An engineer gets a bug report: “Checkout is broken.” No steps to reproduce. No screenshot. Just a sentence in a ticket and two hours of guessing ahead of them.
A support agent asks: “Can you describe what you were doing?” The customer can't. Nobody can. You don't narrate your experience while it's happening.
Meanwhile, thousands of other users hit the same problem silently. They don't file tickets. They just leave.
The feedback loop between users and engineering teams doesn't exist. That's the problem we're solving.
We named this company after Maurice Merleau-Ponty, the philosopher who argued that perception isn't a report about reality — perception is reality.
You can't understand a user from a metric. You can't understand them from a stack trace or a survey response. You understand them by seeing what they saw — the hesitation, the retry, the moment they gave up.
The closest you can get to being your user is watching what they experienced.
But you have ten thousand sessions a day. You can't watch them all. That's where AI comes in — not to replace observation, but to scale it.
Ponty records every session. AI watches all of them — summarizing what happened, detecting friction, and flagging issues automatically. No manual tagging. No sampling. Every session gets the same attention.
When something goes wrong, AI doesn't just flag it. It generates a test, opens a real browser, and proves the issue is reproducible. The result isn't an alert — it's evidence.
When you ship the fix, the same AI agent re-runs the test to verify it's actually resolved. If your UI changed in the process, the test adapts — no brittle selectors, no flaky scripts.
Detect. Reproduce. Fix. Verify. The loop closes itself.
What we believe
Observation over instrumentation
You shouldn't have to predict what will go wrong and instrument for it in advance. Record everything. Let AI find what matters.
Context over tickets
A ticket that says "checkout is broken" helps no one. A replay showing the exact 14 seconds where a user got stuck helps everyone.
Tests that understand intent
Traditional tests break when a button moves. AI agents understand what the test is trying to do, not where a CSS selector points. Tests should heal themselves.
Close the loop or it wasn't fixed
Shipping a fix isn't enough. If you can't verify the issue is gone for real users, you're just guessing. The loop has to close.
Ponty. Close the loop.