Model
Your PAR sheet describes the math. Kairos shows how players are likely to respond.
What it is
Kairos treats a slot engine’s spin stream as a language. Each spin is tokenized into control, what the table did; behavior, what the player did; and outcome, what the engine paid. The model learns the player’s next action, not the RNG’s next result, because the engine already knows that. On this basis, it can predict each player’s session length, D2, and wager amount for any math model you put in front of them.
Status
Kairos supports player behavior modeling and simulation. Connect your game-round data to compare configurations for your player population. Loop uses those findings to guide the next round of evaluation and tuning.
From one sheet to the full portfolio
Import a PAR sheet, weight table, or control configuration to check RTP, hit frequency, and cost. Run a browser simulator first, then decide what needs deeper optimization.
Kairos learns player behavior from your spin stream and predicts session length, next-day retention, and turnover. Compare different outcomes by segment under the same math configuration.
Loop simulates candidate designs, compares metrics, and checks budget and risk constraints. Feed the results into the next cycle to continuously manage games and player portfolios.
Read
Tokenization, coarse-to-fine heads, pretraining followed by fine-tuning, and multi-path sampling.
Open769 simulation runs of a real-time dynamic control system.
Open802 runs across ten behavioral mechanisms.
OpenPre-release notes for the Kairos repository.
OpenWork with us
Certified studios
Compare player outcomes alongside the numbers in your PAR sheet.
Dynamic control
Evaluate compensation cost alongside retention to find configurations worth adjusting.
LLM studios
Generate, simulate, validate and refine in one loop.
Start with your next decision
Want to reduce pre-submission iteration, evaluate a compensation budget, or validate an AI-generated design? Tell us which metric you want to optimize, and bring a PAR sheet, weight table, or draft. Together we’ll define the evaluation scope, deliverables, and next step.