Mesocort AI (@MesocortAI) header
Mesocort AI (@MesocortAI) profile picture

We build AI models that understand and predict human behavior under pressure.

Joined September 2026
Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)Photo by Mesocort AI (@MesocortAI)
Lose a round in CS2 and your voice chat turns into a crime scene. We sat through 476 hours of CS2 voice messages... Transcribed every line, flagged every f-bomb in any language, pinned each one to its round.
1
2
41
In games, players shout info to each other all the time. But how often is it actually true? We built ECHO to find out. It listens to the whole team, matches every word to what really happened in the game, and can even tell who will win just from how people talk.
1
6
99
Can you tell what someone will do under pressure, before they actually do it? We tried it in Counter-Strike 2, where players get hit from behind all the time and have a split second to turn and fight or run.
2
1
3
191
What do a store robot and a Counter-Strike player have in common? Both keep losing sight of someone who is still moving. So we taught a robot with player data to guess where that person is now.
1
8
92
A CS2 bot you control with words, not code 👇
1
2
174
We ran a two-day hackathon where everyone signed up for the topic they wanted to dig into. Seven teams finished with seven projects. None of them is going on a shelf 🫡 We're already working out how to implement every one and over the next days we'll take them one by one here.
1
5
80
Everyone has a dataset now. The question is how much you can pull out of a single second. A woman turns around on a busy street. Traffic, music, a dog, voices... We check them all. One second holds more information than you might think. Not just what happened. But WHY.
1
1
2
135
A cheater often looks like a very good player. Even experts can't tell from a replay. So we stopped asking "does this look like cheating?" and asked: did this player act on information no human could have had?
1
1
7
159
We’ve customized the drone simulator based on yesterday’s @Tesla presentation. We take staying ahead of the curve seriously. 🫡
Real life, baby
3
55
We're building tools to explore patterns in 50M+ hours of human activity data. This helps us move from massive amounts of raw data to the moments that actually matter. How do people make decisions, how do they react, and what influences their behavior in critical moments?
3
3
103
Imagine having millions of hours of driving data. Instead of going through it all manually, you could simply search for situations where a driver had to react suddenly to a pedestrian. This is exactly the kind of technology we're working on at Mesocort AI.
12
Similar situations can happen to different people around the world on the same day. And everyone might react differently. But what if we could find, compare, and analyze all these situations?
11
Imagine you're driving home from work. Suddenly, a pedestrian steps in front of your car, and you have only a split second to react. Some people immediately hit the brakes. Others instinctively turn the steering wheel. Some might react a little later.
43
Counter Strike only gives us part of the story. So we built our own visibility engine to fill in the missing data and reconstruct what players actually see. 👉 More accurate data, more precise stats and a better understanding of what really happens in-game.
2
52
What happens when you train an AI to navigate in CS2 and then let it loose in a real office? Well, we tried it. Our AI learned to navigate in CS and then adapted to a completely different environment in under an hour. Unlike some of our colleagues, who are still figuring it out.
1
2
9
120