WebeleonThe first three minutes of a practice session are often a negotiation with yourself. The fun song against the wedding song. The new tune against the rusty one.
That negotiation is where sessions go to die. Deciding costs the same energy as playing.
The fix is to take the decision out of the moment. Decide once, in the cold light of a rule, instead of nightly, in the warm dark of the practice room. My rule has three lines. Songs with a gig on the calendar come first, ordered by the nearest date. Then the weakest scores. Ties break by the longest unplayed. And the first song is never decided at all: a dice opens one at random, so the session starts before the argument even begins. Nothing hidden, nothing weighted by a model I cannot explain. If the tool cannot read its own rule out loud, the feature is wrong.
The decision is the tax
Nobody counts choosing as practice. The meter runs anyway.
By the time you have argued yourself into the ballad, you have spent the easy minutes of the evening on an argument you will have again tomorrow, from zero.
The daily practice loop exists for this reason too. A session that starts with structure starts playing sooner and stops negotiating earlier. But the loop governs how a session runs. This rule governs what it runs on. And it has to be made before the instrument is in your hands — otherwise it is just another opinion.
A rule you can read out loud
Mine, in full, in the order it applies.
First: a song serving an upcoming gig surfaces ahead of everything. Among gig songs, the nearest date wins. A wedding in twelve days outranks a festival in two months.
Second: everything else, ordered by the weighted score from rating the take twice. The honest trouble spots float up.
Third: ties break by age. Longest unplayed wins. A never-played song counts as oldest — so a new tune for next month's set cannot hide in the library forever.
Three lines. That is the whole point. A rule this short lives in your head, and you can audit any decision it makes. Why is the ballad near the top of tonight's list? Gig in twelve days. Why is the new tune above the old standard? Never played, therefore oldest. When the list arrives already ordered and you can reconstruct its reasoning from memory, you stop arguing with it. You already argued with yourself when you chose the rule.
One song, at random
The rule orders the list. It does not open the session.
Open Edgars and the first thing you see is a dice. Tap it and it draws one active song at random, recording already armed — two taps from a closed app to a downbeat, and neither of them is a decision.
Random at the door is the point. Any ranked opener is still a choice, and a choice is still the negotiation. Hand me the rule's top pick and I will ask whether tonight is really a gig night. Hand me five ranked and I will ask about the sixth. The dice does not argue. It opens one song and the practice starts.
Nothing is lost after that first tap. Below the dice sits the ranked list — every active song, ordered by the same three lines, each row carrying its plain reason. That rule is what carries the session from the second song on. The opener is random. The order is earned.
What I refused to build
A confession from the build side. The boundaries are the product here.
No streak counter. A streak is a debt you owe the app. The day you break it, the app has made itself the villain of your own practice. No guilt copy, no flames, no "don't break the chain." An empty week renders empty. The app tells the truth without scolding.
And no "the algorithm picked this." There is no algorithm, in the engagement-industry sense of an opaque optimization target. There is a published rule, readable in the app in one tap, exactly as long as the one above.
The line I will not cross is between surfacing and nagging. A tool that answers "what should I play next?" in three seconds has earned its place on the stand. A tool that makes you feel watched has forfeited it — no matter how good its picks are.