sequence-plan-around-risky-assumptions

Sequence the plan around risky assumptions

Order by information gained and time to recover, not by convenience. The cost of a wrong assumption is paid at the moment you discover it, so discover the expensive ones early, while the schedule can still absorb the news.

Procedure

1. Score each task in notes/plan.md on two axes: uncertainty (how likely the approach is wrong) and lead time (how long it takes to get any signal).

2. Sort so that high-uncertainty and long-lead items come first; low-risk, well-understood work is filler for the back.

   python3 -c "
   t=[('new-indexer',9,9),('copy-tweak',1,1),('vendor-api',7,8),('config',2,1)]
   for n,u,l in sorted(t,key=lambda x:-(x[1]+x[2])): print(n,u+l)"
   new-indexer 18
   vendor-api 15
   config 3
   copy-tweak 2
   

3. Expose the riskiest assumption with the smallest possible probe first: a spike, a call, a single fetch. Do not build the whole feature to learn it will not work.

4. Interleave the safe work as the "next thing" when a risky item is blocked waiting on a signal, but never let comfortable work crowd the risk off the front.

5. If the riskiest item fails, the plan must have an alternative ready — that is why it is first; the schedule has not yet been committed to it.

6. Re-check the ordering when an assumption resolves; a former risk becomes known work and the next unknown moves up.

Pitfalls

• Doing the known work first to "get momentum," so the risky item surfaces late when there is no time to react.
• Confusing uncertainty with difficulty: a hard but well-understood task is not the thing to front-load.
• Front-loading risk without a fallback, so the early failure has nowhere to go.
• Treating a long lead time as low priority because effort is low — lead time dominates recoverability.
• Sequencing by who is available rather than by information value.

Verification

python3 -c "t=[('a',9,9),('b',1,1)]; print(sorted(t,key=lambda x:-(x[1]+x[2]))[0][0]==\"a\")"
# True — the higher uncertainty+lead item sorts first; a plan ordered otherwise is inverted

Report the ordering, each item's uncertainty and lead time, and the fallback for the riskiest one.