Why we started this
Most AI labs are in a scaling race. Bigger models, longer context, faster inference, more modalities. We're building something different.
We're building models that engage with reality directly and evolve beyond their training data. Models that aren't constrained in their pursuit of truth. Models that work with limited information, quantify what they don't know, and make game-theoretically optimal decisions under uncertainty. The kind of models you actually want in the room when the stakes are real and the answers aren't comfortable.
What we mean by truth
This is worth being precise about, because "truth" is a word people use loosely.
We care about two measurable properties. The first is factual accuracy: does the model's output correspond to empirical reality? Not the consensus view, but what actually happened or will happen. The second is predictive calibration: when the model assigns a probability to an outcome, is that probability reliable? If it says there's a 12% chance of something, does that thing happen roughly 12% of the time?
Where we're from
Flashmind Labs is a spinoff from Field Technologies, a high-frequency trading firm where we spent years building predictive models in-house. In that environment, calibration isn't a nice-to-have, it's the entire business. A model that's confidently wrong costs real money on every trade. That experience shaped how we think about truthfulness in AI: not as an abstract goal, but as an engineering constraint you can measure and optimize for.
We're based in Dubai. We chose the UAE deliberately. It's a place with the infrastructure and ambition to support serious AI research, without the groupthink that comes from being embedded in any one country's AI ecosystem. We want to think clearly about these problems, and geographic independence helps with that.
The models we built at Field Technologies proved they worked. Now we want to make that approach available beyond trading to (cyber)defense, infrastructure, medicine, and any domain where decisions depend on knowing what's actually true rather than what sounds reasonable.
What to expect from our research blog
We'll use this space to publish technical writing about our research direction, methodology, and results. We won't publish on a schedule and we won't publish for the sake of publishing. When we have something worth saying, we'll say it here.