Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
4b11b4
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
31.
▲
by
4b11b4
2mo ago
Root agents.md should be human authored. Even if you copy and paste into it. That's your only lasting thread into having any intuition on the behavior that will result
32.
▲
by
4b11b4
2mo ago
I've been experimenting with not only refactoring but rewriting history. I'm always moving around commits, inserting, splitting things up. The amount of times I've instructed to "tell the story right the first time"
33.
▲
by
4b11b4
2mo ago
Lately I just use an isolated worktree per conversation and I have all these threads sitting which need to be merged then I just let codex orchestrate that itself. I tell each thread to name itself with a prefix and it's picked up by a
34.
▲
by
4b11b4
2mo ago
In codex threads can inspect each other, rename themselves , send each other messages. communication is default primitive
35.
▲
by
4b11b4
2mo ago
Uhh yeah I don't let agents edit files that I'm writing - that's how you get psychosis. I enforce this via snapshots before and after tool calls
36.
▲
by
4b11b4
2mo ago
There were papers about this years ago
37.
▲
by
4b11b4
3mo ago
Seriously: for the sake of all future children, the elderly, and everyone in between
38.
▲
by
4b11b4
3mo ago
What world are you living in, I don't want to go there
39.
▲
by
4b11b4
3mo ago
Generally no, but if I'm using voice I might be thinking out loud and make a connection to something else funny
40.
▲
by
4b11b4
3mo ago
misleading to people, people think claude is "smart", people think their ideas are better than they are, sycophancy, people are drawn to confide in a model over other people, etc
41.
▲
by
4b11b4
3mo ago
I love emacs
42.
▲
by
4b11b4
3mo ago
In general I'm referring to the contrast between Claude and GPT... where Claude might follow some tangent idea you mentioned and tell you how its interesting and give you some elaborate response about that little one remark you made wh
43.
▲
by
4b11b4
3mo ago
I strictly prefer when models ignore any human quirks in my responses. Claude trying to be your friend, saying LOL to your jokes is ridiculous and frankly, harmful
44.
▲
by
4b11b4
3mo ago
I stopped reading Claude's ridiculous distracting nonsense walls of text months ago. Honestly, I prefer Codex being boring and basically ignoring any human utterances I make which are unrelated the work to be done.
45.
▲
by
4b11b4
3mo ago
Really liked the framing on this
46.
▲
by
4b11b4
3mo ago
Hah Tetris that really matches my "everyone has to draw their own line constantly".
47.
▲
by
4b11b4
3mo ago
I'm sorry, what! 400k...?
48.
▲
by
4b11b4
3mo ago
What do you mean by integration. Do you have a model driving emacs via server?
49.
▲
by
4b11b4
3mo ago
looks cool but a little rough. If you had another 3d scene how would you integrate this?
50.
▲
by
4b11b4
3mo ago
And drenches them in sound
51.
▲
by
4b11b4
3mo ago
The beginning of Shinra?
52.
▲
by
4b11b4
4mo ago
I like the idea and I know you explicitly address this but wonder if still it could search for human made works for you to learn from first If it does find some, maybe it could supplement them instead of just from scratch
53.
▲
by
4b11b4
4mo ago
I'm increasingly over it. Don't care how good the model/harness gets, the second I can't hold every bit of the mental model in my head is the second it's all over. And that's a very fine line and very easy to c
54.
▲
by
4b11b4
4mo ago
Fuck private equity practicioners at every end of the spectrum, I don't care what role you play and how big or small the firm is. I heard some crazy anecdote a few months ago about them buying up firefighting software and then just cha
55.
▲
by
4b11b4
4mo ago
Never say please but I do tell em they're going to do excellent world class foundational work
56.
▲
by
4b11b4
4mo ago
Haven't used Claude in a month. Haven't desired to once either. In that time I once asked Claude to review some shit but it was such over verbose garbage I wondered how I tolerated it for that long. On top of it, CC is garbage
57.
▲
by
4b11b4
5mo ago
It depends
58.
▲
by
4b11b4
5mo ago
I mean, idk why anyone is surprised. Was obvious goog was slow playing their harness
59.
▲
by
4b11b4
5mo ago
Opposite, I trust less and less the more I use these things. Must write certain things myself, especially types. Sometimes I'll even touch a new file myself (I know, creating files yourself?!) and write some pseudocode in a new file al
60.
▲
by
4b11b4
5mo ago
RL we're gonna find out will get abandoned cuz we don't even know what is getting "aligned", just my naive gut feeling don't take it seriously
More ›