The Indie AI Playbook

Why now

The marginal cost of shipping a feature that uses AI dropped to near zero between 2022 and 2025. Anyone with taste and a credit card can wire up an LLM and have a working product before lunch. That's both the opportunity and the trap. The opportunity is obvious. The trap is that the same low cost applies to every other person on Earth — most “AI features” you can imagine, ten other people are also shipping this week.

The indie AI builder's edge is no longer I built an AI thing. It's taste, focus, and follow-through — the same edge indie devs have always had, just routed through a faster-moving substrate.

Five principles I work by

1. AI is leverage, not the product

The customer cares about the outcome, not the model. “I made a portrait of myself in 30 seconds with zero learning curve” is the product. The LLM call inside it is a means. If you can hide the AI completely — no chat interface, no generating… spinner, no model name in the UI — the experience usually improves.

The reverse holds too: features that scream “AI” tend to age fast. Features that quietly use AI to remove a step tend to compound.

2. Distribution is half the work

A great indie product with no distribution surface is a hobby project. The App Store still rewards consistent shipping in narrow categories. Web tools still benefit from search and word-of-mouth in tight communities.

Pick a distribution surface where you already understand the audience. Don't try to be everywhere. For me that's the iOS App Store (utility-shaped problems) and a couple of niche web tools (depth in a specific workflow).

3. Cost-aware from day one

LLM costs scale linearly with usage and don't have the same fixed-cost-amortization story as traditional software. A viral indie hit on the wrong model can wipe out a quarter of revenue overnight.

Cheap defaults: Haiku-class models for anything where a small model suffices, prompt caching on every multi-turn flow, batch APIs when latency permits. Reserve frontier models for the moments where they actually change the user's outcome.

4. Solo, end to end

The whole stack belongs to one person on purpose. No handoffs, no shared context to maintain, no let me check with the designer. This is slower than a team of three on day one and much faster after week two.

The cost is breadth: you have to be passable at product, design, infra, copy, and support. The payoff is a feedback loop measured in minutes.

5. Calm UI, not dazzling

Software people use every day shouldn't feel wow. It should feel like an extension of their hand. That means careful defaults, no dark patterns, low color and visual noise, and respect for the user's attention.

Calm UI is also an indie advantage: dazzling UI requires a design team and a brand exercise. Calm UI requires taste and restraint.

What I avoid

• Subscription-first products with thin AI veneers. They burn trust.
• Heavy backend infra. Cloudflare Workers, static assets, and KV get you 95% of the way. Save Postgres for when you genuinely need it.
• Roadmaps as marketing. Ship the thing or don't mention it.
• Marketplace-style products. Two-sided markets are not an indie game.
• Anything that needs venture funding to make the math work. That's a startup, not an indie.

What I look for in collaborators

If you bring deep knowledge of a specific domain — medicine, accounting, music education, freight, family wealth management — and the right AI product would be obvious to you but you don't want to build it alone, we should talk. That's the Vertical AI partnership track.

If you're an operator who wants to deploy AI inside an existing company without hiring a consulting firm, that's the Enterprise AI deployment track — OpenAI-style deployment services, indie-scale.

If you're another indie maker who wants to co-build something on equity terms, that's Indie / Maker collaboration.

If you just want a sounding board, that's Advisory / Office hours.

The pitch is the apps

The most honest indie pitch is go look at what I've shipped. Browse the iOS apps or the web tools. Each one is a small concrete bet on these principles. Some land, some don't — that's the cadence.

If any of this resonates, say hi.