Skip to main content

Same seed, same world: why I made LIVEX deterministic

Reproducibility is the difference between a story and a result.

Updated 2 min readMarkdown
Part of seriesLIVEX: Behind the Simulation (EN)

When a simulation produces something surprising, the first question is always the same: can I see it again? If the answer is “maybe, with a bit of luck”, then what you have is an anecdote.

In LIVEX, the answer is yes by design. The same seed and the same configuration replay the same trajectory, bit for bit. Each run carries an identifier built from its seed, so a result can be named, shared and replayed.

Why it matters for emergence

Emergent behavior is slippery. A group forms, a routine appears, and you want to know why. With a deterministic engine you can rewind, change one parameter, and compare two runs that differ by exactly one thing. Without that, you are comparing two different luck draws.

What it costs

Determinism is a constraint you pay for continuously. Every source of randomness must come from the seed. Ordering must be stable. Anything that depends on the machine or the clock becomes suspect. It also pushes you to keep the engine strictly separated from observation, which is why SYNE simulates and ECHOS only reads traces afterward.

Measuring instead of claiming

The project publishes its reproducibility and measures with dates and sources, and I try to keep claims smaller than the evidence. It is less exciting than a demo, but it is the only way I know to make the word “result” mean something.

If you work on agent-based models, game simulation or anything replayable, I would genuinely like to hear how you handle the same problem. The code is on GitHub, and issues are open.

About the author

  • Creator and developer of LIVEX, an open-source lab for multi-agent simulation. Notes on engines, observation tools and the surprises that emerge.

No comments yet. Be the first to comment.

LIVEX: Behind the Simulation (EN)

Part 5 of 6

The story of LIVEX in plain English: why it exists, how it works and what it takes to make a multi-agent simulation reproducible.

Up nextWhat if you could watch a society emerge from simple rules?Meet LIVEX, an open source multi-agent simulation built to observe what nobody programmed.