Principles
The principles we hold to.
The version of how we operate that's written down. We try to live by these in every engagement. When we don't, we owe you the why.
Engineering
Working code beats slides
Every demo shows running software. If we don't have something to show, that's a status problem we resolve before the meeting — not a slide we hide it behind.
Exception-first design
We design for the refusal cases, the escalation paths, and the failure modes before the happy path. The happy path falls out.
Boring infrastructure
Postgres over Kafka. REST over GraphQL. Hosted API over fine-tuning. The clever path is for the AI itself, not the plumbing.
Observability from commit one
Every prompt, retrieval, tool call, and output logged with reasoning. Replayable traces are non-optional.
Eval suites are permanent systems
Not a launch checklist. The eval suite outlives the engagement and grows from production failures.
Sales + scoping
Discovery is paid
We don't promise scope and price before we've done the work to deserve confidence in it. Free pre-sales discovery causes the bad fixed-fee math everyone complains about.
We turn down work that doesn't pencil
If the ROI math doesn't support the build, we say so. Not a flex — it's the only way fixed-fee actually works long-term.
Pricing is public
$25–$35/hour, fixed-fee sprint envelopes, retainer ranges. If you can't see the rough number on the site, we're hiding it from you.
We don't sell what we can't ship in 8 weeks
If it can't be a v1 in 8 weeks, we scope it down or pass. Year-long discovery projects are someone else's business.
Client communication
You talk to engineers
Not account managers. The engineer on your project is your point of contact for technical questions. We don't insert layers.
Weekly demos, no exceptions
Even if it's a half-built thing. Especially if it's a half-built thing. Slipping weeks happen — silent slipping weeks are the real failure.
Bad news first
If something's going off-track, the engagement lead surfaces it before you ask. Surprising the client with timeline misses kills trust faster than the missed timeline does.
Honest scoping over optimism
We'd rather quote 8 weeks and ship in 7 than quote 4 weeks and ship in 9. Compounding trust beats winning the bid.
Hiring + team
Senior on every project
No juniors learning on your dime. We staff seniors first, then add reach where it helps.
We hire when demand calls
Not when investors expect headcount growth. We've kept a short bench warm precisely so we can scale without aggressive over-hiring.
We borrow from Alian Software's bench
Same hiring bar, shared infrastructure, same code-review process. The 15-year engineering practice is part of the AI division by design.
Async-first written communication
Most of our team is async. Decisions get written down, debated in writing, and revised in writing. Meetings exist for the things writing can't resolve.
What we won't do
No vendor lock-in
Code, prompts, configs — yours from commit one. We don't license you 'our framework.' If you fire us, the AI still works.
No 'AI strategy' decks without code attached
We can do strategy phases. We also ship build artifacts at the end of them. Strategy decks alone aren't a product we sell.
No agents that take irreversible action without approval
v1 agents draft. Humans approve. System executes. Period.
No 'set it and forget it' deployments
If you don't keep running the eval suite, the model drifts. We name an owner before launch.
Sound like you?
Book a 20-min call. Or read the engineering methodology that makes these principles tactical.
