m_w_

Although unsurprising given that there’s perhaps a trillion dollars at stake, it is disappointing that this article opens with “Anthropic’s redacted the best parts”. The table of shared/redacted sources is unfortunately dominated by no/partly.

Leaks, for what they're worth, indicate that the next GPT model is soon (tomorrow rumored), and that Fable 5.5 is already ready-to-go. That said Fable is likely being held to drown out said GPT release, since Anthropic is already ahead w/ Opus 5.5.

1. Claude Code (Ultra) + Codex (Plus) 2. Yes, but not for coding. In my experience Opus 5.5 is the only model I'd code with now, but other tasks (especially when not subsidized / part of product) use many other models. Latest deepseek checkpoint is very solid 3. CC + extensive config / hooks 4. Usually supervising myself w/ 5-6 sessions, if orchestrating more than that it's a hand-rolled swarm tool 5. Tried NanoClaw, and I have access to Claude Duet, but between getting on and off the waitlist for Duet, I hand-rolled a personal agent that does what I want.

I was hoping this link would be to VoidTools’ Everything [0] which - unlike this repo - is a very useful utility.

I’m all for recreating software, especially to be faster, cleaner, smaller, and more bespoke - but trying to one-shot a whole suite will never work. Building these foundational office-y tools requires an immense amount of testing and attention.

[0]: https://www.voidtools.com/

Best available does not mean “not actively misleading”, which it obviously is.

OpenRouter is not remotely close to a representative source for inference distribution.

This is an interesting benchmark - I like that time is limited, as it often isn't in other benchmarks.

It would be interesting to see the reverse, how long does each model take to get >50% WR against bots of various tiers.

Definitely would concede this, but I haven't had any issues so far. It's been able to make things look exactly as I've dictated them, even with relatively complex visualization of git branching. I get the sense that, with how capable Opus 5.5 is, "less capable" no longer means "not capable enough".

If it is, it hasn't been a problem so far. I have been focused heavily on memoization, to what extent that's possible when fully vibe coding.

Fwiw, what it's trying to replace (GitKraken) is slow, clunky, and resource hungry, so it's not hard to beat.

I would go one step farther - if you're vibe coding, skip Tauri altogether. I've been on-and-off vibecoding a GitKraken replacement using egui instead, and the memory footprint is very low. Otherwise can't tell the difference, as I'm not reading the code either way.

Great parody - very "Focus, M"-esque. Often, it feels that we've lost sight of "If it's not fun, why bother".

Well this reply is quite late, but in case you do end up seeing it - I’m not an expert on restoring them, but my understanding is the LEDs are generally the way, and if you’re capable, there’s a real premium paid for restored units.

The lights are generally not reliable on the 949 - particularly the radar display on the front-left (vumeter per channel, basically), so that’s also worth a service, if you can.

After that it’s the caps, as you probably know. The thing is with the quad ones, there’s literally two of everything in there.

Best of luck with restoring and selling it.

This article would have been much better if it chose a specific game and cheat to illustrate - I’m not sure “everyone cheats at a hypothetical souls-like” is relatable enough to demonstrate the point the author is trying to make.

Haha, it’s fitting that SharePoint is #1 here. If you want a glimpse into why that’s the case, look at the claims format [1], probably the most brain-destroying encoding format I've ever had to deal with.

Listing “5” for an email address, and then "4" for DNS and "6" for a gender much later in the article is perfectly emblematic of all the nonsensical decisions therein.

[1]: https://learn.microsoft.com/en-us/answers/questions/349797/u...

> but getting those two men into the same room at the same time can approach impossibility.

Yeah, you said it. I've also got a QX-949, but luckily no vintage refrigerators.

As much as the over-built, last-a-lifetime quality is a big component of owning of these old appliances, I can imagine how that fades when it's an object that weighs 800 pounds instead of 80.

Woah woah woah - don’t lump 1970s era stereo equipment in with that. If any of these friends have vintage Pioneer (SX-##50/80) that’s “junk” to them, I could take it off their hands ;)

Gee Eddy_Viscosity2, how come your mom lets you have TWO big boxes of cables? I'm sure I have multiple VGA cables in mine, and I'm not sure a single device I own even has a VGA port on it anymore.

> Along the way, it wrote 13 million lines of Lean and proved 29,500 intermediate theorems.

Pretty insane. I suppose it lends further credence to the idea that anything that can be shown to be correct can be done by a model.

The review that "IBM Bob isn't just another autocomplete tool" tells you everything you need to know about this product. lol.

lore does not make it a good or professional choice

For one, I think AI music is ontologically evil and have extreme opinions on the matter.

More than that though, I completely disagree that anything I'd ever choose to listen to is elevator tier / background noise. Just because it is playing while other things are happening does not mean it never receives focus or active listening.

This was a very pleasant read. There's (justifiably) a lot of debate about AI prose on this forum, which has become tiring. It's refreshing to read something so grounded. Doubly so while looking at a packet of Belvita breakfast biscuits on my own desk.

As far as lo-fi goes, I must recommend Leavv. Mind Garden and Tales of a Flowing Forest. Both pre-AI music albums that are lovely to work/study to.

This is super cool! I have always loved projects that try to categorize color space. Usually my work has gone in the other direction - isGray, isCloseToWhite, isPastel - but it's the same insofar as it's defining shapes like your matplot visual.

Wish I had this on hand like this three or four years ago when I was working on VR avatar software - we had hard-coded skin tone options at the time - would have been a cool addition!

Very impressive headline benchmark numbers. I expected a step change, but not past Fable. That said - it all depends on whether the classifiers make the model unusable...

This is a good idea - and I might pick it up for my writing, since a lot of it still happens in Obsidian. If a document is worked on across multiple days / revisions / app sessions - how is that handled?

It seems that they're trying to push up-market, or at least they were.

Given the extremely competitive releases of GLM 5.2 and DeepSeek V4 (both pro and flash), I don't think there'll be appetite for it.

It's a bit disheartening to see no comparison to other models here - and I'm not sure this pushes the curve anywhere. 3.6 flash is more expensive than GLM 5.2 - but seemingly worse, although this post is really light (lite?) on details.

It seemed for a time that Google had finally gotten the ball rolling, but I'm doubting that more and more as time passes. We'll see what happens with 3.5 pro I suppose.

> There is no joy of discovery of new music that you haven't heard. There is no connection to other humans through music.

Hard disagree here.

For new music, Discover Weekly is great, if you take some time to engage with it on a routine basis. Even better, if you have an artist/genre you already like, the Fans Also Like or Discovered On will link you to other artists and playlists. Super easy to go down rabbit holes of new artists and playlists.

As far as connecting with others, I do like the spotify DMs (in-app share), the friend activity tab, and particularly the share attribution. When you share a song via link (url with ?si=), you'll permanently be linked to it. For a number of my favorite songs, I see "From {friend}" at the bottom while listening. Makes me feel super connected to friends I've bonded with over music.

Yes - this is certainly true. I would have to concede that I could not speak to how my thesis would map onto a truly colossal codebase.

That said - I would (again, maybe naively) suppose it's not hugely different - much of the work I do occurs in code where many people have and will work on it, and where the size of the codebase dwarfs model context windows.

In that case, I feel the same - current frontier models, when properly oriented to a task, with some assist on the big-picture thinking - are more than capable of generating good code that can slot into big codebases with many moving pieces. Of course, I'd have to point to other people's work to defend this, but I think that's still pretty reasonable especially against the declared "LLMs are worse than useless for generating code".

Obviously I'm just one guy, but I maintain what I said w/ typing happily at 120+wpm. I think I could break 150 if I switched to colemak or dvorak, which I've been considering.

I don't think my job was ever to type fast, nor do I make any claim that I'm ontologically better than someone writing code by hand - but for what I need to do, I'm way faster now.

You might think he's an AI shill, but I was pretty compelled by simonw's post and idea - "your job is to deliver code that works" [1]. I think I can deliver more code that I can prove to work, faster now. The productivity is nice, but as I've said on other parts of the thread, it's also just fun to spend more time thinking, less time hitting the semicolon on my keyboard?

[1]: https://simonwillison.net/2025/Dec/18/code-proven-to-work/

Edit: should have used [1] in the first place instead of an asterisk.

Yes - I think that's fair. I'm happy to concede the term coding - I still enjoy the process of telling my computer to build something I've envisioned. It's now much faster, and I get to do more of the big-picture thinking and ideation, which I thoroughly enjoy.