Arete

Arete

ἀρετή — excellence earned through effort, not given.

LLMs are great at writing code. They're terrible at deciding *what* to build.

I kept hitting the same wall. I'd give Claude a vague prompt, and ten seconds later I had a docker-compose.yml, a schema, and three new dependencies. No questions asked. I was missing this kind of, you know, annoying senior engineer challenging my decisions. So I was like, I'm going to build one.

That's Arete. A guided brainstorm that forces you to pause, think, and design properly before any code gets written. Five phases, each with exit criteria.

Two tracks

Arete detects whether you're solving a technical or a conceptual problem and routes the questions accordingly.

How it works

Five phases, each a skill. They run in order. If stress-testing finds gaps, you loop back instead of pushing through.

Ground. Make sure the problem is real. Probes the trigger, the pain, who hurts, the cost of inaction, and the user requirements. Refuses vague pain — "it's slow" gets "how slow, for whom, under what load?" There's a kill switch: if the stakes are vague ("it's not ideal"), the conversation stops. The thing is, most "problems" don't survive ground.

Explore. Diverge. Surfaces multiple approaches before narrowing. No solutioning yet. One question at a time, building on your keywords. You'd be surprised how often "I already know what I want" doesn't survive three minutes of this.

Decide. Converge. Forces a trade-off matrix — effort, risk, reversibility, NFRs. Reversibility is the most undervalued axis in tech decisions, of course, so it's always there. The decision is a provisional bet, not a marriage.

Stress. The grind. Challenge every claim from explore and decide, and on the technical track, sharpen rough user requirements into…