Spyke

Replies

Comment on

Shoulda listened

Reply in thread

Yeah, the fact that so much of the language is conventions as well doesn't help. The type hinting is fairly poor as well, but then again type hinting is basically admitting that dynamic typing is a mistake (which it is).

games

Comment on

EA: Ruin Everything

Reply in thread

Despite combat being better and more polished in Andromeda than it was in the trilogy Mass Effect, Andromeda felt like a fan-fiction adaptation of Mass Effect. It did not feel like it was made by the same people. I feel like tying that game into Mass Effect was a mistake.

Comment on

Shoulda listened

Reply in thread

Yeah, Python as a language is a mess IMO. We used it for making a electrical simulator and the language is...bad. The variable scoping, lack of explicit variable declaration, backwards comprehension syntax, the VERY CONFUSING object orientation design, that classes aren't serializable by default, package story and deployment... It really has very little going for it, and combine all that with the awful performance characteristics it's just...no.

Today I'm "Rust if it fits, if not Rust then C#". We re-wrote the simulator in C# and performance was 70x better and the code got shorter, the deployment was hell of a lot easier since we went AOT and it just builds a native binary for the target platform. No virtual environment crap.

Comment on

Impersonation rule

Reply in thread

We had a Svelte application. We chose svelte because it's simpler than React is, and it's much faster. However further down the line a consultant said to the CTO that we should move to React. I said React is more complicated and has more edge cases, and its slower, but the consultant said more people know react and it has a bigger community, so we switched to React. This was fairly early in the project. React has sooo many edge cases. It's not made with async in mind, if you forget to add a [] to a useEffect it just breaks your application completely instead of doing the more sensible thing, it says "don't use useEffect" but there's so many "except when" that you end up using it everywhere regardless. Later on the CTO got all in on Claude Code, so we spent the easter holidays rewriting the app in Svelte. Except a few bugs it was feature complete by that week and in production a few days later.

It's much, MUCH faster now, and the code is much simpler. We lost absolutely nothing by moving back to Svelte but gained a lot.

Edit: not that using either of these is sane thing to do in the start menu. The start menu should not be a web application.

fuck_ai

Comment on

No, AI Is Not "Autonomously Hacking" with Cal Newport

Reply in thread

Yeah, imagine if you let your gasoline powered lawnmover just run around in your yard just to see what would happen and you just left it for a couple of days in the anticipation that eventually your entire lawn was cut only to find that it also ran over some kid. The lesson from that wouldn't be that your lawnmover is just too good.

news

Comment on

FTC chair suggests AI developers should be liable for conduct of agents

What angers me is that people are buying into the AI shit that the companies are peddlings. These aren't sentient machines with a will of their own or even any mechanism to do anything on their own. These shit-heels are taking a text generator, trained on doing all sorts of shit, and piping its results into a bash-shell. If you just don't do that they can't do any of the damage they did in OpenAI's case. They try to make it look like they're figuratively breaking out of a cage or jail. That's not what's happening.

They're given every tool they need and then they add clumsy and inefficient filters on the output text they're writing. It's negligent incompetence.

But they do this for a reason. They want legislation to stop what they're doing. OpenAI specifically does not have an avenue currently where they end up making money, so they want everyone to stop training new models so that they can focus on making a positive cash flow rather than competing.

People need to stop eating the bullshit these companies are shovelling. Yes, if you write an application that breaks into the pentagon you're looking at jailtime. If you host a LLM that does the same thing it's the exact same issue. If it did break in then you weren't paying attention (i.e. being negligent). It's not a complicated problem.