Hey {{first_name | there}},

We keep hearing about AI models getting smarter, faster, and better. But what happened recently might be much bigger than another model release.

So yesterday, OpenAI announced three major scientific discoveries in mathematics, computer science, and physics.

And one of them could change something researchers have believed for nearly 30 years.

OpenAI claims its AI has disproved the Exponential Time Hypothesis (ETH).

Now, before you skip this because it sounds too technical, hear me out. This is actually pretty interesting.

Imagine you're trying to solve a really complicated puzzle.

As the puzzle gets bigger, the number of possible solutions grows ridiculously fast. Even a powerful computer could take millions of years to solve certain versions of it.

That's basically what the ETH is about.

It says that certain computational problems, including a famous one called 3-SAT, can't be solved as quickly as we'd like.

Researchers have relied on this assumption for decades to understand the limits of algorithms.

It produced a research paper describing an algorithm that could solve these problems in subexponential time.

In simple words, the algorithm could be theoretically much faster than what ETH says is possible.

There's a catch, though.

The proof has been checked using a mathematical verification system called Lean, but independent experts are still reviewing the broader claim.

And no, this doesn't mean your laptop can suddenly crack impossible problems overnight.

But if the result holds up, a major assumption in computer science will need rethinking.

OpenAI also announced AI-generated results involving the Kakeya conjecture and quantum chromodynamics.

Pretty wild, right?

Forget the complicated math for a second.

We're watching AI move from answering research questions to actually producing new research.

Think about that.

Until recently, you could ask ChatGPT to explain a scientific paper.

Now, frontier AI systems are producing original mathematical arguments that researchers have to sit down and evaluate.

And that's where I think a huge opportunity is opening up.

Because once AI can help generate research, we need a whole new set of tools to support it.

1. AI research assistants

I'm not talking about asking ChatGPT to summarize a PDF.

Imagine AI that can read thousands of papers, spot missing connections, suggest hypotheses, and help researchers test them.

We're already seeing this happen.

IEEE recently launched its AI Research Suite, which searches millions of peer-reviewed publications and provides answers linked to actual research.

2. AI research verification

Here's the problem with AI-generated discoveries.

Someone still needs to check whether they're correct.

That's where proof assistants, formal verification tools, and automated research-checking systems come in.

If AI starts producing thousands of research claims, tools that can reliably verify them become incredibly valuable.

3. Turning discoveries into practical products

Imagine you're running a logistics company.

Somewhere, a researcher discovers a new algorithm that could improve route optimization.

Would you even know about it? Probably not.

Now what if there’s a tool that monitors new research and tells you exactly which discoveries could affect your industry and how.

I'd seriously consider building something like that.

I'd start by asking one question:

What does my industry currently consider too difficult, expensive, or time-consuming to solve?

Maybe it's optimizing deliveries, reviewing legal documents, discovering new materials, or analyzing complex financial data.

Then I'd ask:

Could better algorithms, cheaper AI, or automated research make any part of this easier?

Not every impossible-looking problem has a computational breakthrough waiting around the corner.

But I wouldn't automatically assume today's limitations will stay the same.

Because that's what excites me most about AI right now.

It's not just getting better at doing things we already know how to do. It's starting to help us discover things we didn't know before.

And the people who can ask good questions, recognize useful results, and actually test their ideas will have a massive advantage.

So I'm curious.

Does any of this change what you're planning to build next?

Hit reply and tell me. I'd love to hear what you're working on.

- Aashish

Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.

Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.

Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.

Try Notetaker Free