hdjrudni

I hired a guy to assemble a desk...he did a shit job. Won't be doing that again.

Amusingly, when I know my peer is just going to point his AI at my feedback, I write for their AI, not for them. I'm much more curt. Maybe not so amusing but I don't feel bad about dumping a laundry list of fixes for them.

I don't even want to put 'my' code or apps out into the wild anymore. I've built a a few things I think are useful and I'm using myself but I'm afraid it'll just get called slop or my only users will be bots. What's the point. So I don't get the joy out of sharing it nor the joy of the achievement. But I've widdled down what I thought was an endless backlog of features to nearly zero, I guess that's something.

This looks pretty nice for humans too actually.

The "for agents" is a bit of a turn off for me. Are they just trying to jump on the AI hype train or is this really AI-centric? Because if its for agents and only agents then I guess I should expect it to get progressively worse for humans but its for humans & agents then maybe it stays a nice clean human-friendly language.

It's actually kind of funny that the purpose of the Turbo button was to slow your computer down so you could play games that were running too fast.

But I guess they didn't want to put a "Slow" button on their PC towers.

I did that too but not for the sake of the LLM, just for myself. Massively helps discoverability. It also protects from backwards-compat if you ever want to change a default without breaking existing users, knowing that whatever they had before has already been saved into their settings so some setting doesn't randomly change on them.

My work recently changed an agent API. It was a certain CLI tool built for agents which I had copied into one of my Skills.

What they did was kind of clever. When the agent calls the old API/uses old args, I guess they told it to tell the human to fix their skill and 'upgrade' to the new API. They even provided a skill upgrader skill.

Anyway, my agent tried calling it the old way, must have received some sort of error message and then diligently reported to me that my skill needed upgrading and we should be doing it some new way.

Kind of like reverse feedback I guess? Not the AI giving feedback to them, but them giving feedback to me via my AI.

There's probably some efficiency to it too. Encoding a dozen images as one and load them in one request will be quicker than 12 separate images each with their own overhead.

Really? <map> was also supported since Chrome and Firefox 1. Is ismap older yet?

https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/...

I don't like that semi-colon. AFAICT `i, s;` is not an independent statement. `i, s: vec` would have been fine, but `;` ?!?! madness.

That one actually shows what the variable is initialized to. This one doesn't. reply

Which variable? `i` is clearly initialized to `0` and `it` doesn't have an initializer per se... it starts at the first value of the container.

It almost seems like an abuse to put 2 unrelated things in the `for(...)` but that's what they did.

I've been wanting to make an RTS but found it a bit daunting. I thought this might be a bit challenging for LLMs too but I guess not! This is nice to see.

I was hoping it’d end in a “so how much did this cost?” so that others team looking at similar migrations have an estimate on what they can expect

It was in the middle. $165k.

Isn't that equivalent? The task here was port Bun from Zig to Rust. He's saying that task will cost more.

It's only 4% unsafe and most of it is single-line pointers that came from C++

At the time of writing, about 4% of Bun's Rust code sits inside an unsafe block (~13,000 unsafe keywords across ~27,000 lines / ~780,000 lines), and 78% of those blocks are a single line — a pointer that came from C++, or one call into a C library.

I don't know much about Rust but I imagine this is safer than 100% 'unsafe' code in Zig or C++.

Box3D just showcased some stuff including deterministic replays. If you wanted to port that, you could probably import the replay and make sure it plays back the same way in your new language. I think it captures the inputs and forces applied, not the pixels.

I suppose rendering is a component of a game engine too though, not just physics. I don't know how to do that reliably. Even if you captured pixels, it'd be annoying. If you've ever tried doing screenshot based diffing on web you will know that slight changes in aliasing in Chrome bugger everything up. Things that should be equivalent randomly aren't but not in a way that any human would care.

At first I was really impressed, I thought granny was the voice. But it turns out it's the same kind of annoying voice and tone. And then Constance starts talking to it and it immediately cuts her off at 1:05, after they just explained it was better at conversation flow.

Seems a bit disappointing, but the 3 overlapping questions example was impressive.

I don't think it needs an 'internal clock'. It needs two things. (1) the timer tool, (2) to know not to lie about having a tool or using a tool. I think (2) is the hard thing.

The two biggest hardware challenges are reducing the size and weight of the ultrasound probe and replacing ultrasound gel with a more practical coupling material, such as hydrogel. We think both are solvable, making it possible for the probe to eventually become a lightweight wearable or adhesive patch.

Not sure I'd want to put an adhesive patch on my neck every morning so I can silently talk to an LLM in the cubicle farm. I hope this is not our future.

Very cool tech though and surprisingly good results for so little training.

I think time might be better spent improving a lip reading model (no adhesive required), assuming we're unable to read brainwaves directly.

The entire article is speculation. The real answer was posted by Ryan Cavanaugh in the official repo over a year a go: https://github.com/microsoft/typescript-go/discussions/411 Why would trust anything else?

I've used a very similar workflow somewhat by accident for some larger changes. I'll be chatting with Codex about the best way to implement and watch my context slowly dwindle down, so I ask it to write out the plan to some dir (e.g. tasks/some-feature/). I recently had it write out a 10 step plan, as in 10 large, detailed files. I just ask it to implement the first file in a fresh context, check it over, get it to fix anything up, then start the 2nd file in a fresh context. It was smart enough to document 'deliverables' and verification for each step. Worked relatively well.

Might not be a bad idea to formalize this process with a skill.

What does "clean room" mean here and why is it important?

Also the task list (checkmarks) render poorly in PDF.

Also why wouldn't rendering the same doc twice produce the same byte output?

This one is a fair point... if all the characters have distinctive names then NPC can plausibly become aware that it is a character in that game and spoil things. I don't know how to stop that, short of training a clean model from scratch, but that seems a bit much for a game.

What "context"? If it's coming from the player's input, then the LLM isn't spoiling anything.

e.g. let's say an important character dies in your game but that hasn't happened yet in this point in the story. You're talking to some NPC. That NPC can't possibly spoil that that character is going to die because the developer wouldn't feed that in as context. The NPC doesn't "know" the character is going to die. You, as the player, might be able to convince the NPC that that will happen and it might respond accordingly, but it won't just randomly divulge that by accident.

Oof, why would you sign up for only the standup comedy specials? 90% of them are garbage. These people are not funny.

Q,U,V,X,Y,Z are interesting. Not really sure how to read the map though. Does black mean its placed there the most or least?

Love to see this! I got started with Box2D back in probably 2006ish. Great to see Erin is still working on this stuff. Thank you Erin for the great libraries!

Codex is not better anymore. It appears they nerfed their quota a few weeks ago. I never used to hit my 5 hr limit, now I always do. Sometimes in like 2 prompts.

Nope. I currently have 64 GB which is about enough. I'd do 128 or more if it were cheaper. But a decent stick size is about 16 GB. If I could get 48GB for the price of a 16GB stick, that'd be reasonable. Then I could get 2 of those instead of 4 16s and not pay horrendously more.

Then do `deno ./my_downloaded_deno_gui` instead. If you trust the copy of deno you downloaded, then hopefully it can be trusted to verify the permissions of random downloaded deno-apps.

Yes, it kind of defeats the standalone binary aspect but if you're really concerned about security, maybe it's a happy medium.