Memory with a history
An organisation should be able to recover not just a past decision, but the evidence, assumptions and later events that explain why it was made and whether it still stands.
RESONANCE AI / ORIGINAL IN-HOUSE RESEARCH
Developing a different kind
of artificial intelligence.
Resonance is an active programme of architecture design, engineering and experimental testing. We are investigating how AI could build lasting knowledge, reason from incomplete information, test what it believes and give organisations greater control over their own intelligence and confidential data.
WHERE THE IDEA BEGAN
A convincing answer is not
the same as lasting understanding.
The project began with a deceptively simple question: why can a powerful AI provide a compelling answer today, yet struggle to preserve exactly what it learned, why it believed it and what changed afterwards?
Modern language models have transformed what computers can do with language, code and complex information. Yet a fluent response is not, by itself, a lasting record of evidence. Across a long investigation, relevant history can be lost, conclusions can be difficult to verify, and a new answer may not build reliably on what was established before.
One analogy in those early discussions was a vast library able to produce answers without always showing which book, piece of evidence or later correction an answer came from. Instead of asking how to make a chatbot remember a little more, we began asking what an intelligence system would need in order to develop, preserve and revise its own body of knowledge over time.
That question became Resonance: an attempt to rethink the foundations of artificial intelligence, not simply add another interface to existing models.
THE ORIGINAL CHALLENGE
Three problems worth solving at the foundations.
Answers that can withstand scrutiny
An intelligent system should distinguish a verified fact from a plausible suggestion, recognise conflicting information and explain what further evidence would settle the question.
Intelligence that keeps developing
Useful experience should accumulate. A system should learn from confirmed results and failed attempts, rather than having to reconstruct the same investigation from scratch every time.
THE ARCHITECTURAL DIRECTION
From producing answers
to building understanding.
The ambition is to create an intelligence system whose knowledge and learning do not depend on a single conversation or a particular language model.
Existing AI models remain valuable: they can communicate, interpret language, help generate possibilities and work with specialised tools. Resonance is being designed to complement them with a different foundation for accumulated knowledge, investigation and verification.
Imagine working with an AI on a complex engineering programme for years. It could retain the history of decisions, recognise when fresh evidence undermines an old assumption, connect a new problem to an earlier failure and suggest the next question worth investigating.
Not merely recalling an answer, but carrying the understanding of an evolving problem forward.
THE JOURNEY SO FAR
From a foundational question to a programme of experiments.
Resonance is being developed and evaluated in-house through successive design iterations, experimental work and independent technical challenge.
01
Question the assumptions
Resonance began with an examination of what current model-centred AI does exceptionally well, and where memory, evidence, reasoning and continuing learning remain difficult.
02
Design a different foundation
The work developed into a wider architectural programme intended to preserve knowledge across time, separate proposals from established facts and coordinate specialist capabilities.
03
Build and test the components
The programme moved into engineering, prototyping and controlled experiments, including tests involving incomplete information, the recovery of useful relationships and ways to check candidate conclusions.
04
Test intelligence as a whole
The next challenge is integrating the useful results into a coherent system and determining, through comparison and repeated testing, whether it can investigate and solve problems more effectively than simpler approaches.
WHAT WE ARE WORKING TOWARDS
Capabilities that could change what AI is useful for.
Knowledge that accumulates
Preserve a growing understanding of an organisation, its projects, evidence and decisions. When something changes, retain the earlier history and explain the new position.
Discovery from incomplete information
Begin with partial clues rather than a perfectly formed question. Search for relevant relationships across different information sources and identify leads that deserve closer investigation.
Recognising what it does not know
Keep uncertainty visible, notice contradictions and identify which missing facts, checks or experiments would make an answer more dependable.
Asking better questions
Move beyond waiting for the next prompt. Investigate unexpected patterns, challenge assumptions and identify questions a human investigator may not yet have considered.
Developing and testing new solutions
Form competing explanations or designs, explore alternatives, test them against evidence and retain the lessons from what succeeds or fails.
Enterprise-owned intelligence
Allow organisations to retain control over their private knowledge, operating rules and accumulated expertise, without making that value dependent on one outside AI provider.
WHY SOVEREIGN AI MATTERS
An organisation should own
the intelligence it builds.
The knowledge accumulated by an intelligent system can become as valuable as the documents and data it began with.
Consider a manufacturer whose engineering decisions, test failures, supplier history and operational experience have been built up over decades. The real asset is not merely the original documents. It is the understanding of how they relate, what has been learned and which decisions that evidence supports.
Resonance's long-term enterprise vision is for that accumulated intelligence to remain under the organisation's control. Confidential knowledge could stay within the boundaries it chooses, while the system works with appropriate external tools and models without surrendering its institutional memory.
This could be particularly valuable wherever intellectual property, regulated information, national security or the continuity of expert knowledge matter.
WHY GO BEYOND ANSWERING QUESTIONS
What if AI could help discover
what to investigate next?
Suppose an engineering team is investigating an intermittent failure that has resisted years of conventional troubleshooting.
The evidence is scattered across maintenance reports, design changes, test results and supplier records. Some observations appear unrelated; others contradict one another. The most useful question may be one nobody has thought to ask.
The kind of intelligence we are working towards would retain that history, connect significant clues, keep competing explanations alive and recommend targeted tests to distinguish between them. As those tests produce results, the investigation would become progressively better informed.
The same principle could apply to scientific research, complex industrial operations, software engineering and other disciplines where the valuable discovery lies beyond the obvious search result.
ACTIVE IN-HOUSE ENGINEERING
Research guided by tests,
not demonstrations alone.
Resonance has progressed beyond its original conceptual discussion into a structured engineering and research programme.
Current work investigates how knowledge can be preserved and re-used, how useful information can be recovered from incomplete clues, and how proposed conclusions should be tested before being relied upon. We are also testing how an intelligence system might decide when a quick answer is sufficient and when a problem needs broader, deeper investigation.
Experimental results are compared with simpler established approaches, and the research is subjected to independent technical review. That discipline matters because an unfamiliar architecture is only worthwhile if it produces real improvements in capability, reliability or efficiency.
We are also exploring the longer-term possibility of more efficient foundations for specialised intelligence computing. The immediate focus remains proving useful capabilities through software experiments.
WHY THIS WORK SITS WITH POLARIS
Research and practical work
inform one another.
Our commercial services use today's AI technology. Resonance is where we investigate the limitations of that technology and how a different generation of systems might work.
Designing and testing AI architectures gives Polaris a deeper perspective on the systems businesses are adopting now: which capabilities can be trusted, where careful verification is essential, and what more demanding applications could require.
That combination of hands-on practice and original research underpins the approach we bring to AI training, business integration and technical intelligence.
PRACTICAL QUESTIONS
Research, not a finished product.
Understand what is being investigated and what can be commissioned separately.
Can I commission Resonance AI as a commercial product today?
No. Resonance is an active research programme, not a finished software service or a generally available capability.
Is this research the same as Polaris software assurance?
No. Software Assurance and Technical Due Diligence are separately scoped commercial services. The research objectives described here are not promises about the capabilities of those engagements.
Why undertake research into a new form of AI?
The programme explores questions about persistent knowledge, incomplete information and control of sensitive information. Its proposed mechanisms and future capabilities remain subject to experiments and independent evaluation.
Can you teach us how to manage our own complex AI project?
That is a separate practical offering: Advanced AI Project Engineering & Training. It focuses on organising and directing substantial AI-assisted work, not supplying the experimental Resonance architecture.
YOUR NEXT STEP
Looking for practical AI expertise?
Explore how Polaris applies its experience using and developing AI to real business problems, from team training to carefully scoped integration work.