Law II
Difficulty Is Load-Bearing
The hard part is what produces the value, when it gives you a second, independent way to check whether the answer is right. Effort that only repeats the same route can go. This covers knowing whether you are right; the effort that builds capacity, like training or practice, sits outside it and is untested.
- StandingRevised under pressure, relied on
- Last tested18 Aug 2026
- RevisedOnce
- AppliesAt any time scale
Narrowed after a test in August 2026. The narrowed wording has not been tested yet.
Ask this of your own work
Does this hard part give you a second, independent way to check the answer, or only more effort along the same route?
An example
In 2015 the US Naval Academy put navigating by the stars back into training. Its reason: “computers are great. The problem is, there’s no backup.” The harder method is a second way to know where the ship is.
Start with
- Confidently Wrong The effort you're handing to AI was doing two hidden jobs. Skip them and you get faster, weaker, and blind to your own mistakes.
- Remembers Everything, Learns Nothing You gave your agent a memory and it still repeats the same mistake. What makes it improve is a loop that tests each failure and turns the ones that recur into procedures.
How it could be wrong
What would break it
What would count as proof it is wrong. When a test meets the old version, the law is rewritten and the change is recorded beside it.
A hard part of knowing whether you are right that gave no independent check, yet removing it wrecked the result. Or one that did give an independent check, and removing it cost nothing.
Rewritten on 18 Aug 2026, when a test met the old version.
What changed, and why
· after a test
The old wording could not fail: whenever removing a hard part helped, you could call it the wrong hard part. A test of five cases split on one line. Checking a computer model by hand is an independent check; without one, the error in the Sleipner A platform’s model (shear under-predicted by about 45%, at a cost of about $700M) went uncaught. Surgical speed, made unnecessary by anaesthesia, and hand-stowed cargo, replaced by containers, were effort along the same route, and removing them helped.
The exact wording, as the framework records it
The hard parts ARE the mechanism producing value — but only difficulty that forces an INDEPENDENT ROUTE to the answer; effort along the existing route is removable.
The full statement
The hard, constrained, effortful, and unmeasurable parts of a system are not bugs to optimise away — they are the structural mechanisms producing value. Remove them and the output collapses. AMENDED 2026-08-18: this holds for difficulty that forces an INDEPENDENT ROUTE to the answer — a second, differently-derived way of knowing whether the thing is right. Difficulty that only adds effort ALONG the existing route is not load-bearing and can be removed freely. Scope limit: the criterion governs epistemic difficulty (knowing), not capacity difficulty (doing) — progressive overload, hormesis and deliberate practice sit outside it and remain untested.