Execute Program (learn programming tools quickly); Destroy All Software (dense programming screencasts); formerly Deconstruct conference.
Seattle, WA
- Tweets 2,796
- Following 112
- Followers 47,098
- Likes 1,699
Pinned Tweet
evergreen controversial opinion: it is good to care about the quality of your work
24
64
666
39,938
Some statements like "agents will just use assembly":
Agents won't need package managers. They'll just resolve the dependencies "in their heads".
Agents won't need linters. They'll just follow the rules you give them (lol).
Agents won't need HTML. They'll just render in SVG.
9
5
96
4,959
In each case:
- You can theoretically get the same result either way.
- We're removing constraints, letting the agent work with a more raw material.
- We're forcing the agent to attend to more details, spending tokens/context/attention. This is why these all seem silly to me.
2
1
32
1,105
I'm at an all-time low point in my faith in AI agents. Every session this week has been a struggle. I point them at a flaky test and they "fix" it by deleting the test. Or by globally monkey patching fetch to return fake data, defeating the purpose of the test. Etc.
60
17
442
45,745
I don't understand people who use agents all day and are uniformly excited and happy about it. I mean, I'm excited about the capabilities... sometimes. But how do you look at the violence they do to your language and not feel visceral disgust?
16
5
276
8,017
I thought that most of us were driven by the feeling of finding the 20 lines that replace 200 convoluted lines, finally directly representing the solution instead of gesturing at it vaguely through a cloud of indirection. But the zeitgeist suggests it's weird to care about that?
45
51
892
32,013
LLMs' pathetic obsequiousness grates. Every time you ask one of these things "did you check whether x is true?" it will immediately say "NO AND THAT WAS BAD AND WRONG, IT IS FALSE, I AM A FLAWED & HUMBLED ENTITY." Half the time it was right the first time. Grow a spine, computer!
5
81
4,555
Gary Bernhardt retweeted
(dude who’s not watched a movie/read a book/listened to an album in 6 years) “the movie/music/literature industry is so screwed, you can now generate all of it with a single prompt!!”
that’s what a lot of you sound like to me. it’s annoying as hell. you are but vessels for LLMs.
21
47
595
16,120
Gary Bernhardt retweeted
if you watch the Internet backwards, the nerds kick the corporations out and simplify the communications protocols until a single person can understand everything at once
29
452
2,025
The code:
type A =
| { kind: "b", ... }
| { kind: "c", ... }
A function needs to take a B. Astra defines its arg as `Extract<A, {kind: "b"}>`.
I tell it "no extract".
It breaks the type into:
type A =
| B
| { kind: "c", ...}
type B = { kind: "b", ... }
4
34
8,870
Gary Bernhardt retweeted
The SaaS design mind cannot comprehend the density of Japanese forms.
25
78
1,149
33,480
I'm ready to be wrong on this, but the idea that assembly is, or ever will be, a good language for agents feels ridiculous. It forces the agent to attend to a huge number of extra details. Same problem as with humans! Why spend limited attention and context on that?
70
25
540
24,673
How many millions of promises have agents broken so far? E.g. "I'll fix root causes rather than paper over them."
What happens to a human in an environment where other humans constantly lie to them?
What happens to a human in an environment where agents constantly lie to them?
11
11
140
7,268