A hard problem sits open for an hour. You reread notes, redraw boxes, and feel busy. Nothing falsifies the current guess. Spinning feels like progress until the clock proves otherwise and the page looks the same as when you started.
Split the hard problem into one next test. Not a full plan. One check that could prove you wrong cheaply. That single test turns fog into a move you can finish inside a short block and record without drama.
Name the smallest claim you believe
Write one sentence: I think X is true because Y. If you cannot write it, you are still exploring, not testing. Exploration needs a different clock and a wider notebook. Testing needs a claim you are willing to risk in public on the page.
When the claim keeps morphing, you are in a decision loop. Put a timebox around naming the claim using the exit habits in timebox looping decisions so choices stop spinning. Freeze the wording when the timer ends.
Design one cheap falsifier
Ask what observation would kill the claim in under fifteen minutes. A quick measurement, a call, a sketch against real data, a prototype click path. Cheap beats thorough when you are stuck. Thorough arrives after cheap tests clear the fog.
Write the falsifier before you gather more tabs. Gathering without a kill test is how problems stay hard for theater reasons. Tab count is not a proxy for rigor.
Run the test before you polish the theory
Execute the falsifier in one sitting. Record pass, fail, or unclear. Unclear means the test was poorly designed, not that you need more philosophy. Redesign once, then run again. Do not open a third theory draft in between.
If unfinished notes threaten the next switch, park unfinished sentences before switches so the test result survives the break and still makes sense tomorrow morning.
Keep working memory free for the check
Close unrelated drafts. Shed load so the test has room. Habits from shed working-memory load before you start work apply here as much as before deep writing. A crowded pad makes every result feel ambiguous.
After meetings, clear residue before you claim you are testing. Residue makes every result feel muddy, which is why people clear attention residue after meetings on purpose before returning to hard problems.
Chain tests instead of building grand plans
When a test fails, write the next claim from the failure, not from pride. When it passes, widen the next falsifier slightly. Chains beat master plans when uncertainty is high and stakeholders want theatrics instead of evidence.
Review weekly which problems sat open without a single test. Those are process failures, not intelligence failures. Require one falsifier before another hour of rereading. The rule sounds strict until you count the hours it saves.
FAQ
What is a next test?
One cheap check that could falsify your current claim in a short sitting with a written outcome.
How small should the claim be?
One sentence you would defend today. If it needs a paragraph, split it into separate claims.
What if the test is unclear?
Treat unclear as a bad test design. Rewrite the check once, then run it again without new theory.
Should I plan the whole solution first?
Not while stuck. Chain small falsifiers until the shape of the solution appears from evidence.
How long should a test block last?
Fifteen to thirty minutes is enough for most cheap falsifiers. Longer blocks invite polishing.
What do I write after the test?
Pass, fail, or unclear, plus the next claim. Then stop theorizing for that block.
