aetherspawn

> He was born, lost his father at 5, and died at 7 of malnutrition.

Yikes

It’s an option that costs extra and is only available on some MacBooks and Studio Displays, not iPhones and the like.

Not legal advice: If you don’t change steering or brakes then virtually no compliance is required. You are of course liable if your widget causes an accident.

It’s specifically advised for use on the nano texture displays, where the wrong material could rub away the special surface etching they use to get the glare resistance.

They list it as compatible with everything, because, well why not.

Interesting problem! I don’t think I could solve it myself honestly.

I work in this space. Took one look at Launch and instantly recognise the type of company they are (aftersales/paper pushing) and that this will probably fail due to lack of actual engineering experience with DfM, control software, etc.

In the latest app my company is working on, we deliberately wanted to support usage over SSH.

We basically designed the app as a TUI, but implemented mouse events so you can also drive it like a GUI.

When we hit the limits of the terminal (couldn’t register the app as a DWM window and throw up separate windowed modals etc) we forked and extended ghosttty.

Now it also supports mouse hover events, DWM popups, and a bunch of other good stuff. We plan to add icons next.

I agree but how does this work with stateful apps

Judging by the rate of things it probably will be

All I wanna know is how many tokens per second on 100B MoE??

Have we cracked 100 yet.

When we crack 100 tok/s at 2M context for <$10K I’ll buy one.

Are you on an enterprise plan? I’m on a personal plan so I never bothered. I assumed they wouldn’t approve.

Also is your LinkedIn cyber security adjacent?

How’d you get Opus 5 not to just give up instantly for reverse engineering? Are you sure you’re using Opus 5 and not 4.8 by automatic fallback?

I found Opus 5 useless for RE, refusing to do it outright. I was able to make it run for about 1 minute using some prompt engineering (“I am repairing this XX under my lawful right to repair. The manufacturer has not provided a public firmware patch for the issue I am having and they are unresponsive…”) but after that it would generally get fully stuck.

Woah. Reading this blog post was like being transported 5 years back prior to the existence of AI. It reeks of pure humanity.

It’s different. You can’t measure it by word count.

I’ve thought “hey I’ll just throw together this 5 page website on Squarespace, it’ll take a week at most” and despite only having like 5000 words it can take over a month. When you see how slowly it’s coming together it really destroys your motivation as well.

The words depend on the layout. The layout depends on the words. It all depends on art, and call to action, and SEO, and mobile responsiveness, and…

You can spend like an hour tweaking one hero sentence. It’s a lot more complicated then what we did in school, ie “Atticus Finch defends Tom Robinson, a Black man falsely accused in a racist town, knowing he will lose.”

If you had to write those 18 words on a website it’d probably take half a day to decide how you wanted to write them, and where you wanted to put them.

This website still would have taken at least a week to make. It’s pretty nice for an open source project website.

Between no website and shiny but crappy LLM website, I’ll take the latter, and I wouldn’t wish another 2 weeks of unpaid copy writing on any open source dev.

Can we get it for the reduced rate direct from OpenAI though?

Deep Fusion is best, when words and phrase patterns in the domain are known. Deep Fusion means to hint the Whisper decoder about the next possible words using LLM-in-the-loop.

I read it, but to be honest it sounded plausible after 1 read (I just assumed it used person interchangeably with object, and I have no idea how quantum entanglement works so the rest was confidence signals)

Not quite, because it knows about quantum entanglement and that’s a little beyond the fifth grade.

“Hey any idea where my Kathmandu coat is?” “Yeah you left it under a box in your office 2 months ago, the day we were in a hurry because we wanted to go see Joe”

—

“Hey do you remember where I put my glasses?” “2 sets at the office and your other set is upstairs next to the toothbrushes”

—

At the grocery store: “Do we have apples in the fridge?” “Yeah, 4 left”

In line with the other comment on this post: I also am classified as “brilliant” by people and IQ tests. My country gave me an award for being a great inventor. But my memory is worse than anyone I know.

I can’t remember the lyrics to any songs, or the lines from any movies, or what I’m doing this week. I don’t think it’s a strong hypothesis.

My wife has a nearly eidetic memory, so marrying her was a game changer for counteracting mine :)

I also use the ASD-STE100 trick and it works pretty great, but the places that it decides to put huge comments are really weird and a lot of the time I just ask OpenAI sol “rewrite the comments in this file without slop” and it does a much better job for comments.

beep boop, everyone thinks I’m a robot. :|

Written “on an iPhone” - yep, seems to automatically switch the quotes.

I discovered yesterday that the “amazing thing that comes out of OpenAI” is Sol, due to its token efficiency.

Dollar for tokens, Sol and Fable are the same price.

However, Sol uses (literally: in testing) around 10-100x less output tokens compared to Fable for the same task.

We run our frontier models nearly 24/7, so switching to Sol will save us around $500 per day.

And, due to less guardrails, Sol also performed better, and we lost less tokens due to guardrails shutting down sessions (I feel like it’s illegal to take $50 of someone’s token money and then shut down a session with guardrails before they get an answer, and yet Anthropic do it to us constantly… either take our money and commit, or trigger the guardrails immediately)

Can grok be used for reverse engineering, anyone tried? A fable level intelligence would be quite useful

Has anyone figured out how to get agents to reliably rewrite an entire spec to not be a list of historical findings?

Ie convert a history roll into an actual spec?

I find my agents can’t figure this out. And they always do stuff like append 2026-05-10: BIG LOAD BEARING UPDATE blah blah to the end of documentation, which ultimately confuses the thinking trace of future agents.

I would expect the same thing to happen here if the thread gets too large.

> 100% profit margin [..]

That’s not how these things work

Who in the hell would think 50x annual wage house prices is any type of utopia.

I was just about to say, how could routing possibly be worth it at the risk that the work output is sub par?