Isaac Correa · Madrid

Bets, with dates

I am Isaac Correa, co-founder of Hellomatik. A prediction without a date is an opinion, and an opinion cannot be wrong. This page keeps the predictions I have put a date on, each with where I said it and how anyone can check it when the date comes. Nothing here is edited after the fact: if a bet turns out wrong, the row will say so, with the date it was settled.

Four bets so far, all of them due in 2028. They are the ones already published in my own words elsewhere; this page only collects them.

1. From 2028 on, an ordinary computer, the one already on the desk, will run a model worth using.

Said
13 August 2026, on X, in a post about why the model that ends up everywhere is not the smartest one. The post.
In full
“My date is 2028. From then on I expect an ordinary computer, the one already on the desk, to run a model worth using. Something fascinating, that a company can put its real work through without feeling it settled for less.”
Check
In 2028, take a computer bought for an office in 2026 and run a company’s real work through a model that runs on it alone. If a company would do that without feeling it settled for less, the bet held.
Status
Open.

2. Products that took the easy road, an existing tool plus a touch, will converge on each other and compete on price.

Said
September 2026, in the working paper The cliff of innovation, DOI 10.5281/zenodo.22704763, section 8.
Why
Because the parts they share are the same parts and the touch is small.
Check
Compare the feature lists and prices of such products in 2028 with those of 2025. If the lists have converged and the prices have fallen, the bet held.
Status
Open.

3. The durable value in this category will belong to whoever holds a company’s written map: its truth, rules, reversibility and permissions.

Said
September 2026, in the same working paper, section 8. This bet favours my own company, and a reader should weight it accordingly; the paper says so in the same sentence.
Why
Because those four things are specific to each company and cannot be bought, so the party that has written them down will be hard to replace even when the models underneath are swapped.
Check
In 2028, ask companies that changed their AI supplier what they kept. If what they kept was the written account of how they decide, the bet held.
Status
Open.

4. The ceiling of a deployment will correlate with the number of people in the company who ask the system for new things, more than with the model, the budget or the size of the company.

Said
September 2026, in the same working paper, section 8; the argument behind it is the last person.
Check
For any set of deployments, count requests per person per month against the number of distinct tasks automated. If the first explains the second better than the model version does, the bet held.
Status
Open.

How this page works

A bet enters this page only after it has been published somewhere with a date on it, in my own words: a paper with a DOI, a post, a page on this site. The row links to that place, so the wording here can be checked against the original. When the date comes, the status changes to held or missed, with a sentence on what happened, and the original wording stays. The long version of the reasoning behind bets 2 to 4 is in The map of a company.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419.

Read more

The map of a company — the long essay, in nineteen sections with notes.
The cliff of innovation — the working paper the three dated predictions come from.
The machine of the future — written in November 2025, on what the next tool is.
The people I look for — two signs, neither of them on a CV.
The questions people ask, answered — short answers, each with a source.