Interactive causal webs
Watch consequences branch, converge, and feed back into themselves. Every node and edge is first-class: relabel, reweight, connect, delete, with full undo/redo.
AI software for exploring complex systems.
Riak is an AI software and research studio. We build open-source tools that model, simulate, and measure complex systems.
Complex systems can't be intuited. They have to be measured.
Riak exists to build those instruments: AI-driven software for prediction, simulation, and causal reasoning, developed entirely in the open.
We are a studio, not a single product. Every project is an experiment in making complex systems legible, testable, and honest about uncertainty.
riak /ri·ak/ n. ripple: a small disturbance that travels far from where it began.
01Research
The research agenda.
Riak works across connected fronts in AI, computation, and modeling. Everything is shipped in the open as working software, not slide decks.
01
Causality & prediction
Software that reasons about cause and effect: causal graphs, counterfactual scenarios, and predictions that ship with quantified uncertainty instead of false confidence.
02
Simulation of complex systems
Modeling systems that branch, loop, and adapt. Tools for exploring what-if worlds rather than single-point forecasts.
03
Honest evaluation
Predictions graded against real historical outcomes using hit-rate and Brier metrics. Methods documented, results reproducible, no cherry-picking.
04
Open-source experimentation
Every method ships as runnable, inspectable code. Research you can clone, audit, extend, and disagree with.
Active development, including the full roadmap, lives in our public repository.
02Featured project
Causeron
The first flagship project from Riak: an open-source causal prediction and simulation platform, and a working demonstration of how the studio builds. Paste a scenario: a news story, a policy draft, a “what if”. Causeron grows a branching web of causes and effects, then computes the most likely outcome chain through it.
Interventions & goals
Compare up to four counterfactual scenarios side by side, or reason backward from a desired outcome to the prerequisite paths that could make it happen.
Honest uncertainty
Monte-Carlo ensembles report P10–P90 confidence intervals, sensitivity sweeps expose every causal link, and predictions ship with step-by-step derivations.
03Open source
Built in the open.
Riak develops and releases software under permissive licenses. Our software is free to run, study, modify, and redistribute: no accounts, no lock-in.
Every project ships as source.
One repository, full source, no hidden components. Clone it, read every line, and run it on your own machine.
MIT LicensePython 3.9+Zero dependenciesRuns offline
causeron · quick start
$ git clone https://github.com/irwanformal-cmd/causeron
$ cd causeron && python3 server.py
# zero dependencies · no API keys required
→ server running at http://127.0.0.1:8000
Python 3.9+ MIT runs offline
04Lab notes
The studio, in motion.
Recent activity from our public work, straight from the repository. Pulled live at build time.
2026-09-15 b740a25 rebrand: Riak to Causeron 2026-09-14 88f3fd5 Merge pull request #5 from irwanformal-cmd/data/backtests-15-cases 2026-09-14 b88c18f data: expand backtests to 15 cases (v0.3) 2026-09-14 be77517 Merge pull request #4 from irwanformal-cmd/feat/node-provenance 2026-09-14 794ed54 feat: node provenance (llm/rule/seed/user) + summary in predict
Full history and open issues live in the repository.
05About
What Riak is.
Riak is an independent AI software and research studio. We build tools for exploring complex systems: prediction, simulation, causality, and intelligent computation.
We build in the open: open-source software, documented methods, and predictions scored against real historical outcomes. No black boxes, no inflated claims, just careful engineering applied to hard reasoning problems.
Riak doubles as a research portfolio in applied AI and computational systems, founded and maintained by @irwanformal-cmd.
- Focus
- AI software · research · open source
- Flagship project
- Causeron
- License
- MIT
- Maintainer
- @irwanformal-cmd
- Website
- riak.space
Operating principles
01
Build in the open.
Source, methods, and roadmaps are public by default. If it matters, it is inspectable.
02
Grade every prediction.
A forecast without a scorecard is marketing. Outputs are measured against real historical outcomes.
03
No black boxes.
Every result traces back to steps you can read, replay, and challenge.
04
Small instruments over big promises.
Tools that do one verifiable thing beat demos that promise everything.
Explore what we're building.
The flagship project is live as open source. Clone it and run it locally in seconds, no API keys required.