San Francisco-based AI firm Logical Intelligence has introduced an energy-based (EBM) reasoning system called Kona 1.0. The model will be tested in pilot programmes with select partners in advanced manufacturing, energy and semiconductor industries later this quarter.

The company has also appointed Yann LeCun, former chief AI scientist at Meta, as founding chair of its Technical Research Board. Patrick Hillmann, formerly a director at General Electric and previously Chief Strategy Officer at Binance, will serve as the company’s new Chief Strategy Officer.

Kona 1.0 is designed to learn by recognising and correcting its own mistakes, rather than relying on probabilistic systems. The company claims that this approach marks a significant departure from traditional AI models and may signal early steps towards developing general intelligence (AGI).

Logical Intelligence has released a live demonstration of Kona 1.0 on its website, featuring head-to-head sudoku challenges against leading large language models.

“We are seeing in Kona the first credible signs of AGI,” said Eve Bodnia, founder and CEO of Logical Intelligence. “It is not the end state, but it is a clear break from narrow AI.”

Logical Intelligence develops AI systems for applications where behaviour must be provably correct rather than statistically likely. The company’s energy-based systems work by mapping out what is allowed and what is not, then finding solutions that stay within those boundaries.

The architecture is aimed at sectors where certification, liability, and auditability are prerequisites for deployment. Kona 1.0 extends the company’s existing work in formal verification and verified code generation, which has been used to prove the correctness of software components in adversarial and regulated settings.

Logical Intelligence’s leadership team includes Fields Medalist Michael Freedman as Chief of Mathematics and ICPC World Champion Vlad Isenbaev as Chief of AI.