roncesvalles

It's basically programming with really bad ergonomics. I'm a programmer by profession and I was immediately frustrated by Factorio because I realized I'd rather just write code and build a project because it would scratch the exact same itch while actually being useful.

It buys you time. I had my Anthropic key on Northflank set as an env var as-is and only found out about the incident from the HN front page 6 hours later (I didn't even get an email).

I rushed to the Claude platform console and was expecting to see a 4 figure usage bill or something. Luckily there was no usage.

But imagine if I were traveling. Even a rot13 of the key would've bought me enough time to rotate.

2 lessons:

- don't name your env var "CLAUDE_API_KEY". Call it "ARG_1" or something

- at least do some obscurin' so that the value doesn't start with "sk-ant..."

Your words are comment-shaped.

I had this exact thought yesterday. Shouldn't AI obliterate PaaS platforms like this one since it's super easy to build a bespoke CI/CD process now? Since it's bespoke, you can lock it down as much as you want. Attackers don't even know it's there. It's time to go back to `deploy.sh` and a VPS.

What you're not considering is that software implementation is a 2-way street. The process of reading what you wrote both elucidates something more about the problem, and elicits from you some aspects about the program's behavior that you do care about but didn't consider before.

LLMs for many months have been at a level where if you know exactly what you want, you can simply let the LLM bang out an implementation and never have to read it. But outside of simple problems, you probably don't know what you exactly want.

That being said, I've had considerable success ending every prompt with "ask me for any further details that I haven't considered" or some variation thereof.

I think "wrong answer" was the wrong phrasing. I meant it wasn't the best answer on the page.

It's a really intriguing phenomenon that on StackOverflow, the top-most voted answer was most often the wrong answer. My initial theory was that it's staleness. That is, it might have been the best answer when the question was asked, but if you're looking at a few years later the right answer changed. But it wasn't always that.

This has very interesting implications for everything in the world, btw. Correctness does not track majority consensus.

The public is the final "greater fool".

Video games are already movies that are video games.

>Software engineers

This Microsoft deal for the large part is not being offered to software engineering orgs.

That's ~41 million tokens. Your codebase better be the size of the entire MySQL codebase.

>do you really need SOTA models for everything?

For work in which you are in the loop, absolutely. Do people not value their own time or something? Why in the world would you use less than the best LLM available to you? You're not going to run out of the $20 plan even if you use Opus 5.5 at xhigh and code 8 hours/day (assuming you're doing it piecewise using a Claude product) so what exactly are you conserving by using something inferior?

For large continuous volumes of programmatically submitted LLM tasks, sure.

Jev's marketing can't be that good because as a dev I was thoroughly unimpressed by it. Isn't it just structured output? If you embed LLMs in your applications, you were almost certainly using structured outputs already. Maybe they did additional training on top of some open Chinese model, but almost always this makes things worse not better. There's no way Jev is better than a 5-5 series Claude model at any kind of classification task.

Like that old adage "that's a feature not a product" - Jev is just a system prompt, not a product.