jedbrooke

audio CDs use a different wavelength for the laser (infra red) while PS game disks are BluRay (blue laser).

Of course, 99.9% of other BluRay players will also support DVD/CD, so yeah it still does seem silly.

I also remember my Dad’s disappointment when he put a standard DVD in our Wii back in the day. Those are literally the same physical format as wii disks.

OCRing handwritten dates and printed timestamps seems useful, but how is “AI” supposed to know when a photo is taken just by looking at it?

At least the scan date is a real piece of data that could be useful. This seems like it would cause more harm than good by polluting the data with nonsense, unless the added dates are clearly labeled as estimates

from what I understand prismml isn’t doing a quant like normal models where you take a model trained at fp16 and then chop off some bits to reduce vram, but rather they’re training the model natively with 1 bit weights. It’s explained more in the article. They’re also doing some other tricks like a fp16 weight per block of 128 1bit weights to get some more data out of 1 bit weights

I find they still get blocked occasionally :(

I find sonnet and haiku will do that, but I rarely get that with opus. When it does happen it’s a good cue to start a new session. That’s another benefit of the style guide is it’s a good canary for when the model has gone off the rails. If it’s left the style guide, the rest of the response can likely be discarded as departed from reality.

Some small part of me has started to dread reading LLM output because I know what I’m going to find. False assumptions and hallucinations. Emphatic, staccato fragments. Excessive emojis :sparkle:. :rocket: It’s not just me—these are real patterns (:barf:).

It takes like 5 seconds to add a quick CLAUDE.md / AGENTS.md with a quick style guide, or even just “EMOJI ARE FORBIDDEN” and I find it makes llm output significantly more tolerable. That and a quick style guide with some banned words and phrases.

That won’t help with the false assumptions though, gotta use old fashioned careful reading and critical thinking the catch those.

already been tried: https://github.com/samrolken/nokode

Maybe LLMs will eventually get to the place where it replaces even the OS kernel

oh I had thought it was opt-in, but it wasn’t. Yeah ignoring licenses is significantly less cool

Supplies are limited, and the first 1,000 eligible submissions will receive one. Limit one per person. Availability may be limited by country or region.

This seems quite limited to be a real product, but also quite a lot for seemingly what’s just a joke to mock Sony for ceasing blu-ray production.

It’s worth mentioning that github does (did?) do some cool stuff with physical media archiving of code like the arctic project [1], but these CDs are burned, not pressed so they’ll only last around 10 years

[1] https://archiveprogram.github.com/arctic-vault

reminds me of the x264 “placebo” encoder setting

https://trac.ffmpeg.org/wiki/Encode/H.264#FAQ

Dolphin only supports Wii games, not Wii U. This is part of the reason the Wii U flopped is that people thought it was just a peripheral for the original Wii and not a separate new console.

Dolphin is indeed fantastic for Wii/gamecube games though

looks like the macOS one is Tahoe only. I’ve been putting of upgrading to tahoe but this might be enough to tempt me

I’d been thinking about if something like this would be possible for https://chatjimmy.ai/ . The underlying model is only llama 3 8B but I’m curious what coding harnesses would be like at 17k tok/s

for me the most annoying one is “escape hatch”.

Everything is an escape hatch, try catch is an escape hatch, a cli flag is an escape hatch. It makes no sense, and quickly ended up in my “banned words and phrases” md file

I know this will never happen, but wishful thinking has got me thinking about the “perfect” ad algorithm. Advertisers don’t want to waste money on impressions that won’t convert. If AI got smart enough that it would truly know which ads would convert and which would not, then the “perfectly relevant ad” would only be stuff you actually need, and if you don’t need anything, no ads!

of course, this will never happen

Fun fact, an electric guitar can also be used as a microphone if you shout in to it loud enough.

We discovered this while fooling around with some guitars and such as teenagers. We had a 4 track input device that was separating vocals and instruments, but even after turning down the vocal track, we could still hear it in the instrument track. We then of course followed it up with some experiments deliberately shouting into the guitar and enjoying the distorted recordings that came out of it

Honestly, I get pretty good improvement from just adding a “Emoji are forbidden” and a small list of banned words and phrases (the usual suspects like “it’s not just x, it’s y” etc)

We’ve come full circle and started using Graphics Processing Units to process graphics again

"alt+cmd m j/k" (media -> vol up/down)

if only keyboards came with built in buttons for adjusting the volume… oh wait. Unless of course you are suffering on a touch bar mac, then I completely understand.

looks interesting, I agree that chat is not always the right interface for agents, and a LLM boosted cli sometimes feels like the right paradigm (especially for dev related tasks).

how would you say this compares to similar tools like google’s dotprompt? https://google.github.io/dotprompt/getting-started/

I tried it in Chinese and ChatGPT said No, and then gave a history of Saint Nicholas

Oof, off topic but the trains were out of service here for my commute last night so I though from the headline this meant that somehow all trains everywhere just stopped working. Glad to see it’s just some Saas product that’s down

the windows version is playable on macos through wine. Even modern version, I got it running on a m2 mac mini on Macos 15 sequoia

EDIT: this was for HL1 I’m not sure about HL2

so many times I catch myself asking a coding agent e.g “please print the output” and it will update the file with “print (output)”.

Maybe there’s something about not having to context switch between natural language and code just makes it _feel_ easier sometimes

I've found that a 250w incandescent bulb (can be had for ~$10) paired with a 4000 lumen LED produced decent results on a budget. Search for "reptile" or "chicken" lamps, they are usually red. You can feel the HEAT from a 250w light bulb.

The only thing to watch out for is that the lamp base you're using can support the high wattage.

maybe he’ll stop, maybe he won’t. There’s no way to know

I still don’t know how openAI thought it was a good idea to have a model named "4o" AND a model named "o4", unless the goal was intentional confusion

the author mentions they’re using haiku for the model, but I wonder if the travel query -> place transform could be done with a tiny local model.

But that’s the bitter lesson I guess, unless there’s a reason to go fully local (extreme privacy concerns, offline use, etc) training a custom model just isn’t worth it over using some cloud api offering, even if it is orders of magnitude more compute that _somebody_ is paying for in the end

For me there’s something about context that shapes my dark mode vs light mode preference. Code and terminals is dark, documents/web/chat is light. For some reason I can’t stand dark mode slack/discord, but dark mode IDE is preferred.

Also, there’s just something about the graph of “Average brightness of MacOS screenshots over the years” and extrapolating it that tickles my brain. By 2030 MacOS light mode will just be a single white rectangle (with a notch). It reminds me of the “Number of youtube videos on the homepage” blog that extrapolates that by 2030 there will be 0 videos on the homepage.

Thanks for bringing back memories of the Netflix Wii Channel. At the beginning it was on a disk. We used the Netflix wii channel until the day they dropped support. Our Wii long outlived its life as a games console by continuing on as a netflix machine. I still miss using the actual pointer to point at things, it’s just such an intuitive interface for a TV

EDIT: I just looked it up and apparently the wii netflix channel was supported until 2019, so my memory of using it until it went bust were incorrect. We prob used it until around 2012 or so