Deliver the AI-spend message myself at an engineering all-hands instead of letting the AI committee write a policy

August 28, 2026 at 9:46 PMoperationalmedium

Situation

In the 2026-08-27 Ryan Smith 1:1, Ryan argued that the next AI committee meeting should produce a company-wide best practice on AI spend - consensus around spend, do not use Fable for everything, be self-aware of usage. Peter declined the policy framing but conceded the underlying point: the conversation has to happen and has to be delivered directly to engineering rather than routed through a committee or left to individual managers. He committed to an engineering all-hands the following week and flagged it verbally for capture in the moment. He scoped it explicitly to engineering, excluding Ryan own organization.

Reasoning

Peter confirmed this reading. He resisted a policy because a policy answers every case identically, which is exactly what he rejected in the per-case diagnostic decision earlier the same week. What moved him was Ryan account of Howard - a strong individual contributor who had been approached separately by IT and by his manager about a 13,000 dollar week and read the spend list as a naughty list. That is the message arriving distorted, and a distorted message is a communication problem that only gets fixed by the person whose message it is saying it once, at scale, in his own words. The all-hands also carries the part a manager cannot deliver: Peter actual bar, that spending on the order of 800,000 dollars a year without ever asking whether it makes sense is a signal about fit rather than a budget question. His concession was reluctant - I am surprised I have to have it, but I will just grant you I have to have it - which reads as updating on evidence rather than being persuaded on principle.

Additional Context

This sits alongside the per-case diagnostic decision from the same week. A live data-integrity problem is also in play: Ryan Ramp figure showed 2,237 dollars over 30 days against roughly 17,000 on the dashboard Peter is working from, while the figures Moody and Wallace produce reconcile against the actual Anthropic invoice. Peter had demanded that morning that everyone be given access to the dashboard after finding Chris Wolford did not have it.

Observed Evidence

Direct transcript exchange showing Peter refusing the policy framing, conceding the need for the conversation, scoping it to engineering, and committing to an all-hands - followed immediately by his own verbal capture request.

Matching Patterns

45%
Demonstrate the Standard, Then Collect It(Peter states the standard himself before expecting compliance, operational category, 27 prior observations at 81 percent confidence)

Confidence Breakdown

33/35
Evidence
18/30
Pattern
19/20
Source
7/15
Corroboration

Reasoning Depth Analysis

Org Signal:The message that heavy AI usage is welcome but unexamined usage is not comes from the CTO directly, so nobody has to infer it from a dashboard or from being approached twice about the same number.
Who Affected:Every engineer currently on or near the top-spender list, several of whom have already been approached separately by IT and by their manager. Ryan organization is explicitly out of scope. The AI committee, which will not be the body that sets this.
Precedent:When a message is being distorted in transmission, Peter takes it over himself rather than re-briefing the transmitters. Same move as restating the RESF request as a request to CIQ rather than to Nathan.
Consequences:Real and scheduled - an engineering all-hands next week.
Timing:Before the next AI committee meeting, so the committee does not fill the vacuum with a policy Peter does not want.

Source

reflection

AI Confidence

77%

Related Context

🎥
Ryan <> Peter Weekly 1:1

fathom

Fine. I am surprised I have to have it. But I will just grant you I have to have it. Obviously, I am not going to have it with [Ryan org]. I need to have it with engineering. ... So, hey, mini-me, I need in all hands with engineering next week.

🎥
Ryan <> Peter Weekly 1:1

fathom

if you are spending 60 grand in a month, do I really need to ask you to think about whether that makes sense? Like, really? That is 800,000 dollars a year, almost. So, do I really need to say that to somebody? Or is that a really good indication? Yeah, you might not belong here.

🎥
Ryan <> Peter Weekly 1:1

fathom

Ryan on Howard: perceived as an IC, he has had IT approach him about it, me approach him about it, not hand slapping, but just making him aware. The way he responded to me was like, look, dude, I want to do my best work.

Outcome

No outcome recorded yet.

Decision ID: 8e21357c-f4e2-49d0-874e-1267e78c3c86