geraneum

4 ms·

61.
by geraneum 2mo ago
I don’t watch sitcoms anymore. I come read HN for these gems.
62.
by geraneum 2mo ago
> outcomes are evaluated purely empirically How does that look like?
63.
by geraneum 2mo ago
I’ve read LLM outputs on a piece of code or topic that I already understand or hand written before, and lately, it’s so confusing sometimes that I need to reread a couple of times or focus too mich to decipher the writing style.
64.
by geraneum 2mo ago
We, developers/tech-employees (safe to assume since we’re on HN) are the “entire class of workers” that the article is talking about. The premise is losing faith in the future of our profession. Which is something that’s openly adverti
65.
by geraneum 2mo ago
Remember, our misery is some potential investor’s financial prospect. We read these articles with different eyes. There are people who are the real audience and salivate when they read these sentiments.
66.
by geraneum 2mo ago
The author has discovered the red tape.
67.
by geraneum 2mo ago
Blimey, it’s a mystery why they don’t let me take my jigsaw to the “community of hand saw enthusiasts”.
68.
by geraneum 2mo ago
> observe that the agent day after day do handle user input safely How do you observe the issues that aren’t apparent via a GUI? Do you notice the circular logic in your reasoning?
69.
by geraneum 2mo ago
> I don't review code that either works or doesn't - most HTML and CSS layout code for example. There I test it on desktop and mobile and commit it if it works. Good example of what not to review if you're working on your
70.
by geraneum 2mo ago
> It's novel because previous rounds of automation were about automating specific tasks or well-scoped functions. Evidently, this has not really changed with LLMs and coding agents. It’s what AI companies are betting on though.
71.
by geraneum 2mo ago
Does the code get reviewed? How do you deal with increased amount of code that may need to be looked at?
72.
by geraneum 2mo ago
Unfortunately people sometimes get defensive against this take. But I think treating the LLM as you described can make you a better LLM user and help get better output. It helps understand the failure modes better, and moderate one’s relian
73.
by geraneum 2mo ago
Then ask it to fix it. When “fixed”, ask the same question again and you’ll get a similar response again!
74.
by geraneum 2mo ago
I’m terribly sorry on their behalf. I hope the expression of their experience has not hurt Claude’s feelings (IPO valuation). Won’t happen again.
75.
by geraneum 2mo ago
> Never write READMEs, docstrings, or comments. I will write those myself later. And yes, I really mean this. The precise and rigorous practice of “engineering” in 2026.
76.
by geraneum 2mo ago
That pesky “scientific method” bothers these corps sometimes, so one can play fast and loose with numbers and conclusions for all sort of reasons. No need to worry about reproducibility, rigor, scrutiny, etc. not good for IPO and stocks.
77.
by geraneum 3mo ago
This is what you get when you prompt claude to avoid –
78.
by geraneum 3mo ago
You’re absolutely right to push back. It was indeed a school. I’ll save <schoo_name> in my memory for the future.
79.
by geraneum 3mo ago
Stata is better in statistics than we are. Why doesn’t that produce guesses?
80.
by geraneum 3mo ago
> Anthropic has never advocated for a ban on open-weights models. If you wonder why this is written as an opening to a list of reasons that advocate for banning the open weight models, it’s because
81.
by geraneum 3mo ago
How does the AI guess?
82.
by geraneum 3mo ago
Even on a less dangerous level, repeated mistakes can be detrimental to your business. Imagine a typical ecommerce app for selling anything. Mess with people’s orders and money and you’ll go out of business or get sued or get a bad reputati
83.
by geraneum 3mo ago
It might be because it was going out of its way before and had too much of a blast radius, and now they could have changed the RLHF (or other tricks in their sleeve) to get what you're seeing now. The reason it swings is that they can&
84.
by geraneum 3mo ago
I read this as the model being less steerable. I was bitten by this just today where I had a local Postgres instance running, and prompted opus 5 to run a server against it, but forgot to give it the password. Instead of asking for the pass
85.
by geraneum 3mo ago
What better way to spend token. I’m amazed that people talk about this as it’s a good thing!
86.
by geraneum 3mo ago
Why not make the agents make what you want with marimo?
87.
by geraneum 3mo ago
> It largely works and it's a massive business success. The engineer is suggesting that it could be done cheaper and maybe with better outcome. Ironically, this is a classic business case.
88.
by geraneum 3mo ago
If you’re puzzled as to why this exists, imagine that, out of the goodness of your heart, you donate $230 to OpenAI to support their mission of rear ending the singularity, and receive Codex Micro memorabilia as a token of appreciation.
89.
by geraneum 3mo ago
It’s because they don’t pitch to the users. They pitch to the investors. Have noticed how everyone and their dog now has an “AI story” on their website? Yeah, because they won’t get funding otherwise. And these are not even the AI companies
90.
by geraneum 3mo ago
I have this funny feeling that someone’s probably gonna ask their favorite frontier LLM about the fruit salad thing to refute your point.

More ›