Skip to main content
Next Wave
An early draft of the future

An AI-researched letter for founders and investors, every second Tuesday. It finds ideas worth building or backing while they are still small, tests each one against a sceptic, and keeps score on its own predictions.

You will get one email asking you to confirm. After that, the letter every second Tuesday and nothing else. See the privacy notice and terms.

Free  ·  Every second Tuesday  ·  Unsubscribe any time

Researched and written by AI. No human reads an edition before it is sent.

Why it exists

The information was there years before the price was

In July 2010 a bitcoin cost less than ten US cents. In June 2018 OpenAI published a language model and almost nobody outside the field noticed.

Neither was a secret. Both were being argued about in the open by small groups of serious people. Most of us were reading something else.

Next Wave is built to find that gap: the distance between how big something will become and how little has been built or priced so far. Sometimes that means a fringe idea. Sometimes it means a trend everyone knows, followed two or three steps past the headline. If AI needs power, what runs short next, and what follows from that?

How an edition is made

It reads every week and writes every fortnight

1
It reads where ideas first surface

Software collects several hundred things a week, none of them press releases, and an AI reads all of them.

preprints · grant awards · Hacker News · Reddit · niche forums · new code · prediction markets · draft standards · working researchers on X

2
It asks one question

How much of this has already happened? Not “do people know about it”. An idea everyone talks about can still be early if little money or adoption has followed the talk.

3
A sceptic attacks every idea

A second AI, with no stake in the first one’s work, tries to kill each one: already priced, no reason it is happening now, or a claim the sources do not support. What it cannot kill goes out, with its best objection printed beside the pitch.

4
It keeps score

Every idea carries a dated prediction with a probability. When the date comes, the letter reports whether it was right or wrong.

What you get

Three to five of these, every second Tuesday

This one is from the first edition, shortened to fit.

Fringe

3. Drying with ultrasound instead of heat

Magnitude  ●●●●●Asymmetry  ●●●●●Evidence  ●●●●●Mainstream  22/100

Ultropia’s NSF award, dated 14 September 2026, lists US$1,243,571 in estimated total funding. The company is developing modular equipment that uses ultrasound, sound above human hearing, to remove water from industrial materials.

Why now. The new second-stage award funds the move from laboratory tests to a modular industrial trial. The change to test is whether the equipment works inside a real production line.

The secret

Removing water before the final heated drying step could cut total energy costs and increase output. A module fitted to an existing line could do that without replacing every dryer.

Ways to play
To build
Measure performance and installation economics for one material stream.
Picks and shovels
Power-control electronics, servicing for sound-producing parts and moisture measurement.
The sceptic
No repeatable full-line advantage has been published against a competitor’s five reported pilots. Uneven contact, clogging, maintenance and final heating can consume the claimed savings.
45%likely
Prediction · by 30 Sep 2028
A named industrial pilot will publicly report at least 25% lower total drying energy per kilogram of water removed, at matched output.
Why it is different

It argues with itself, then shows you the argument

The objection is printed

Most newsletters sell you the idea. This one prints the best case against it in the same breath, so you can disagree with either side.

The predictions are dated

“This will be big” cannot be wrong. “45% likely by 30 September 2028” can. Every idea gets one, and the results are reported.

The record is the point

Over years it builds a record of which ideas moved from the fringe to the mainstream, which died, and what the early signs looked like.

Who writes this

An AI does, and nobody checks it first

This is not a newsletter curated by a person. Every edition is researched and written by AI, attacked by a second AI whose only job is to reject, then assembled and sent by plain code.

I built the system. I do not write, choose or approve what it sends, and I do not read an edition before you do. It can be wrong. Every idea links to its sources so you can check it, and that is what you should do before relying on any of it. If you find an error, reply to the letter. Replies go to me, not to the AI: it never reads them and never answers them. Errors I confirm are corrected in a later edition.

It is general information, not financial, legal or investment advice. It recommends nothing and knows nothing about your circumstances. See the privacy notice and terms.

There is also no track record. The first edition went out in September 2026 and nothing here has been proved right. That is why every idea carries a dated prediction with a probability: in two years you will be able to check whether this was worth reading.

What would you do differently if you had read about it two years early?

You will get one email asking you to confirm. After that, the letter every second Tuesday and nothing else. See the privacy notice and terms.

Free  ·  AI-generated  ·  Every second Tuesday  ·  Unsubscribe any time