AI Just Spent 88 Hours Solving a Problem Humans Have Wrestled With for Decades. So What?

OpenAI says one of its experimental AI systems has produced a proposed solution to a famous mathematical problem that has challenged researchers for decades.

It took about 88 hours. That is an impressive headline. But the more interesting part is not really the maths. It is how the AI worked.

This was not somebody asking ChatGPT a clever question

According to the reports around the announcement, OpenAI deployed roughly 10,000 AI agents to work on the problem. They explored different approaches, exchanged information, ran calculations and concentrated resources on ideas that appeared promising. Researchers directed the process and checked the results.

That is a significant change in the working model. The system is not only retrieving an answer that already exists somewhere. It is being used as part of a research process where the route to the answer is uncertain.

From answering questions to investigating problems

Most businesses currently use AI for tasks such as:

• writing emails
• summarising documents
• researching subjects
• creating marketing content
• analysing spreadsheets
• generating ideas

Useful, certainly. But these activities mostly involve helping people work with information that already exists. The person provides the objective, the context and often the judgement. The AI makes the work quicker or easier.

The more interesting development is AI being used to tackle problems where the answer is not already sitting in a document waiting to be retrieved. That moves us closer to systems that can be given an objective and allowed to work towards it.

What does a decades-old maths problem have to do with your business?

Probably nothing directly, unless your Monday morning management meeting has recently become unusually mathematical. But the working model matters.

Imagine giving an AI system an objective such as:

Or you might ask:

Or:

That is a very different proposition from asking ChatGPT to write an email.

There is an important catch

“88 hours” sounds almost magical. It was not. Thousands of AI agents were working simultaneously, using substantial computing resources and building on decades of human mathematical research. Researchers were also involved in directing and verifying the process.

This is not evidence that AI has suddenly become an all-knowing machine. It is evidence that the scale and type of problems AI can be applied to are changing quickly.

The same caution applies in a business. An AI system can investigate a problem without understanding the business in the same way an experienced owner or manager does. It may use incomplete data, follow an attractive but wrong explanation or mistake correlation for a useful cause.

The question for businesses is changing too

For the last couple of years, the practical question has largely been:

Increasingly, the more interesting question may become:

Start with a problem, not a pile of data

You do not need 10,000 AI agents or an unsolved maths problem to start thinking about this. Start with a problem that has a clear owner, a bounded set of evidence and a reviewable outcome.

1. 1State the problem in one sentence.
2. 2Define what a useful answer would help you decide.
3. 3Give the system only the data and tools it actually needs.
4. 4Ask it to show evidence, assumptions and alternative explanations.
5. 5Have a person review the findings before changing the business.

For example, “find unnecessary technology spending” is a better starting point than “optimise our IT”. The first has a purpose and possible evidence. The second is so broad that an AI system could produce a confident report without giving you a useful decision.

It is also worth separating investigation from authority. An AI system might be allowed to inspect approved records and suggest an explanation without being allowed to cancel a contract, change a customer price or buy a replacement service.

The IT Club view

The story here is not that AI has solved everything. It is that AI is moving from retrieving, summarising and drafting existing information towards working on complex problems where the answer is not already known.

For an SME, that shift is worth watching because the most valuable use of AI may not be producing more content. It may be helping a small team investigate questions it has never had enough time, data or people to examine properly.

The question to start with is simple:

The Race to Replace: Why AI Needs to Start Doing the Work →

Explore the AI Governance Knowledge Centre →

Want to talk through an AI investigation?

Describe the business problem, the evidence you already have and what decision you would make with a reliable answer. The Ask the Advisor service can help you think through whether the problem is suitable for AI investigation and what checks should sit around it.

Ask the IT Club Advisor →

Found this useful? Forward it to someone who might too.

Sources and further reading

This is original IT Club commentary based on the reported announcement and the business implications of multi-agent research. Claims about the proposed mathematical result, the approximate time and the reported agent count are attributed to OpenAI and the background source below; the result should not be treated as independently accepted without further verification.

Background reading: AI Cracks 90-Year-Old Maths Problem In 88 Hours →