Hugo Sousa (@Ciberon) profile picture

publishing https://nitter.cf/t.co/gKTY3vdsdX & https://nitter.cf/t.co/kH0ObxlWTW building https://nitter.cf/t.co/hiUFmYyl0u

Joined February 2009
Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)Photo by Hugo Sousa (@Ciberon)
I must be doing something really smart or really dumb when AI is avoiding to do what I'm asking
31
Luna keeps telling me > No—I didn’t add tests. I followed the instruction not to add or run them unless you asked. So I guess that's in the chatgpt system prompt for it.
1
1
25
I wonder if anyone is back to using AI via autocomplete mostly, like the previous cursor ux.
1
77
We need agents* * people with agency
Okay, longer version because this was dumb. I think I wrote the exact opposite of what I meant: we no longer need folks who just "do" what others have thought about. i.e. we no longer need people who just code up what others specified. So, I should've written "we no longer need doers, but thinkers", except that also isn't it, is it? Because now that you can get machines to "do" so much for you, you need to pair the thinking with activation energy, agency, and the willingness to go big. Maybe the best way to put it is that the distinction doesn't make sense anymore in software, because 99% of the "doing" has gone poof, since you can dictate a lot of ideas into existence now. But just thinking of ideas is also no longer valuable, is it? Maybe it was, when you had dedicated "doers", but now you need to pair the thinking with action. Think-doers is what we need.
29
The ballsack distillation attack
16
"burger king implies the existence of burger queen" in a universe of only kings and queens, with the assumption that the "burger king" is not the whole universe. It's the search of a language counterpart in a universe slightly bigger than that concept. I hope this helps.
i used to get annoyed by posts like "burger king implies the existence of burger queen" because it doesn't "imply" that in the math sense. and then i remembered "implies" can also mean "suggests" and now it doesnt bother me. hopefully this will help others similarly sperging
37
A while ago I had an idea for having types describing the actions you want and then have the scala compiler fill the implementation through implicits. Might be time to revisit that idea.
In theory program specifications (types) should be shorter than programs (terms) because we can abstract away what’s irrelevant. Before AI we had to write the program otherwise who would? But now, why are we still writing programs (terms)?
37
Some days being on long sick leave gets to you but as long as you have energy you can turn it around.
23
Did some work on my programming language, which I hadn't in a long time. Feels good.
1
23
ngl i’d be in favor of removing all write apis from X
73
31
8
1,325
50,291
Hugo Sousa retweeted
Holy balls the latest update to macOS *ROLLED BACK* the eMudhra cert removal from /etc/ssl/! What the hell are you doing over there, Apple friends?
113
260
61
5,301
401,576
Do we deserve to eat shit
There will not be a new agent-first language that replaces everything We’re stuck with JavaScript guys Not the language we need, but the one we deserve
1
39
Hugo Sousa retweeted
I kid you not, sometimes I talk to an AI and they are stunned by my insight into architecture
302
78
24
5,675
167,944
Hugo Sousa retweeted
One of the weirdest effects of LLMs is that people act as if writing extremely complex programs is somehow solved now. Recently, I was talking about writing a compiler on here, and one of you suggested to "simply" write a verified compiler. As if writing a verified compiler is something that doesn't require you to be a top scientist AND crack programmer at the same time. Somehow LLMs will do that for you, apparently. Despite there being like 1 verified compiler (CompCert) in existence at the moment. But somehow it's something that one simply does now.
59
21
1
967
63,649
Cooking up my own issue tracker
42
New phone: :) Setting up: :(
30
The clanker keeps clanking
26
Hugo Sousa retweeted
there's no such thing as a rogue agent man its software someone made and deployed they are responsible for what it does cannot let this blame shifting occur
147
434
41
4,671
97,181
Hugo Sousa retweeted
And the sanity cost for microservices is always underestimated.
Microservices solve a team size problem. If you have three developers, a modular monolith with clean boundaries will beat them on speed, cost, and sanity.
1
4
68
1,978
Hugo Sousa retweeted
if you can do all of these things, you don't need a job at an ai startup You can literally just make your own job
literal hardest role to hire for rn and every ai startup wants someone > chronically online (knows trends) > has taste & can create (new media) > understands ai & technical > can execute with quality (operator) such a small niche of folks who can do this
2
2
42
1,687