Coding errors in a dating app cause frustration, but bugs in an airplane or a medical device can cost lives. London based startup Hypercritical just secured £2 million to solve this high stakes problem using a new breed of artificial intelligence that uses math to guarantee safety. Their technology promises to fix the hallucination issues plaguing current AI models in heavy industries like aerospace and defense.
Big Backing for a Serious Problem
Hypercritical has successfully closed a £2 million pre-seed funding round led by Join Capital. This deeptech startup also attracted attention and capital from heavyweights like Octopus Ventures and the Tiny Supercomputer Investment Company. Plug and Play also joined the round to support the vision of safer industrial automation. The company plans to use this fresh cash injection to double its engineering team and buy the massive cloud computing power needed to train its proprietary models.
This investment comes at a crucial time for the software industry. We have seen how fragile global systems can be. While popular AI tools like ChatGPT or GitHub Copilot are great for writing emails or basic web scripts, they are famous for “hallucinating.” They confidently make up wrong answers or write buggy code. That is a deal breaker for a company building self driving cars or nuclear reactor controls.
Hypercritical is building a system that does not guess; it proves.
hypercritical ai industrial control software funding round london
“We are moving beyond the era of AI that merely suggests code to an era where AI guarantees it,” the team noted regarding their mission.
The involvement of Join Capital is significant. They are known for backing industrial tech firm. Their support signals that the European market sees a massive gap in how we build software for physical machines.
Stopping the AI Hallucinations
The core innovation here is distinct from the large language models most people use today. Hypercritical uses a logic driven architecture. This means the software is built on strict mathematical rules rather than just predicting the next likely word or line of code.
Standard AI models operate on probability. They ask themselves what piece of code probably comes next. Hypercritical’s system functions as a domain specialized agent. It designs and verifies software within mathematically defined safety limits. If the code does not pass the strict logic tests, the AI does not write it.
Here is how the two approaches differ:
| Feature | Standard Generative AI (LLMs) | Hypercritical’s Logic AI |
|---|---|---|
| Method | Probabilistic (Guesses best fit) | Deterministic (Math based rules) |
| Reliability | Prone to hallucinations/bugs | Mathematically guaranteed correctness |
| Use Case | Creative writing, web apps | Aerospace, Defense, Robotics |
| Verification | Requires heavy human review | Automated verification built in |
This “correct by construction” approach is vital. It allows engineers to focus on what the system should do, rather than getting bogged down in line-by-line debugging. The system ensures the output is bug free before a human even looks at it.
Speeding Up Heavy Industry
Industries like automotive, aerospace, and defense are notoriously slow to update their software. This is not because they are lazy. It is because of regulation. Every line of code in a passenger jet must be certified. This process costs billions of dollars and takes years.
Hypercritical aims to slash these costs. Their flagship product is called Hyperpilot. It is already being used by engineering teams to automate the development of control software. The promise is simple but powerful: define the tests the system must satisfy, and the AI generates the algorithms to meet them.
- Cost Reduction: Automated code generation cuts down engineering hours.
- Faster Deployment: Passing tests automatically means faster regulatory approval.
- Safety: Mathematical proof reduces the risk of catastrophic failure.
This technology turns the testing phase from a bottleneck into a blueprint. Instead of writing code and then testing it, engineers define the tests first. The AI then builds the software specifically to pass those tests. It is a complete reversal of the traditional workflow.
A Future Without Coding Errors
The vision extends beyond just making work easier for coders. Hypercritical wants its methods to become the new gold standard. They aim to have their technology incorporated into ISO standards. This would modernize global software certification.
Currently, certification is a manual, paper heavy nightmare. If Hypercritical succeeds, we could see a future where “compliance” is just another automated step in the software pipeline.
The company divides its tech into two categories:
- Copilots: These act as specialized assistants, like a QA engineer or a systems engineer, helping human teams spot issues immediately.
- Autopilot: This produces unsupervised software that passes 100% of tests without human intervention.
This dual approach allows companies to trust the system gradually. They can start with the Copilot to assist their teams and move to Autopilot as they see the results. With the new funding, the race is on to prove that this logic driven approach can scale across the most demanding industries on Earth.
The London based team is hiring now. They are looking for the brightest minds to help build an industrial future where software failure is a thing of the past.
