Start a project / A direct conversation
Have something real to build?
Tell me what you are trying to make. I will tell you what I can own, what needs proving first, and the smallest useful way to start.
Scope / Three kinds of work
Where I can be useful.
The shape can change. The boundary does not: the product has to remain understandable, testable, and honest about what the model cannot own.
- 01
AI products people can use
A working product in the browser, with the model, data, evaluation, and interface connected.
Useful for
- Retrieval over private material
- Agents with bounded actions
- Generation workflows with durable jobs
Boundary
- No unsupported benchmark or user claim
- No pretending a prototype is production
- 02
Business workflow automation
Repeated work moves through a visible, recoverable workflow while people keep authority over consequential decisions.
Useful for
- Long-running workflow orchestration
- Risk and fraud scoring
- Scheduled infrastructure tasks
Boundary
- The model does not own lifecycle state
- High-impact actions keep an explicit gate
- 03
Interactive product storytelling
A web experience that explains how the system works without hiding the useful page behind motion.
Useful for
- Data-led visual stories
- Accessible interaction and reduced motion
- Static fallbacks that keep the whole argument
Boundary
- Motion never owns navigation
- A visual never invents project data
Process / Four clear moves
We make the uncertainty smaller.
This is the default path, not ceremony for its own sake. A project can begin smaller when the evidence says it should.
- 01
Start with the actual problem.
We get clear on who is stuck, what exists today, and what a useful change would look like.
A problem worth solving
- 02
Prove the risky part first.
Before a large build, we test the assumption most likely to make the idea fail.
Evidence, not momentum
- 03
Build the smallest useful version.
The interface, model, data, evaluation, and recovery path are connected around one real job.
Something people can use
- 04
Ship it, measure it, leave it clear.
We document the decisions, keep the failure states visible, and decide the next step from what happened.
A product you can continue
Fit / Better to know early
A good fit is more useful than a forced yes.
If I am not the right person, I will say so. If the idea needs proof before implementation, that is where I will suggest we start.
Likely a fit
There is something real underneath the idea.
- There is a real workflow, user problem, or product idea behind the AI request.
- The riskiest assumption can be tested with actual material or a bounded prototype.
- A human keeps authority over decisions that affect money, access, health, or people.
- You want the useful page and reliable system, not only the impressive demo.
Probably not a fit
The request needs the story to outrun the evidence.
- The brief needs invented proof, hidden limitations, or a claim the evidence cannot support.
- The model is expected to own consequential decisions without a review boundary.
- The request is only “add AI” and there is no user problem to examine yet.
- The project depends on pretending a prototype is already a production system.