How the lab runs
The process every product runs through, from first observation to a live domain. It has run seven times across six sectors.
Five steps, every
time, no exceptions
Skipping a step is how a build turns into an abandoned repo.
Find the leak
DaysStart from money or time already being lost, not from a feature idea.
Every product in the index began as an observed leak: calls going to voicemail, roles filled before they were posted, therapy blocked by geography. If the loss cannot be described in one sentence, it is not worth four weeks.
- One-sentence problem statement
- Who is already paying for the pain
- Why now, and why nobody has fixed it
- The cheapest possible version that would prove it
Design the wedge
~1 weekCut the idea down to the one thing that has to work for anyone to care.
The wedge is the narrowest surface that still delivers the promise end to end. One flow, one persona, one outcome. Everything else is deliberately deferred, and written down as deferred rather than forgotten.
- Single core flow, mapped screen by screen
- Visual identity and design system decided up front
- Explicit cut list of what is not being built
- Success defined before a line of code
Build it whole
2 to 4 weeksFrontend, backend, infrastructure, and integrations. No partial products.
A half built product teaches nothing, so nothing ships until the loop closes: auth, data, the actual logic, the integrations, error states, and the boring parts. One owner from schema to empty state removes an entire category of handoff drift.
- Real data, real auth, real integrations
- Error and empty states treated as features
- Deployed continuously from day one
- Analytics wired before launch, not after
Put it in front of people
OngoingShip to a live domain and find out whether the leak was real.
The product goes public under its own name and domain, with SEO, structured data, and a way to get in touch. Launch is not a milestone here, it is the start of the measurement. Opinions stop mattering and behaviour starts.
- Own domain, own brand, own analytics
- Technical SEO and schema at launch
- Direct line to early users
- Instrumented for the one metric that matters
Double down or walk away
ContinuousUsage decides what gets the next month of attention.
Some products earn more investment and some do not. Keeping one alive because it was expensive to build is how a lab turns into a graveyard, so the decision is made on evidence and made out loud. What survives stays live and maintained.
- Evidence over sunk cost
- Winners get depth, not breadth
- Live products stay maintained
- Learnings carry into the next product
The rules that
do not move
Ship whole, ship narrow
A small product that works completely beats a large one that works partly.
Boring stack, sharp surface
The infrastructure is deliberately unexciting so the novelty budget goes into the product itself.
Build where people can see it
Progress goes out publicly while it is still uncertain. It compounds into distribution and keeps the work honest.
Deliberately boring,
so the product can't be
The novelty budget belongs in the product, not the infrastructure.
Interface
Server-first React, typed end to end
Mobile
One codebase, both stores
Services
The parts that hold state
Delivery
Deploys should be boring
Reach
Shipping is half the job
A process is only worth
what it produced
Seven runs through this loop, seven products still live.