APLOT P-A
Zero to launch
A new AI product, from problem definition to paying users, built with your team in the room.
Partnership — three plots
Your platform, built by the lab that builds its own.
The lab takes a few partner platforms a year — the same people who ship Averil and LobeStack. You own the repository from the first commit.
Pick the plot that fits.
Three ways an engagement starts. Say which one in the form — or ignore the labels and describe the work.
BPLOT P-B
Scale what works
A platform with traction that needs more users, markets, or channels without breaking.
CPLOT P-C
Intelligence inside
AI capability added to software you already run — measured, gated, honest about limits.
You own the repository from the first commit. A few partnerships at a time, finished properly. No outsourced development, no billable-hour retainers, no website builds.
Four phases. Each ends in running software.
Not a document about software you might get later. You own the code and the decisions throughout.
Phase 01
Find the real problem
Two weeks with your team and your data. We work out where AI genuinely helps — and where it would just be expensive.
Ends with: A scoped product thesis and an honest no-go list.
Phase 02
Prove it works
A working prototype with evaluation attached, so quality is a number you can watch rather than a feeling.
Ends with: A prototype your users touch and an eval baseline.
Phase 03
Build and instrument
Weekly releases behind flags, with logging, cost tracking, and alerting in place before the first real user arrives.
Ends with: Production access, dashboards, and a runbook.
Phase 04
Hand over the keys
Your team owns the code, the models, and the roadmap. We stay on support for as long as it's genuinely useful.
Ends with: Repo transfer, documented architecture, trained team.
We take on three partner products a year. That is how many we can build without the quality of our own products slipping.
Fit
Work we take, and work we don't.
Work we do well
- A product that needs to exist in months, with a team that will keep running it
- AI features where being wrong is expensive and quality has to be measured
- Existing software that works but nobody enjoys using
- A founding team that wants to be in the build, not briefed on it
Work we turn down
- Staff augmentation, or bodies billed by the month
- Projects where AI is the requirement and the problem comes later
- Fixed scope written before anyone has spoken to a user
- Anything we'd have to hire a team we don't know to deliver
Open plot
Tell the lab what you're building.
Plain language is fine. We reply within two working days — including when the answer is no.