126MB color picker? in this economy?!
126MB color picker? in this economy?!
both user and coding agent larp as 2022 coders when it's actually more like project manager toddler and eldritch entity with dementia
bruh
In retrospect: I used this whole faulty ThoriumLXC setup because Claude recommended me it ๐
Today I tried AzerothCore instead and it was so easy to setup, the only complex moment was to add playerbots module which required making override to docker compose but other than that very easy
I was instead wasting time totally convinced I missed some step from the guide but it was rather obvious now that I know
Docker logs got flooded with something trying to put stuff into tables that don't exist. This was a hint to check *.sh and *.sql files in a database container to find tons of unused ones in the startup script.
That all happened through tiresome copy-pasting of text around web interfaces and terminal, I bet if I instead gave CLI agent direct access to docker images it could've found that issue way quicker just running few careful greps
But it's not like AI one shotted this issue, no no no. I talked to Gemini, GPT, Claude even Grok bruh, pasting outputs of commands, logs etc. For quite some time until we finally with Claude figured everything once and for all (though I think I arrived at answer earlier than realized)
I spent several hours debugging broken World of Warcraft local server docker image and AI helped a lot. Turns out database image has broken script where they completely forgot to run bunch of queries to create important tables for playerbots to work.
problem -> solution ๐
I wish this was true, my trial runs out
You as delusional as you claim them to be if you're so quick to disapprove consciousness so easy
test driven development is good until your agent gets "all tests pass" psychosis
this was an early GPT issue but now when they train tokenizers there's special regex that makes so that bunch of spaces become single string for BPE to "learn" tokens from
finished Antimemetic Division and ooh it was good
then went asking Claude to explain few confusing parts I missed, at first it completely hallucinated nonsense, asked it to keep doing web searches but best responses came from dropping entire pdf and asking questions piecing plot back together
for optimists:
imagine a future genie that can make most wishes come true and start writing down every good wish now to be prepared, maybe some wishes will become completable earlier than others, order them by complexity and importance
for pessimist:
preparing for every fail scenario where that doesn't happen
solar panels, homelab, downloading and storing entire internet knowledge while it's still up and open, some other post-apocalypse preparation stuff
Using image model while reading a book to confirm whether I properly understood confusing scene described
do zoxide next
should I?
They caught the Rust disease of never releasing anything into stable
Gmail. Most emails now have Unsubscribe button and it silently sends unsubscribe email from you.
They now even have separate tab listing all services that sent you annoying stuff recently with unsubscribe button near each of them
maybe they're migrating for new model or something because that thing is dead, even flash barely work
They're good for integrating p5.js or canvas into Astro page without meddling with passing IDs around when you want to have multiple of them on same page
the council has spoken, turn on the A/C
Nobody, source article more about password that made their way into the code.
But even checking GitHub with their provided examples most of these are comments and placeholders. Though few actually use passwords with those characters ๐ญ, but then why passwords in the source code in the first place
what in the cognitohazard
github.com/openai/grok
waste of time and money