Principles

These are the beliefs and habits that currently shape how I build, decide, and work with people. They are not fixed — when something here stops being true, I update it. Old positions are evidence of learning, not failure. But stale principles are worse than no principles.

How I build

Default to shipping

Ship something real before you have the full picture. The worst version of a live product teaches you more than the best version of a plan. I'd rather run a concierge experiment in a week than spend a month on a business plan. This applies to building — not to every decision. Some things (partnerships, fundraising, team composition) deserve deliberation. Knowing which mode you're in matters.

Lots of small bets to drive signal

Don't bet everything on one hypothesis. Run cheap experiments in parallel, watch what the data says, and double down only when you have evidence — not when you have enthusiasm. Conviction should follow signal, not precede it.

Hypothesize with gut + evidence, continue only with evidence

Intuition is a fine place to start. It's a terrible place to stay. Use instinct and experience to form the hypothesis, then test it. If the numbers don't show up, move on. The hardest version of this is killing something that feels right but isn't working.

Test it in prod

AI has changed this calculus. When the cost of building a working version drops to near zero, the right move is to ship it and learn from real usage — not to plan, spec, and de-risk in a vacuum. Controlled environments lie. Production doesn't.

The wedge comes before the platform

Start with a narrow use case that solves a real pain point. Prove it works. Expand from the foothold. Most founders try to build the full platform on day one. The wedge earns you the right to broaden.

How I think

Friction is the real competitor

Most products don't lose to other products. They lose to inaction. Understanding why people don't act is usually more valuable than building one more feature.

Behavioral insight > feature insight

The best product ideas come from understanding why people do or don't do something — not from inventing a shinier workflow. "People know they should return things and still don't" is more useful than "our returns flow has fewer steps."

Opportunity cost math

Every yes is evaluated against what else that time, energy, and risk could be buying me. Every commitment has a shadow cost. This isn't ruthlessness — it's honesty about what matters most right now.

Data > vibes

I'm skeptical of claims that aren't grounded in something measurable. "Great traction" means very little. "500 signups, no marketing, word of mouth only" means something. I hold myself to this standard.

Does it compound?

For any strategy or product: does this get stronger with time, or does it stay the same? Features usually don't compound. Distribution, data, reputation, and network effects can. I bias toward things that compound.

How I work with people

Radical candor

Say the thing directly — but care about the person you're saying it to. Being honest without being kind is just cruelty with a productivity label on it. Being kind without being honest is worse — it lets problems compound until they're unfixable. The goal is both: direct feedback delivered with genuine respect for the person receiving it.

Intentional communication

I'm measured in what I say and when I say it. Words carry weight — especially in high-stakes conversations like equity negotiations, co-founder dynamics, and investor relationships. Ship fast, speak deliberately.

Best idea wins

I don't care where the idea comes from — a co-founder, an intern, a customer, a competitor. If it's the best idea, it wins. Title and tenure don't get a vote.

Make it safe to disagree

Best idea wins only works if people feel safe surfacing bad news, half-formed thoughts, and disagreements. I'd rather hear "I think this is wrong and here's why" than find out three months later that everyone saw the problem and nobody said anything.

Learn or it was wasted

Failures are fine. Failing and not learning from it is not. Every dead end should produce a clear lesson — what we tested, what we learned, what changes. If it doesn't, the failure was wasted.

Ownership > titles

If I'm going to pour founder-level energy into something, it has to come with founder-level ownership. The title is not the point. The stake — and the control that comes with it — is. This applies to how I evaluate my own commitments and how I think about building teams.

What I believe about the craft

0→1 is where I'm most dangerous

I've scaled products — Trilogy grew to $100M+ ARR, icdcodes.ai to 300M+ impressions. I can do the scaling work. But the phase where nothing exists yet and you have to figure out what to make, how to make it, and whether anyone cares — that's where I'm most engaged and most effective. I'm at my best with extreme ownership over a problem that doesn't have a playbook yet.

AI changes the relationship, not just the output

AI doesn't just make code faster to write. It changes what it means to be a programmer — the craft moves upstream to taste, judgment, and knowing what to build. There's something worth mourning in that shift even as the overall arc bends toward progress.

Being bullish doesn't get you a round. Being a bull does.

Conviction matters more than optimism. Investors don't fund enthusiasm. They fund founders who are already charging at the problem whether or not they get funded. Show the work, not the energy.

What I'm skeptical of

• Abstraction without execution. Vision decks with no working product. Grand strategies with no shipped first step.
• Premature scaling. Hiring before the work demands it. Raising before you know what you'd spend it on.
• Consensus-driven building. Products built by committee usually blur into nothing. Strong opinions, loosely held, rapidly tested.
• The "passionate about" crowd. Passion without output is a hobby. I care about what you've shipped, not what excites you.
• Over-planning. Plans are hypotheses. The sooner they hit reality, the sooner they become useful.