Pressure-test a strategy you own

The most valuable thing an agent does for a product manager is not writing. It is being the only party in the building with no stake in your being right.

Used badly, this produces polite generic objections. Used well, it produces the question you have been avoiding.

Step 1 — State the position so it can be wrong

An attackable strategy has a claim, a bet, and a thing it refuses to do:

Our activation problem is not discoverability, it is that a new workspace is empty. We will invest the next two quarters in getting one real piece of work into a workspace within ten minutes of signup, and we will not build a guided tour, because tours solve the problem we do not have.

Write yours in that shape. If it does not contain a refusal, it is not a strategy.

Step 2 — Attack it three ways, separately

• Evidence. “Using only @research/, which parts of this are supported, which are unsupported, and which are contradicted? Quote the contradictions. If nothing contradicts it, say so plainly.”
• Mechanism. “Assume the diagnosis is right and the bet still fails. Give me the four most plausible reasons, ordered by likelihood, not by how interesting they are.”
• Cost of being right. “Assume it works. What do we lose by having spent two quarters here? Who is worse off? What did we not build?”

The third one is the one people never ask, and it is where the real objection usually is.

Step 3 — Check the outside world, with receipts

If you have the browser integration enabled, this is where it earns its place:

Check how three competitors handle the first ten minutes after signup. For each: what you saw, the URL, and a screenshot. Do not describe anything you did not load. If a page needed a login, say so and stop.

Then keep the artifacts. The screenshots and the recording are what let you say “checked on the 4th” in a meeting six weeks later, and they are the only defence against quietly repeating a competitor claim that was true a year ago.

Step 4 — The question you have been avoiding

Given everything in this conversation, what is the one question about this strategy that I have not asked and would least like to answer? Ask it. Do not soften it and do not answer it.

This is the most useful single prompt in the course. It works because it inverts the default: instead of producing what you asked for, it has to find what you avoided.

Step 5 — Write the memo, own the decision

Draft a one-page memo following @context/rules.md: the position, the strongest objection to it, my answer to that objection, and what would have to be true in six weeks for us to abandon it. Use my words for the answer. Do not invent a rationale I did not give.

The last section is the one that matters. A strategy with no stated falsifier is a belief, and beliefs do not survive a quarterly review.

Where this stops working

An agent cannot tell you what your organisation will tolerate, which executive has already promised something to a customer, or whether your engineering lead has the appetite for a two-quarter bet. It can tell you your reasoning is thin. It cannot tell you whether you can get away with it. That gap is the job.