Semantic Reasoning Labs
We are building a new foundational model.
AI you can trust by design: inspectable, explainable, deterministic.
New to neurosymbolic AI? Start with the two-minute explainer.
Who we are
- We are a frontier AI research team based in Prague.
-
We do not work on larger or better language models,
but on a fundamentally different foundational model to complement LLMs where they cannot be trusted. - Our mission is to make deep reasoning inspectable, explainable, deterministic, and order of magnitude cheaper.
Imagine an AI where
- Knowledge lives in an explicit, human-readable form rather than in a set of weights, and you can inspect and audit what the model has learned.
- Reasoning is mostly an algorithmic and produces a trace of actual steps taken. All reasoning steps are thus explainable.
- The search is deterministic. Same question, same answer, every time.
- When something is wrong, you can easily fix the model with O(1) complexity, so the same failure does not recur.
- This also means, the entire knowledge base can act like the task context. Therefore, the context is limited only by the physical memory.
- Inference is order of magnitude cheaper and can thus run on existing hardware locally.
FAQ
Consider a student learning calculus. The data she needs are modest: a textbook, lecture slides, and at most a few hundred examples (so at most a few tens of MB of data in total). As for the compute required, the human brain runs on roughly 20 W of energy. So even if the student spent all waking and sleeping hours of an entire semester learning just calculus (unlikely), her brain's total energy budget would power a single Nvidia B200 for approximately 2 days.
It is clearly physically possible to learn from much less data with much lower compute than with current GPT-based approaches. The method to do so had simply not been discovered yet.
Much of modern machine learning resembles dog training: trial --> reward --> reinforcement until the desired behavior emerges. It works and is also an essential part of human learning. But it is not how humans typically learn structured high-level knowledge.
What comes out is a distribution over tokens. An LLM does not know what it knows and what it does not; when it hallucinates, it is because it cannot tell the fabrication from the truth. Harnesses, tools and retrieval improve it a great deal, and they get you to 95%. However, no harness closes the last mile.
Humans read. Humans think. Sometimes new concepts click immediately, sometimes only after revisiting the material or finding a different explanation better matching our existing mental model. At Semantic Reasoning Labs, this is the sort of learning and thinking we are trying to recreate computationally. Two kinds of learning building two kinds of thinking: that is the neurosymbolic idea, and the explainer walks through it.
No. Our technology cannot write a haiku in the style of Shakespeare. For code, copy, video and conversation, LLMs are the right tool and will stay so.
We complement LLMs brilliantly for the other kind of work in: contracts, compliance, tax, legal, pharma, regulatory and scientific environments, where 95% correctness will not do.
Current LLMs take a conceptually different path. It is not possible to break their fundamental limitations by hiring thousands of engineers to tweak them incessantly. We need a fundamentally different approach.
That is where we come in. A team that has background in research, engineering, philosophy and AI. The team that has spent years thinking differently about AI. The team with the conviction and depth to build something genuinely new.
Extraordinary claims require extraordinary evidence
We agree. Will a demo showing deep reasoning (on a single topic) running on a laptop count? Let us know: curious (a) semanticreasoning.ai