Discovery

A queue of specific tactics pulled out of things people published. Each row is one claim, the number they gave if they gave one, the source and date, and what testing it would actually cost you. You do one of three things with each: try it, reject it, or leave it for later.

Every other screen in this product improves what you are already doing. The allocator moves budget toward the message that is working, the scoreboard tells you whether a lane is worth another week, the experiment engine kills an arm that has had its forty sends. All of that is search inside a space you defined when you set the venture up.

None of it can find a tactic nobody here thought of. A system that only optimises converges on a local maximum and then stays there, looking healthy the whole time, because every number it watches is improving. The only way out is to go and read what other people claim worked and try the ones worth trying.

That is why this screen shows claims rather than conclusions. We have not run any of these. We cannot tell you whether the person who wrote one was right, whether they measured it properly, or whether it only worked because of something about their market that is not true of yours. Judging that is the part you are good at and the system is not, so the system does the reading and the deduplication and hands you the shortlist.

The design constraint that matters most is that it stays cheap. Three minutes a week, keyboard-driven. A queue that takes longer gets opened once, and a discovery loop whose queue nobody opens is worse than not building one, because the miner keeps filling it and you get a growing pile of unread homework instead of a habit.

1. 01Open it once a week. The keys are j and k to move, a to try it, x to reject, l for later.
2. 02Read the claim and the credibility badge together. Somebody who ran it and gave a number is worth more than somebody repeating advice, and the badge says which.
3. 03Check the cost line before accepting. It tells you how many sends and roughly how many weeks before the result means anything, which is the real price of the decision.
4. 04When you reject, give a reason if you have one. The reason is read back by the miner to stop showing you that class of claim.
5. 05Open the source when the claim sounds too good. The original text is on the row, and the link is on it too.

• Learning
• The scoreboard
• Approvals