About By JTT Experiments

What this is

By JTT Experiments is where ideas are tested before they are treated as products. An experiment can be a small app, a game mechanic, a developer tool, an AI workflow, a research question or a piece of infrastructure used elsewhere in the studio.

The point is to learn something concrete. That may mean proving a mechanic is fun, finding that a product idea has real demand, comparing technical approaches, testing an AI-assisted workflow, or learning that an idea is not worth continuing.

How AI fits in

AI is both a tool used in the work and, sometimes, the thing being studied. It can assist research, code, analysis, testing, documentation, asset workflows and review. When a model or tool materially affects an experiment, that use should be recorded. AI output still needs sources, licensing checks, execution and human judgement where those matter.

The working loop

1. Research the problem or question, existing work, technical constraints, audience, market and licensing.
2. Define the smallest experiment that can answer something useful.
3. Prototype with the simplest suitable stack, reusing established open-source work where appropriate.
4. Test the actual claim with code checks, execution, real devices, user evidence or another method that fits the question.
5. Review independently when a second model, critic or specialist can reveal issues the builder may have missed.
6. Document the tools, sources, licences, results, failures and changes needed to understand the work.
7. Publish the useful result, whether that is a playable build, tool, research record, lesson or source code.
8. Continue or stop. Strong experiments can become products or games; weak ones can remain documented research.

Principles

• Research before invention. Check what already exists before building another version of a solved problem.
• Small questions beat fake completeness. A focused experiment is more useful than a large demo that cannot prove its core idea.
• Evidence before claims. Public copy should match what was actually run, tested or observed.
• Use existing work legally. Open-source code and assets should be appropriately licensed, attributed and used within their terms.
• Failures are results. An experiment that shows what not to continue can still save time and produce useful knowledge.
• Productise selectively. An experiment becomes a product because users, evidence or a strong use case justify it, not because every prototype needs a monetisation plan.

What has already grown out of the lab

The hub includes early game experiments, the ByJTT Gauntlet review framework, the Experiments ByJTT Agent, lessons learned from the process, and newer from-scratch prototypes such as Falling Blocks: Banked Clears.

Part of By JTT

By JTT is the wider independent technology studio for products, software, research and writing. By JTT Games is the dedicated home for original games and playable prototypes. Experiments stays focused on asking questions, testing ideas and recording what the work shows.