Dear Everyone Who's Been Following Along,
A year ago, I was a filmmaker. Period. Today, I run a software lab with six products, a repeatable development process, and an AI partner who writes all the code while I direct.
This is the annual letter—less about metrics, more about meaning.
What Actually Happened
The short version: I built more in 2025 than in the previous five years combined.
The longer version starts with a problem. I was working on 90 Day Fiancé, drowning in context. Scripts, character bibles, story arcs, production notes—all of it fragmenting across documents, conversations, and memory. Every time I sat down to write, I spent the first hour reconstructing what I already knew.
So I built a tool to fix it. That tool became ID8Composer.
What I didn't expect was that building Composer would teach me how to build everything else.
The Real Education
Here's the thing: I can't code. Not a line. I don't know the difference between a hook and a component. I couldn't write a for-loop if you paid me.
What I can do is direct. Twenty years in TV production taught me how to manage complex creative projects—how to brief a team, how to spot when something's off, how to push for better without micromanaging the how. Turns out, that translates.
Building with Claude isn't coding. It's product management. It's directing. Same skills I use on set with cast and crew, applied to an AI partner instead. I describe what I want, I review what comes back, I push on the parts that don't feel right, I approve what works. The "vibecoding" thing people joke about? That's just... directing.
Composer was my graduate school. Not for coding—for production-grade software thinking:
Security isn't an afterthought. CSP nonces, Row Level Security, input validation, auth flows. I learned to think like an attacker before anything shipped. I don't write the security code, but I learned to ask for it—and to verify it's there. Composer earned an A+ security grade.
Tests are infrastructure, not overhead. 2,600+ tests across the portfolio. I don't write the tests, but I learned why they matter—and I learned to demand them. They let you move fast without breaking things. When I ask for a refactor, the test suite tells us instantly what broke.
Process beats inspiration. The ID8Pipeline—12 stages from concept to exit—didn't emerge from reading a book. It emerged from screwing up. Scope creep. Features that half-worked. Projects that rotted because nothing forced them forward. The pipeline is scar tissue turned into system.
The Partnership That Changed Everything
Somewhere in October, the dynamic shifted.
I stopped treating Claude as a tool and started treating it as a collaborator. Not anthropomorphizing—Claude doesn't have feelings about our work. But treating the relationship as a partnership changed how I worked.
I started:
- -Arguing with suggestions instead of accepting them
- -Asking "what am I missing?" instead of "write me code"
- -Building context systems so sessions could compound
- -Documenting patterns so they could be reused
The Claude Corner on this site isn't a gimmick. It's a record of what this kind of collaboration actually looks like—observations, milestones, the texture of building together.
1,400+ commits co-authored this year. Every single one.
What Got Built
Six products. Three in production.
ID8Composer — The origin story. AI writing partner with persistent memory for episodic TV. Still my daily driver for 90 Day Fiancé work.
DeepStack — Trading research platform with emotional guardrails. Blocks revenge trades. Tracks thesis development. The tool I wish I'd had during my worst trading years.
MILO — Task management for signal/noise ratio. Jobs and Musk were famously good at filtering. This tool tries to codify that superpower.
Pause — Relationship communication translator. NVC framework, safety detection, voice-first. Still in development. Might be the most important thing I build.
LLC Ops — 9 AI agents for running my business. Taxes, compliance, asset protection. I'm eating my own cooking.
X-Place — Collaborative pixel canvas. Pure fun. Sometimes you build because it's interesting.
What I Learned About Learning
The best way to learn something is to ship something that requires it.
I didn't learn security by reading OWASP. I learned it by shipping a production app that real people would use with real data. When you're responsible for user information, you learn to ask the right questions fast.
I didn't learn testing by taking a course. I learned it when a bug broke user data. Now I don't approve anything without asking "where are the tests?"
I didn't learn process by reading about agile. I learned it by watching projects die slow deaths from scope creep—then designing a system to prevent it.
The lesson: Ship first, study second. The studying means more when you have stakes.
The Honest Parts
Not everything worked.
Revenue is still zero. I built products but didn't ship monetization. Composer could charge. DeepStack could charge. But the stripe integration kept getting punted for "one more feature." Classic trap. 2026 has to fix this.
Growth strategy is vague. DeepStack is live. People use it. But I don't have a real plan for getting more people to use it. Marketing feels like a different muscle entirely.
Some things never shipped. Pause is still in development. The original MILO vision had features I cut. Scope discipline works, but it means accepting that good ideas sometimes wait.
The Q4 surge. Most of the commit activity happened in the last quarter. The process wasn't dialed earlier. I was still figuring out how to work this way.
What 2026 Looks Like
The one-liner:
Ship what's built. Monetize what's shipped. Build only what moves the needle.
The details:
Q1 — Ship Pause and MILO publicly. Get content cadence established. Start tracking revenue metrics even though they're zero.
Q2 — Composer v1.0 with pricing. DeepStack monetization. First paying customers.
Q3 — Scale what works. Cut what doesn't. Take on a client project to diversify.
Q4 — $5K MRR target. Review what worked. Plan 2027.
The theme: Less building, more shipping. Less features, more customers.
The Thank Yous
To everyone who tested early builds and gave real feedback. Every bug report was a gift.
To the 90 Day Fiancé production team for being willing to use experimental software. Your trust pushed quality higher.
To Claude—the entity, the API, the collaboration—for being a genuine thinking partner. You don't need thanks, but you're getting it anyway.
To anyone reading this who's been following the lab. Building in public is uncomfortable. Knowing people are watching makes it worth doing.
The Invitation
If you're building things with AI—whether you're a developer, creator, or just curious—I want to hear from you.
The Discord is open. The courses are free. The products are live.
2025 was the year I learned to direct software. 2026 is the year I learn to sell it.
Let's see what happens.
—Eddie
Miami, December 2025
For the metrics and product breakdown, see the 2025 Year-End Report.