There is another way to act well without solving the whole enormous problem: don't. Just work out the best move from where you actually are, right now.
Chapters 6 and 7 computed a policy offline — a complete rule worked out in advance, covering every state. Online planning takes the opposite stance. It computes nothing ahead of time; instead, whenever a decision is due, it looks ahead from the current state, evaluates the plausible futures, picks the best immediate action, takes it, then repeats the whole exercise from the new state. It spends effort at decision time rather than storing a policy — and for project work, it mirrors how good practitioners actually think.
1Look ahead, act, repeat
The simplest form is receding-horizon planning: look ahead a fixed number of steps, choose the best first move as if the world ended at that horizon, then slide the horizon forward and redo it after each step. You never commit to the distant future — you only ever commit to the next action, and you recompute constantly as reality unfolds. This is rolling-wave planning in everything but name.
2Searching futures cleverly
The art is in exploring possible futures without drowning in them. Forward search expands the tree of actions and outcomes to a set depth. Branch and bound prunes branches that provably can't beat the best found so far. Sparse sampling keeps the tree manageable by sampling a limited number of outcomes per action. And the standout, Monte Carlo tree search, grows the tree selectively toward the most promising lines: it repeatedly selects a path (balancing promising moves against under-explored ones), simulates a rollout to the end, and backs the result up the tree. It concentrates effort where it matters and has powered some of the strongest game-playing systems ever built.
Online planning refuses to solve the whole problem and instead solves the only part that's due: the next move. It trades a stored policy for fresh computation at each step — and re-plans continuously as the project reveals itself.
You do not need a pre-baked answer for every situation a project could reach — you need the best next decision from the situation you're in, and the discipline to recompute it as things change. Simulating a few moves ahead from the current state, then committing only to the immediate action, is rolling-wave planning made rigorous. Look ahead far enough to see the consequences that matter, act, and plan again from where you land.