The Bet
In late 2025, when we started designing the studio's methodology, the core decision was simple:
Do we build for ourselves first, or only for clients?
Most agencies choose only for clients. The math is simple. Client engagements pay. Internal products don't (until they do, eventually).
We chose to do both. Here's why.
The Credibility Argument
A studio that has never shipped its own AI product is selling theoretical capability.
"We know how to build AI for you" is a promise. "We've shipped these AI products of our own, here are the live metrics, here is the methodology that survived contact with our own reality" is a different kind of promise.
The first is marketing. The second is a contract.
We chose the contract.
The Methodology Argument
Building for clients teaches you what clients want. Building for yourself teaches you what works.
Verdikt taught us:
- Eval-first development survives contact with reality. Prompt-first development doesn't.
- Model routing pays off in week three, not week eight.
- The 5-section structure was a 4x conversion lift over stream-of-thought analysis.
- Cost-per-output is the only cost metric that matters at unit-economic scale.
These are the lessons we couldn't have gotten from client work alone. Client engagements have client constraints. Our own products have only the constraints we set.
The lessons compound. Every new client engagement starts with the methodology Verdikt validated.
The Learning Loop Argument
When a client engagement hits a problem, the client usually has an opinion about what to try. Sometimes the opinion is right. Sometimes it's wrong. Either way, the constraint is real.
When our own products hit a problem, we can try anything. Different model. Different prompt structure. Different data source. Different UI affordance. The experiments we run on our own products are the experiments we wouldn't be allowed to run on a client's.
The learning loop is faster, broader, and more honest.
The Trust Filter Argument
The single best filter buyers can apply to any AI vendor is:
"Show me one AI workflow you've shipped in your OWN business."
Most agencies cannot answer this. The ones who can are the ones to hire. We are deliberately one of the ones who can.
What It Costs
Building Verdikt cost us:
- 6 weeks of focused engineering and design time
- ~$8,000 in model provider costs during development and alpha
- Opportunity cost of two simultaneous client engagements
What we gained:
- A live product generating revenue (small, but real)
- Three more product builds (Capsule, Lanes, Triage) already in motion
- A validated methodology
- A credibility artifact every prospective client asks about
- A case study that converts inbound
The math has been worth it. We expect it to compound further.
What We Don't Recommend
We don't recommend that every studio do this. Building your own product is hard. It takes founder time and engineering time and design time that could otherwise go to billable work.
What we do recommend: if you're evaluating an AI vendor and they have no shipped product of their own, weigh that against the alternatives. The vendor who's never built will not understand what they're selling.
What This Means For Clients
Three concrete commitments come out of the dogfood manifesto:
1. Our engagement playbook reflects what we've learned. Eval-first. Model-agnostic. Cost-per-output. These aren't theories. They're rules we follow on our own products too.
2. Our case study is honest. We document what worked and what didn't on Verdikt. The lessons live in the Verdikt case study. When a client engagement makes us re-learn one of those lessons, we update the case study.
3. Our advice is calibrated. When a client wants to skip evals "to ship faster," we know exactly what they'll pay for it because we paid for it ourselves once. The advice is grounded in our own scar tissue.
What We Ship Next
Verdikt is live in alpha. Capsule, Lanes, and Triage are in active development. We commit to shipping at least one of those to public availability in 2026. We commit to publishing the case studies as they ship.
This is the methodology. The receipts are the products. The receipts come first.
If you want a studio that built before it sold, we are here.