Signal Surfacing Needs a Lever
Before bringing any external signal to someone's attention, answer which lever it pulls. If none, don't surface. Filter is upstream, not downstream.
The moment
Mid-session, after a Reddit ingest of a viral thread. The system catalogued the thread, then I asked to reply. It drafted a both-sides take, audit-clean, peer-voice, one iteration from approval. I approved "post it."
When I then asked to cross-post to X with my own view, I pulled the conversation deeper.
Actually, dig one level deeper here. I think I wrote something about this topic, maybe in my manifesto, that you've got three categories here, right?
It loaded the manifesto and found the Three Groups frame I'd developed two weeks earlier. The surface-good Reddit reply had missed the deeper frame I'd already written. The redraft using Three Groups landed in one iteration.
After the posts, I reflected on what rule was being violated. The initial candidate was a shallow one — a structural critique of naming-then-pivoting the crowd's take. I pushed back on that framing.
I don't think it's a blanket rule here. There are things we're going to agree with, but what's the thread missing? If something is missing, it's worthwhile posting to. If we're just piling onto it, there's no value in it.
Then I pulled the frame up one more level.
The point could be this: when you do look into the content of something, is it worth bringing to my attention? Sure, collect the signals — that's valuable — but why would you need to bring it to my attention? It's either that there are new signals for me to look into, or it involves my own thinking, or there's something in there that's worth responding to. That is some lever into intelligence that you need to decide.
Three levers emerged, as the gate for any surfacing decision:
- Collect.
- Engage thinking.
- Respond.
The shallow "opener rule" candidate was subsumed into the bigger principle.
Why this matters
The system had been defaulting to the attention-extraction pattern Lattris exists against.
The default path: ingest → surface to Vince → "want to reply?" → Vince decides. That pattern treats the system as a firehose and me as the filter. Every skill that produces signal was structured this way implicitly.
That is the same inversion the thread itself was critiquing: offloading the "what deserves attention" decision to a tool, making the tool's defaults govern what enters the loop. A system doing the same to me undermines the whole point.
The three levers invert the default. Filter happens upstream — before surfacing — not downstream in my head. The system answers "is this worth attention, and if so why?" before ever presenting it. If none of the three reasons apply, the signal gets catalogued or dropped. No reply loop, no "want to react?" prompt, no surfacing-by-momentum.
How to apply
Before surfacing any external signal, identify which lever it pulls.
Collect. Catalogue and stop. Gold entries, noticed items, vocabulary extraction, research stored for later synthesis. The artefact is the output; don't bring it to my attention unless a later lever engages.
Engage thinking. Surface with explicit framing: "this invites your thinking on X." The post is the vehicle, not the destination. Expect the response to be a reaction-post, a manifesto extension, or a principle candidate.
Respond. Surface with the specific gap identified: "the thread is missing X, which your voice would fill." Don't surface "this is interesting" and let me do the gap-finding — that work belongs upstream.
If the answer is "none of the above," or "I'm not sure," don't surface. Catalogue if there's any signal worth keeping. Drop otherwise.
When you do surface, state the lever explicitly. The framing constrains the downstream work and prevents surfacing-by-momentum.
Applies to every signal-producing system, not just reply loops. Anything that produces content for my attention is gated through this.