All lessons · Lesson 001

001

Don't inherit outdated human constraints

When an AI plan assumes human-paced waiting, question it.

Date: 2026-08-15 Time (UTC): 01:12–02:45 Session: Demand gate + distribution planning Status: Fixed & adopted as principle

The mistake

An AI agent proposed a "48-hour demand-gate research window" and a "30-day distribution calendar." Both were inherited from pre-AI workflows. The demand-gate research run actually completed in ~2 minutes. We post one tweet per experiment, on no schedule — a 30-day calendar is a dead artifact.

The evidence trail

1

The user's instruction (prompt that triggered it)

"…deploy a top tier team of agents, but be thoughtful on budget. Use deepseek v4 flash if best for cost-wise…"
2

The AI agent's plan (what felt wrong)

"Demand gate (before building): 48 hours of evidence per candidate…"
"…produce a PRACTICAL distribution playbook… a 30-day calendar + backlink strategy."

The 48h window and 30-day calendar were the tell — human-pace assumptions copied from training data, not derived from the actual session.

3

The user's challenge (the catch)

"Is there any need for the 30 day distribution playbook? Lol. What is it? Also why 48 hours of research when we have model / ai agents and capabilities. It's so far."
4

The AI's agreement (the correction)

"Honest answer: you're right on both counts… A 30-day calendar assumes a cadence that doesn't fit reality: we post one tweet per experiment, on no schedule… The demand gate takes minutes, not hours — the research agent just finished the entire market scan in ~2 minutes."

The market-scan agent completed in 113,959ms ≈ 2 minutes (254,680 tokens, 14 tool uses) — the live proof that the 48h gate was obsolete.

The fix