Think about how you learned
the genuinely hard things.
Not from the documentation.
From a thread where somebody
under no obligation to anyone
hit the same wall in 2014,
worked at it for four days,
and then came back
and wrote down what it had turned out to be.
They got nothing for it.
Some points on a website.
A little standing among people
they would never meet.
That was enough,
for about twenty years,
to produce the entire written record
this profession runs on.
Now watch what is happening to it.
The person hitting that wall today
asks a model,
has an answer in nine seconds,
and moves on.
Nothing is written.
The working out happened
inside a private window
and ended with the tab.
Multiply that by everybody.
The forums go quiet,
which makes them less worth visiting,
which makes them quieter,
and the thing producing the answers
was never the website.
It was the habit of posting.
Two consequences,
and the second is the sharp one.
The first is that the record
thins out at roughly now.
Everything written about a library
released next spring
will be its own documentation
and whatever the vendor says,
and the awkward truths
that used to surface in public
after six months of real use
will be found individually,
by each of us,
alone.
The second is that the models
are fed by that record.
They are excellent on the years
when people wrote things down.
They will be exactly as good
on the years when we stopped.
None of this is an argument
for refusing the tool.
It is an argument for the ten minutes
after you solve something horrible.
Put it somewhere public.
Briefly, badly,
with the error message in it
so a search can find it.
You are not being generous.
You are putting back
what you have been taking out,
and the answer you will need next
does not exist yet
unless somebody does this.
– Serguey Asael Shinder
