you split the work across five agents to go faster, and it did go faster
every one of them loads your claude.md before it reads the task
mine is sixty seven lines, about seven hundred and forty tokens,
a third of a percent of a two hundred thousand window
five agents make that three thousand seven hundred, twenty make it close to fifteen thousand
and twenty is the documented ceiling on concurrent agents,
so that is the top of the range, not a scare number
yours is probably three or four times mine, because mine is short
the agent editing code needs those conventions
the agent grepping a log needs none of them and pays the same
one field turns it off for that worker, omitClaudeMd, a single line in the file
you did not parallelise the work
you parallelised the preamble
THE MONKEY SHOVED A BANANA ICE CUBE INTO A VOLCANO
ten seconds. one animal. one block of ice. an open door over lava.
the clip contains almost nothing:
→ no cut until the dump
→ no speech, no look at camera
→ ice goes in, fruit comes out
→ nothing labelled, no reason given
the object itself is the implied antecedent. viewers have to finish the sentence the footage started: why the cube, why bananas, why the volcano, why the monkey decides.
nobody can be wrong. every comment supplies the missing noun.
a like is cheap. inventing the motive is work. platforms weight the work.
the footage stops the scroll. the unanswered why makes them type.
nothing rational happened. people will still explain it.