Refuse to write the RLC-AI contrivance down - say it to Justin, Nathan and Andrew, never in a company-wide channel

September 11, 2026 at 8:09 PMoperationalmedium

Situation

Nathan asked Peter to post the full rationale for the deliberately contrived RLC-AI Spark build into the public RLC-AI channel. Peter declined outright and said he would say it to Justin, to Nathan and to Andrew, but would not write it anywhere the whole company can see. The same boundary appears in Slack the same morning, where he told Brett that he wanted the message going to Brett, Andrew and Jeff rather than wider, and said plainly that it is not a message for the whole company.

Reasoning

The build is intentionally not what CIQ would normally ship - latest driver, one benchmark, no long-term supportability - and the reason it is acceptable is a narrow, time-boxed commercial goal. Written down and broadcast, that rationale reads as a standard rather than an exception, and it is exactly the kind of thing that outlives its window. Keeping it verbal to the three people who must execute it preserves the exception without eroding the norm.

Additional Context

Runs alongside the earlier decision not to publicly defang Dawson in the product channel. Peter told Justin the reason plainly: because we are calling this thing RLC-AI, and because we are playing a little fast and loose with what RLC-AI is.

Observed Evidence

Direct quote from the Fathom transcript plus two corroborating Slack messages the same morning setting the same distribution boundary.

Matching Patterns

45%
Small Circle for Sensitive Operations(limits information to the people who must execute)
40%
Purpose Is the Decision Procedure(keyword match, same category)

Confidence Breakdown

33/35
Evidence
22/30
Pattern
19/20
Source
13/15
Corroboration

Reasoning Depth Analysis

Org Signal:An exception is allowed to be an exception. Peter will personally absorb the cost of repeating himself verbally rather than let a shortcut become a written norm.
Who Affected:Everyone downstream of RLC-AI who would otherwise read the rationale as the new bar; Andrew Jorgensen in particular, who Peter described as really uncomfortable and who internalises it once it is explained in person.
Precedent:Sets the precedent that the written record is a standards artifact, not a transparency artifact. A thing can be fully explained without being posted.
Consequences:Real cost - Peter has to hold the same conversation three or four separate times rather than once, and accepts that as the price.
Timing:Now because the build has to ship this weekend and the temptation to broadcast the reasoning once is highest exactly when the deadline is tightest.

Source

reflection

AI Confidence

87%

Related Context

🎥
Justin <> Peter Weekly 1:1

fathom

Nathan said, hey, can you put all of this like in the RLC-AI public chat? And I am like, no, no, I cannot. I will not put this in a place where the whole company can see it. So I will say it to you. I will say it to Nathan. I will say it to Andrew. I will not write it down publicly.

💬
Group DM with Nathan Blackham and Brett Mastbergen

slack

No I do not want it that wide ... I will post it to you and Andrew and Jeff. Or you can. ... But it is not a message for the whole company.

Outcome

No outcome recorded yet.

Decision ID: e2be4382-e2f0-4a1f-9a48-9e1f67d56f79