Detune Precision by Time Horizon
The shorter the horizon, the more detail; keep long-range plans deliberately hazy to avoid false precision
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 3
- Confidence
- 86%
Ambrosino's planning rule is simple: the shorter-term something is, the more detail it needs, and anything far out should stay hazy on purpose. Adding precision to a 9-month plan in a fast-moving AI environment produces false precision that just wastes time, because whatever you planned in November won't be what actually happens.
Origin
Andrew Ambrosino describes this as deliberately un-clever, basic planning practice for the applied product side at OpenAI (noting research plans differently).
Core principles
- 01Detail should scale with proximity: near-term detailed, far-term hazy
- 02Precision added to a long-range plan is false precision that wastes time
- 03You can still state direction for 9 months out — just don't over-specify it
- 04Applied product plans decay fast; what was true for December wasn't true by later
How to run it
- 1
Set detail proportional to nearness
Give near-term work concrete detail and specificity; leave anything months out at the level of rough direction only.
Watch out Any precision you add to a 9-month plan right now is false precision.
- 2
Anchor long-range plans on model capability timelines
For distant work, reason about what you think models will be able to do on what timeline, rather than specifying exact features.
Pro tip State direction ('we can say stuff') without committing to specifics that won't survive contact with the next model.
- 3
Re-plan as capability lands rather than defending the old plan
Accept that the applied roadmap will churn and update it against reality instead of forcing the original plan to hold.
Watch out Anything you could have planned in November may have been true for December but isn't what happened.
In the wild
On the applied product side, Ambrosino notes that anything planned in November may have held for December but diverged after — making detailed long-range planning genuinely hard and low-value.
→ Teams keep the far horizon hazy and re-plan frequently; people are 'frustrated with me all the time' because things constantly ship and change.
Common mistakes
Adding precision to a 9-month plan
Detailed long-range plans in a fast-moving AI environment are false precision — they feel rigorous but waste time because the underlying model capability and market will have shifted.
Defending the original plan when reality diverges
Treating the November plan as a commitment rather than a hazy direction blinds the team to what actually happened by December and beyond.
Is it for you?
Best for
Product and eng leaders planning roadmaps in fast-moving, capability-driven environments
Not ideal for
Research or infrastructure programs with genuinely long, stable lead times where detailed long-range planning pays off
From the transcript
“the shorter term something is, the more detail it needs”
“any amount of precision that you add to a 9-month plan right now is false precision”
“anything that you could have planned in November may have been true for December but like isn't what happened”
From the episode
OpenAI Codex lead on the new shape of product work
Andrew Ambrosino