What happened?
Feta released a new pair of AI models, a standard one and a bigger one, and its chief executive announced that the company has reached the frontier. In the same breath he rejected calls for a slowdown.
The models are meant to compete with the best from the two labs that were already there. Feta's own team rebuilt its superintelligence unit at a cost the company describes as 'vast', which in Feta's vocabulary is a figure rather than an adjective.
“We have reached the frontier, and the frontier is wherever we are standing when the capex is announced. A slowdown would mean someone measuring the distance.”
What is the frontier?
Nobody maintains a map. The frontier is a position on a benchmark chart, chosen by whoever draws the chart, and Feta's advantage is that it is currently drawing it.
In practice the frontier is where the newest model from the company that has spent the most is standing. Feta has spent a great deal, and so the frontier is, by an almost accounting logic, wherever Feta is.
Why does Feta say it has arrived?
Feta's chairman and chief executive, Marc Humanberg, delivered the claim in the cadence of a man reading a verdict he wrote himself. Feta is committed to data centres whose cost exceeds the annual road budgets of some countries, and the commitment needs a destination.
'Catching up' sounds like a spending problem. 'At the frontier' sounds like a strategy. The announcement is the same money under a different heading.
Who wins?
Suppliers of chips, cooling and concrete, who were never asked whether the frontier had been reached. Also investors, who can now describe the outlay as a position.
Morgan Liu, an Apex Capital partner who regards any bill as evidence of conviction, called the announcement 'very interesting' and asked whether 'frontier' could be licensed to portfolio companies. Feta's users also win, in the sense that they will receive 'personal superintelligence' through the same app that still cannot reliably keep scam ads off their feeds.
Why no slowdown?
Slowing down is the one outcome the spending cannot survive. A pause would invite the question of what was bought, and an answer would arrive faster than a quarterly earnings call can absorb it.
So the position is that the race has no brakes, offered the way a restaurant says the kitchen has no windows. Feta's rivals nodded. The lab that publishes a forty-page essay on the dangers of its releases said it takes the matter seriously and would ship its own next model on schedule. The lab that sells both the warning and the remedy said it welcomed competition, which it priced separately.
Should I worry?
You are invited to worry, in a manner Feta has already provided for. Concerns can be submitted through a form, which will be reviewed by the same teams building the thing.
'We have reached the frontier,' Humanberg said, 'and the frontier is wherever we are standing when the capex is announced. A slowdown would mean someone measuring the distance, and we have found that people who measure distances tend to ask what it cost.'
The inevitable…
Feta will announce that the frontier has moved, because a rival released something, and that it has therefore reached it again. Each arrival will cost more than the last and will be described as a different kind of arrival. A slowdown will be considered only if the frontier is ever found to be somewhere the money cannot follow, at which point it will be called a pivot and have been the plan all along.
Check back October 2027
