danielbln

6 ms·

61.
by danielbln 5mo ago
Which game would that be apart from Minecraft?
62.
by danielbln 5mo ago
Ed Zitron is a terrible source, he is so staunchly anti AI that he is effectively blind. Once in a while he'll be right by pure chance, but I wouldn't rely on any of it.
63.
by danielbln 5mo ago
Seems like that NFTBro hit a nerve with this art lover.
64.
by danielbln 5mo ago
My local model humming next to me will always be available. Is it as good as a foundational model? No. But it'll work just fine for most pedestrian tasks and I don't need to keep now useless mechanical knowledge in my brain.
65.
by danielbln 5mo ago
Same in Berlin, can now saunter through with liquids. Then there is Ben Gurion Airport TLV where you could always take liquids with.
66.
by danielbln 5mo ago
A highly beneficial side effect is that it disables the main YouTube shorts stream. You still get some shorts for your subs and such, but the bottomless heroin feed if Shorts is gone, and good riddance.
67.
by danielbln 5mo ago
Ah, I see, you weren't making a point, fair enough.
68.
by danielbln 5mo ago
I came around from fancy everything to just reasonably good beans and an Aeropress. And in the office I'll just saunter to the press-a-button-get-coffee machine, I eventually got tired of the faffing around.
69.
by danielbln 5mo ago
I have the glass and metal Aeropress, so if you fancy the Aeropress but don't fancy boiled plastic, that's an option.
70.
by danielbln 5mo ago
I don't disagree, but I'm not sure I see the point you're trying to make.
71.
by danielbln 5mo ago
prompt + all other bits of information the context has been seeded with before the output was created (documents, web searches, other sources) in which case it might be more efficient to just consume the final deliverable (yourself or via L
72.
by danielbln 5mo ago
Ironically a task that an AI agent would have no problem doing.
73.
by danielbln 5mo ago
There is also discussion, ping pong with the agent, exploring parallel paths, quickly experimenting, analyzing code, researching things. A code agent can do more than "write me as much code as possible, go!".
74.
by danielbln 5mo ago
Alphabet makes $30 billion profit per quarter.
75.
by danielbln 6mo ago
Why don't you enlighten us?
76.
by danielbln 6mo ago
At least oats don't have to be perpetually kept pregnant while taking their offspring away from their mothers. See, snide comments cut both ways.
77.
by danielbln 6mo ago
So, how does aj engineer new to a code base add new features? They read the code base, understand the architecture and structure and make changes. An agent can do the same.
78.
by danielbln 6mo ago
Only a Sith deals in absolutes.
79.
by danielbln 6mo ago
Yeah, LLMs are a solution to the cold start problem plus they are easy to integrate and if you know what you're doing in terms of evals, post processing and so on you can get excellent performance out of them, plus they can do semanti
80.
by danielbln 6mo ago
Im a software engineer and know how to be a software engineer, yet I find LLMs quite helpful. Why should a plumber be any different.
81.
by danielbln 6mo ago
Production ready!
82.
by danielbln 6mo ago
You have these grifter videos for everything, from OpenAI features to hot takes on the strait of Hormuz. These are best ignored entirely.
83.
by danielbln 6mo ago
It's also more often than not good enough, which for a specialist is bad, and for most everyone else is absolutely sufficient.
84.
by danielbln 6mo ago
Also in Europe and can only agree. Granted I'm on the 20x plan, but I have yet to hit a limit once and I'm using Claude 12h+ per day on multiple projects.
85.
by danielbln 6mo ago
git log --all --reflog -- path/to/secret-file
86.
by danielbln 6mo ago
There were no "dark ages", that's the same common wisdom blunder like "in the middle ages everybody was dressed in drab grey clothing, ate gruel and walked through mountains of poop everywhere". It was a time of tra
87.
by danielbln 6mo ago
Open weight models that run under your desk are not frontier model level, but they are getting closer. Improvements in agentic post training and things like TurboQuant mean that even if all frontier labs pull the plug tomorrow, we will stil
88.
by danielbln 6mo ago
Every AI/agentic thread on HN follows the same tension: builders want to build and solve problems. Code or task completion are implementation details to be done on the path to the actual prize: solving the problem. And then there are t
89.
by danielbln 6mo ago
How often do humans make mistakes? That's the better comparison.
90.
by danielbln 6mo ago
That's why I drive with 10 pounds of paper maps in my car. I won't have any of this new fangled GPS tech atrophy my map reading skills that I've honed so much.

More ›