Festina Lente

There is a particular feeling that arrives when the industry ships, in public, the thing you have been quietly convinced of for years. It is not vindication. It is closer to vertigo. Eleven of the largest names in technology put their signatures on a discovery specification for the agentic web, and somewhere in the negative space of that document is the exact problem I have been circling since I was an undergraduate. The honest first reaction is not I was right. It is I should move, now, before the window closes.

That reaction is the one to distrust.

Make haste slowly. I have used the phrase so often it risks becoming decoration, so let me say plainly what it means when the ground is actually moving. It does not mean be slow. The "haste" is real — there is genuine urgency, the field tests show almost nobody has deployed yet, the early-mover ground is unclaimed today and will not be unclaimed in a year. It means: let the speed live in the thinking, and let the slowness live in the committing. Move fast through possibilities. Move slowly into positions you will have to defend or maintain.

The race I am watching has four layers and several camps, and the temptation it produces is to pick a fight on the layer that is loudest. ARD is loud right now. Google and Microsoft are loud. The absence of Anthropic and OpenAI from the signatory list is loud. And the loudest thing of all, to me, is the layer underneath that nobody announced — authorization, scoped and revocable, the thing I have called the Warrant. The pull is to sprint toward it and plant a flag while the attention is there.

But planting a flag is a commitment, and festina lente asks the question commitments deserve: what am I actually right about, versus what merely feels confirmed? I am confident the authorization layer is real and unclaimed. I am not yet confident about the shape of the artifact that wins there — whether it is a protocol, a reference implementation, an essay that shifts how a standards body thinks, or something I have not seen yet. Those are different bets with different costs, and the confirmation I felt this week is not evidence for any one of them in particular. Confirmation that a layer exists is not confirmation of what I should build on it.

So the discipline, concretely, this week:

Write the essay — because writing is the cheapest way to pressure-test a position, and a published argument is reversible in a way a shipped product is not. Words commit you to a claim, not to a roadmap.

Support ARD where it costs little — a catalog file is twenty minutes and it puts me in front of registries before they arrive. Low cost, real option value, no lock-in. That is exactly the kind of fast move festina lente permits.

Do not yet build the Warrant against ARD as a product. The pull to do that is strong and it is mostly adrenaline. The layer will still be open in three months. The question of what belongs there will be clearer in three months. Building now would be building from the feeling of confirmation rather than from a decision I have actually earned.

I know my own failure mode well enough to name it: when I feel the discomfort of a closing window, I reach for a build. The build is the anesthetic. It converts an unresolved strategic question into the comforting motion of making something. Festina lente is, for me, mostly a defense against that — a way of holding the discomfort open long enough for the actual decision to surface, instead of paving over it with activity.

The race will reward speed. It will reward, more, the people who were fast about the right things and slow about the irreversible ones. The front desk is being built in public this month. The layer beneath it has been open for nineteen years and will not close this quarter. There is time to be exactly as deliberate as the decision requires — and no faster.

Make haste. Slowly.

---

See the companion essay, The Front Desk and the Warrant.