Maple Spark Labs
The Lab

What new AI can and can't do in real work.

We test emerging AI capabilities in our own work, then show what exists, what we learned and what remains unproven. These experiments can inform client work; they are not client outcomes.

GraduatedProduct · 2026

Droplet Learn

The question
Can one founder use AI coding tools to take a learning product from idea to a live service, with safeguards and operating controls designed in from the start?
What exists
A live AI learning product at dropletlearn.com. Walter directed the product, design and review, using AI-assisted development; the safety layer shipped in the first release.
Design decisions
Every request is classified and checked before it reaches the model, and every lesson is scanned before the learner sees it.
Anything flagged waits in a review queue for a person, and every request runs inside a cost budget.
What we learned
Discovery doesn't stop being a phase just because building got fast.
What remains unproven
How well the safety controls perform has not been measured, and no adoption or business results are claimed.
Read the full note →Visit ↗
Pilot / early useExperiment · 2026

A planning staff for volunteer coaches

The question
Can AI serve as a planning staff for a volunteer baseball team without taking decisions away from its coaches?
What exists
A working, file-based process with a volunteer youth baseball team. AI helps organize inputs and prepare research-informed plans, a coach proposal, and a weekly brief and practice plan, with a review and debrief rhythm. The coaches approve every decision and message to families.
What we learned
A sophisticated plan failed the real-world fit test until it was simplified around volunteer capacity, transitions and what the children and coaches would actually do.
What remains unproven
It is early: a full season has not run, not every planned workflow is in place, and no effect on games or player development is claimed. It is not a commercial app.
Read the full note →

How the Lab works

Exploring
A question and a first build. Expect rough edges.
Live
Working and in use. Still an experiment.
Graduated
Became a product or a published tool.
Archived
Stopped. The note says what we learned.

Lab rules

  • No client information, ever. Experiments use public, synthetic or our own data.
  • No personal information about real people in a public experiment. Anything involving children, a team or its families stays private.
  • Every experiment names the question it is testing and the year it started.
  • Every experiment ends with a note: what worked, what didn't, what we'd do next.

Want this kind of building pointed at your operation?

Ways to work with usMethod and templates