manoo

yooo, i was literally trying to find something like this ystd

houses only do this when they are severely distressed

one quadrillion dollars

les français on une fascination weird ou un genre de "noble savage view" des québecois sur l'internet LOL

Funny cat picture in a bath captioned "I have to say weird stuff or I will explode"

hospital food is SO ASS BRO. i have to eat the grey SLOP for the 50th time

chilling in a hospital bed isnt the worst thing in the world

mac and cheese with any hot sauce is fire

oooooooooo

it ez what it is

guess what, im in the hospital for like 3 days fuck

being productive is a scam

I got Invisalign but this shit is making me suicidal

the problem with switching pds is that:

1. I'm lazy
2. I don't trust any lone wolf to run a pds
3. I especially do not trust myself

don't belieb this ....

yeah some people need to turn off the pedantic *pushes back glasses* "um actually" comments

my profession is changing day-to-day. it's going in an odd direction that I don't know how to feel about.

but I am dedicated to NOT becoming one of the last blacksmiths who outright denies the effectiveness of CNC machines.

tldr: "it is totally useless because it might be wrong or insecure" is a bad argument if you were trained as an engineer.

our entire curriculum is literally geared towards calculating tradeoffs because humans are not perfect and computers aren't either

this is also why I'm totally against just vibe coding without looking.

but even then, there *is* an argument to be made that review-less coding is fine for *some* cases (if you don't gaf about security)

people also forget that you can verify the output!

a human must see and read the code if you want to be absolutely sure, but that takes very little time compared to what it was before.

it's all a game of balancing how you spend your time.

in a lot of cases: yes! it's more useful to spend some money on the black box that generates something that might work than spending time trying to figure it out!

time constraints, manpower limitations, etc etc. are all things that affect code quality and are very real things to think about

the engineering problem with AI shifts to: how can I leverage these black box dice rolls that have an 80% chance of being correct to speed up the creation of what I want?

ive never worked in a professional environment that:
- formally verifies code for correctness
- unit tests 100% of the code
- has never had a security incident

and it doesn't matter for like 80% of what you are doing.

i think a lot of the "it makes buggy code" critiques come from a naive view that humans have a very high % chance of never committing buggy or incorrect code, high enough that it outdoes AI.

this is absolutely false (at least it is now, with more recent LLMs)

i think one important detail that AI deniers need to internalize is that 80% of software doesn't need to be 100% correct. we should strive for it sure, but most businesses run on a model of "it's good enough" and a huge part of engineering is figuring out what % correctness you need

is this a pattern for an atproto record:

write -> one application only (let's say bsky)
read -> everyone else

like is there a way to create read-only data within an appview

ive realized that "reading error messages thoroughly" will put you at the top 1% of the software engineering profression

We mourn our craft I didn’t ask for this and neither did you. I didn’t ask for a robot to consume every blog post and piece of code I ever wrote and parrot it back so that some hack could make money off o…

feeling this one today

no more takeout until further notice

im health maxxing now