Give the principle then ask for a map of the ChatGPT license ask - fund frontier-tool access for key people, but see the shape first

September 11, 2026 at 3:51 PMoperationallow

Situation

In the Sept 10 weekly 1:1, Steve Wallace raised that Moody had been pinged in one of the AI channels about getting more ChatGPT licenses, said he wanted to run it by Peter before enabling anything, and offered that what he owed Peter first was probably what this would cost. Peter neither approved nor refused. He named a specific person he was comfortable licensing, confirmed Sultan already has one, and asked Steve to get a map of the ask. He then supplied the governing principle: these tools leapfrog each other, so some of CIQ key people having access to the competing vendor tools is critical, and it is fine.

Reasoning

Peter gave the principle before the process, which is what keeps the request for a map from reading as a stall - Steve leaves knowing the answer is almost certainly yes and that the only open question is scope, so he can move without a second approval round. The stated reasoning is competitive rather than budgetary: because the frontier tools leapfrog each other, what is being bought is coverage of the frontier rather than a productivity tool, which is why which people matters more than how many licenses. Naming one specific person he is comfortable approving anchors the category by example rather than by written rule. The decision also sits deliberately against the AI-spend discipline Peter has been pushing in the same period - he distinguishes per-seat access for key people, which he funds readily, from uncontrolled agent token spend, which he does not, and the map is the instrument that lets him tell those apart.

Additional Context

Follows the Aug 28 decisions to diagnose AI overage case by case without squashing usage and to deliver the AI-spend message personally rather than let the AI committee write a policy. Steve had also just shown Peter a Datadog AI costs dashboard covering Anthropic and OpenAI spend, so cost visibility tooling exists and the map request is cheap to satisfy.

Observed Evidence

Direct quotes from the 1:1 transcript giving the named example, the request for a map, and the competitive principle stated immediately after.

Matching Patterns

40%
Purpose Is the Decision Procedure(restates what the spend is for - frontier coverage - rather than adjudicating the individual license requests, same category)
32%
Resource Optimization Through Triage(asks for the cost and shape before committing effort or budget)
30%
Constrain the Mechanism, Not the Access(grants the access readily while asking for visibility into the mechanism and its cost)

Confidence Breakdown

33/35
Evidence
24/30
Pattern
18/20
Source
7/15
Corroboration

Reasoning Depth Analysis

Org Signal:Peter funds frontier-tool access for key engineers readily and still wants to see the shape of the request - cost discipline here is about knowing rather than denying, which is a different message than a spending freeze would send.
Who Affected:Jason Scott, named as approved by example. Justin and Sultan, already in the conversation. Moody, who fielded the original ask. Everyone in the AI channels watching whether the answer comes back yes.
Precedent:Establishes that a spend request of this class gets a principle plus a data request rather than a yes or no, so the next one arrives pre-scoped and Steve can act without a second approval round.
Consequences:Modest and real - Steve owes a cost map before enabling, and the licenses are effectively pre-approved once the map exists.
Timing:Before enabling, because Steve deliberately brought it forward rather than acting - which is itself the behaviour the AI-spend messaging was meant to produce.

Source

reflection

AI Confidence

82%

Related Context

🎥
Steve Wallace weekly 1:1, Sept 10 (recorded as Impromptu Zoom Meeting)

fathom

Steve: Miner got pinged in one of the AI channels about getting more ChatGPT licenses. I wanted to run it by you before I started enabling that... What I owe you first probably is what would this cost? Peter: Yeah, and the people, like Jason Scott, I am comfortable giving one to, Sultan already got one, right? ... So, yeah, let us just get a map of the ask. ... But in general, you know, these tools leapfrog each other. So some of our key people having access to their tools is critical. It is fine.

Outcome

No outcome recorded yet.

Decision ID: dfb3e0a9-591c-4c01-9994-f31f386f1eeb