; the person you’re trying to help is the steering wheel. You need both, and the steering wheel comes first.

Why this is good news

It’s genuinely freeing to realize you don’t need a frontier budget to use a frontier method. The good ideas are mostly portable, mostly old, and mostly free. What separates operations isn’t access to the vocabulary — it’s the discipline to use the sound parts honestly and stay clear about what you actually have.

So borrow boldly from people doing better than you. Take their best method. Just don’t take their vocabulary as a costume, don’t let a big valuation do your thinking for you, and keep the whole thing pointed at a real person. That’s how you get sharper without getting fake.