Gary Bernhardt retweeted
Nathan: Do you see why this is funny?
Elizabeth: Because all these certificates have to be revoked?
Nathan: Yes. But do you understand why 4 is a terrible secret key?
Elizabeth: Well... it's equally likely as any other value drawn from the uniform distribution.
When an LLM says:
"Don't turn assumptions implicit in an everyday scenario into explicit input restrictions"
I have some intuition for how it got there instead of just writing:
"Don't state obvious assumptions about inputs"
and that scares me a bit.
the AI era has brought a new world of pain regarding "do I need to hit shift+enter or alt+enter to insert a newline", where choosing the wrong one often submits the text instead of inserting the newline, and I am genuinely mad about this in real life
I've been back on codex this week for reasons, and it makes me miss pi. Pi is simple, has very little UI jank, and it doesn't tell me to "visit the community forum" and other nonsense. It waits for me, then does exactly what I said, which is how terminal tools should work.
Gary Bernhardt retweeted
I always listen to the confident, loud declarations about the future of programming (don't want to get left behind)! That's why I always start with a UML diagram, then let tooling generate aspect-oriented WS-* services, with the data stored in a network database.
blockchain
This is my experience too. It's a huge, obvious effect. Now put this alongside the utterly confident claims that "AI writes better code than you ever did." xitter.cf/Sirupsen/status/211290…
I fully appreciate the feeling, and I feel it too, but which "value"?
"What will someone pay me a wage to do?"
"What do I enjoy doing?"
"What will improve other people's lives?"
"What will reduce the probability of Dijkstra haunting my home?"
xitter.cf/housecor/status/211255…
I, for one, consider 4 a perfectly fine integer between 1 and 39402006196394479212279040100143613805079739270465446667946905279627659399113263569398956308152294913554433653942642
The Prompt Engineer's Prayer:
AI didn't do that.
And if it did, it was good.
And if it wasn't, you used the wrong model.
And if you didn't, the next model will fix it.
And if it doesn't, that's a skill issue.
curious that AI behavior we like is evidence of superintelligence, but AI behavior we don't like is a skill issue on the human operator's part
Gary Bernhardt retweeted
a compiler is a function that turns a sequence into a tree, turns that tree into a graph, and then turns that graph back into a sequence
Gary Bernhardt retweeted
Right now someone is learning a tidy but wrong theory about your personal area of expertise.
Gary Bernhardt retweeted
1999: there are millions of websites all hyperlinked together
2019: there are four websites, each filled with screenshots of the other three.










