Week 9: Breaking a large subject into small tasks, and building it as software

The ladder gives you the inside of thirty minutes and nothing beyond it. Long-term planning is the other half of the job, and the way a Souls game does it is to let a large subject come apart into small tasks one attempt at a time, rather than all at once on day one

3 May 2027

You have built both halves now, and the temptation is to think of them as two versions of the same thing. They are not, and using one where you needed the other is the most common way all of this falls apart.

(The second half of today is the second assessment.)

Two mechanisms, two jobs#

You have now built both. It is worth being exact about what each one is for, because they are not two flavours of the same thing and using one where you needed the other is the most common way this all falls apart.

The ladder buys you the inside of thirty minutes. Absolute focus and fast feedback while you are sitting there. It is extremely good at this and nothing else does it as well.

What does it offer over a semester? Honestly: encouragement. If you climb quickly and end up at a high rank, you acquire a plain, unadorned confidence — the kind that comes from a number you trust going up, which is not nothing and is not a plan. The ladder cannot break a large task into small ones. It never could. Rating systems rank performances; they have no opinion about what you should do next, and no amount of tuning will give them one.

Getting started, and getting small, live in the Souls half — and the thing to be exact about is that neither one is planning.

Nobody sits down in front of a new Souls boss and decomposes it. You walk through the fog and you die in ninety seconds. What ninety seconds gets you is one observation — it always sweeps after the second swing — because that is all a ninety-second attempt has room to carry. So the next attempt has one new goal. Then the next has one more.

Twenty attempts later the boss is dead, and you never made a plan. The fight came apart into small pieces on its own, a lesson at a time, and it did that because the loop was short enough to only hold one lesson. Decomposition was an output of the attempt cycle, not an input to it.

That matters for the obvious reason: when a subject is genuinely too hard to start, you cannot decompose it first. You do not know enough about it yet to know what its parts are. Any plan you write on day one is a guess, and a confident wrong plan is worse than none. What you can do is make an attempt short enough to survive, take the one thing it gives you, and go again.

So the skill here is how to get a grip on something too hard to start, and how a thing that big becomes a thing that small. The ladder was structurally incapable of either.

What to build#

A boss tree — but build it as a record of what the attempts have told you so far, not as a plan you write on day one and then execute. It should be mostly empty when you start and should get more detailed as you fail. Three properties follow from that.

It rolls up. A boss’s remaining work includes its children’s. If you mark a parent done while its children are open, the tree is lying to you, and it will lie in the flattering direction.

It survives subdivision. Halfway through, a boss will turn out to be three bosses. That is normal and is not failure; it is the fight becoming legible. If splitting a boss corrupts your history or resets your progress, you will avoid splitting, and you will keep a wrong plan because the tool made the right one expensive.

Nothing has a length you fixed in advance. This is the one I would fight for. Estimate the work, and let the duration fall out of the estimate — a campaign’s length is a forecast that moves when the load moves, never a commitment you made at the start. Fix the length up front and you have not set yourself a campaign, you have sentenced yourself to a term, and the difference is the whole reason one of them is bearable.

The unit that finally holds a result#

There is a gap between the two mechanisms and it is worth naming, because it is where most people’s motivation actually dies.

A sitting is thirty minutes. A boss might be a hundred hours. A day is the unit we all reach for, and a day is too short to contain a result worth feeling anything about — which is why so many evenings end with nothing to point at and the vague sense of having wasted one.

The unit that works is several days. Long enough that something finishes inside it, short enough that you can see the far end from where you are standing. Pick a handful of bosses, declare them the current push, and do not put a date on it. When they are done, close it and start another. A few days is enough time to become noticeably different at something, and that is the feeling you have been missing.

Assessment 2 demos#

Second half. Paper or coded, and you may have chosen differently from Assessment 1 — several of you should have.