About
We’re building an intelligence that understands intelligence.
Diablo’s job is to understand AI systems, and to make them better in ways you can check.
Why
AI changes every week. Understanding doesn’t keep up.
New models, new prompts, new tools. When an AI system starts behaving differently, most teams guess, patch and hope.
Diablo replaces the guess with an investigation: the possible causes, an experiment that tells them apart, and a conclusion only as strong as its evidence.
What Diablo is
Not another tool. A new category: AI evolution.
Evaluation, observability, dashboards and coding agents are ingredients. The product is the loop that turns them into understanding, and understanding into improvement.
Principles
Four rules we don’t bend.
01
The AI reasons. The system measures.
The model proposes and explains. Code runs the experiments and computes every statistic.
02
Turn uncertainty into knowledge.
Start from what is known and what isn’t, not from a fix.
03
Every investigation leaves a record.
Hypotheses, experiments, verdicts, and the raw outputs behind them.
04
Every change must be undoable.
Any change Diablo makes will be explicit, tested, reversible and audited.
Where it’s going
From answering questions to improving itself.
- Now
A working loop, on sample systems
Live on a real model: a study designed for your question, paired experiments, real statistics and a checked answer.
- Next
Your own AI system
A live reasoning model, and a connector that runs experiments on the AI you ship.
- Then
Autonomous discovery
Diablo notices that something changed or looks unusual, investigates on its own and remembers what it found.
- Later
Self-improvement
Diablo applies the same loop to itself: its reasoning, planning, memory and choice of experiments.
Spin it. The better Diablo investigates, the better it gets at investigating.
See the loop for yourself.
Ask Diablo a question about your AI. Sign in with Google and watch it investigate.