SuperTuned
Site map
- Home — https://supertuned.ai/
- Team — https://supertuned.ai/team
- Contact — https://supertuned.ai/contact
- hello@supertuned.ai
01. The bet
A specialist model that runs inside the live-ops loop.
Stealth until summer 2026.
02. What we do
We've trained a model on what actually happened.
SuperTuned is post-trained against reward signals derived from 200 million players of real behavioral telemetry — sessions, retention curves, monetization funnels, balance outcomes, live-ops decisions that worked. The model is tuned on the underlying operator preference, not on the surface artifacts that produced it.
Trained on what shipped, what stuck, and what converted.
03. Why it has to be us
Four prerequisites. One team that has all four.
Data
200 million players' worth of live behavioral signal. Not licensed, not scraped, not borrowed. The by-product of running the games.
Domain
Twelve years operating live economies. Sixteen years shipping games. Co-founders for twenty-four years. Every company since 2002.
Method
A specialist post-training stack run by our training partner. Reward modeling and GRPO against operator-elicited preference signals.
Distribution
First commercial deployment, summer 2026. Stealth until launch.
Most teams claiming to build AI for games have one of these. We have all four.
04. The model
Ambient. Embedded. Inside the live-ops loop.
SuperTuned doesn't sit in a chat tab. It runs against live telemetry inside the game's own loop — sessions, missions, purchases, events, live-ops calls — and surfaces what matters before anyone asks. A specialist model post-trained against the multi-dimensional preference real operators run their games on — retention × monetization × fairness × feel. As the games run, the signal sharpens, and the model with it.
- Players in the training signal: 200M+
- Shipping games: 16 years
- Operating live economies: 12 years
- Founding partnership: 24 years
- Prerequisites met: 4 / 4
- First deployment: Summer 2026
05. On the name
Tuned is fitting to the past. Super-tuned is fitting to the preference behind it.
Most specialist models are trained on (input, output) pairs from historical work. They curve-fit to the surface — the artifacts a team shipped, not the judgment that produced them. SuperTuned is post-trained with GRPO against an operator-elicited reward signal that captures the multi-dimensional preference real operators run live products on — retention × monetization × fairness × feel.
- TUNED: Fitted to historical designs. Strong inside the portfolio. Weak outside it.
- SUPER-TUNED: Fitted to the underlying preference. Generalizes across genres, formats, and stacks.
06. What's next
A tuning layer for the games industry. Quietly.
One model goes live in August. What comes after is still taking shape. If you're running live signal of your own, we should talk.
Team
Two founders. Twenty-four years.
Roby John and Navneet S Waraich ("W") have co-founded together for twenty-four years, every company since 2002. SuperTuned is what they've been building toward.
- Roby John — CEO. Operator. Sixteen years shipping games. Twelve years operating live economies. https://www.linkedin.com/in/robyjohn
- Navneet S Waraich ("W") — CTO. Architect. Two decades on distributed systems and game infrastructure. https://www.linkedin.com/in/nswaraich
Contact
Talk to us.
We reply within one business day.