Experimental Lab

Prototype unusual ideas quickly enough to learn what is real.

Feasibility spikes, interface experiments, and creative demos that expose the promise, the constraints, and the next question.

Some AI ideas need a working surface before anyone can judge them. The lab is for quick builds that expose the promise, the constraints, and the next question.

What gets built

  • Feasibility spikesNarrow prototypes that answer one important question without pretending to be a finished product.
  • Interface experimentsNew ways for people to interact with AI — voice, gestures, mixed inputs, unusual outputs — tested against real users.
  • Emerging model probesHands-on evaluation of new APIs and open-source models, with notes on where they break and what they newly enable.
  • Learning memosShort, honest write-ups of what worked, what failed, and whether the idea deserves another round.
  • Reusable building blocksTooling, prompt patterns, and SDK wrappers extracted from past prototypes — so the next experiment starts at iteration three instead of iteration one.

How the work goes

  1. Frame the bet

    Write down the question the prototype is supposed to answer. Agree on what would count as 'real' before building.

  2. Build the narrowest slice

    One feature, one path, enough to feel. Fidelity only where the question requires it.

  3. Run, capture, decide

    Let real people touch it. Capture what happened. Write the memo that tells you what to do next.

  4. Close the loop

    Archive the prototype honestly, pull the reusable pieces into shared tooling, and hand the memo to whoever picks up the next round.

A prototype is cheap; pretending you know without building one is expensive.

— why the lab exists

What you take away

Working prototype

A runnable artifact that answers the question — not a polished product, but honest about what it is.

Findings memo

What we tried, what happened, what we would change, and whether to continue — in a page, not a deck.

Next-step proposal

If the bet paid off: a concrete plan for where to take it next, with effort and risk laid out.

When to pick this

A founder or leader with a risky hypothesis

The idea could matter, but nobody has touched a working version of it yet.

A team that needs to learn a new model or API

Before betting a roadmap on it, build the smallest thing that exposes its real shape.

A pitch or partnership that needs a working demo, not a deck

The next conversation moves faster with a thing people can touch — even a narrow one — than with another slide deck full of promises.

Bring the risky hypothesis. Leave with a prototype that tells the truth.

Prototype an idea