redfloatplane

I was hoping that the video was a walkthrough of your process - do you think you might share that at some point?

> I'm not a programmer anymore. I'm something else now. I don't know what it is but it's multi-disciplinary, and it doesn't involve writing code myself--for better or worse!

Yes, I agree. I think the role of software developer is going to evolve into much more of an administrative, managerial role, dealing more with working with whatever organisation you're in than actually typing code. Honestly I think it probably was always heading in this direction but it's definitely quite a step change. Wrote about it a little incoherently on my blog just this morning: https://redfloatplane.lol/blog/11-2025-the-year-i-didnt-writ...

Well, on my Macbook, some of that already exists. In the Shortcuts app you can use the "Use Model" action which offers to run an LLM on apple's cloud, on-device, or other external service (eg ChatGPT). I use this myself already for several actions, like reading emails from my tennis club to put events in my calendar automatically.

Whether or not we'll see it lower down in the system I'm not sure. Honestly I'm not certain of the utility of an autonomous LLM loop in many or most parts of an OS, where (in general) systems have more value the more deterministic they are, but in the user space, who can say.

In any case, I certainly went down a fun rabbit hole thinking about a mesh network of LLM nodes and thin clients in a post-collapse world. In that scenario, I wonder if the utility of LLMs is really worth the complexity versus a kindle-like device with a copy of wikipedia...

An interesting thought experiment - a fully local, off-grid, off-network LLM device. Solar or wind or what have you. I suppose the Mac Studio route is a good option here, I think Apple make the most energy efficient high-memory options. Back of the napkin indicates it’s possible, just a high up front cost. Interesting to imagine a somewhat catastrophe-resilient LLM device…

Well, it’s very much appreciated. So much of the weird one-off experimentation seems to happen on sites like that, and otherwise I’d have to either lump it or eat the radioactivity. It’s an interesting thing that even though I do plenty of my own weird little experiments and similar tool-building escapades as you, but I rarely post about them on my own blog (https://redfloatplane.lol/blog) (and thus, nowhere). Perhaps it’s that posting them on my blog feels like taking more responsibility than just saying “tried this experiment lol” on twitter.

> Simon often finds ideas within walled-garden platforms (e.g., TikTok, Twitter) and simply brings them to the open web

I find this is a surprisingly valuable thing. The AI space is moving fast, and a lot of the interesting, imaginative experimental stuff is happening on Twitter, Reddit, and other platforms I really don't want to engage with - but I do want to keep roughly up-to-speed with what's happening there.

That camera analogy is very thought provoking! So far the only bright spot in this whole comment thread for me. Thanks for sharing that!

Sort of related: About ten years ago there was a device called the Tessel by Technical Machine which you programmed with Javascript, npm, the whole nine yards. It was pretty clever - the javascript got transpiled to Lua VM bytecode and ran in the Lua VM on the device (a Cortex M3 I believe). I recently had Claude rewrite their old Node 0.8 CLI tools in Rust because I wasn't inclined to do the javascript archeology needed to get the old tools up and running. Of course then I put the Tessel back in its drawer, but fun nonetheless.

"Fun" fact: This is partially because Wolfram owned a patent on the "code notebook" invention. Once it expired, Jupyter was created.

It's quite hard to know if the DAC is actually decent quality though. I've bought two from Amazon (admittedly at the low price point) and both of them have line noise - one of them even has a ground loop buzz, which surprised me, since it's powered by USB-C. I'm unconvinced that any of the higher price points (that are still within my budget) aren't just these cheapo ones in slightly fancier cases.

My old TV had real analogue out for speakers and it really did sound a lot better than what I've been getting through TOSLink and this cheapo DAC. Same Hi-Fi and speakers. I'm sure the problem could be solved with a more expensive DAC, but which one? How could I know?

I find this is one of those things where it's quite hard for the uninitiated to see through the cloud of 'audiophiles' saying that you must buy gold cables or your audio will sound like garbage, and still getting decent quality audio.

Yes, I’m sure! I had mostly stopped using WL by the time ChatGPT came out and so my only experience of their integration is a lot of hallucinated syntax by GPT3. I’ve not been tempted to upgrade to the newer, more LLM-integrated versions yet, particularly because I’ve been trying to cut down on expenses and spending more money on LLM api calls versus just using my existing Anthropic subscription for Claude code isn’t that appealing. I’ve also been going through a bit of a crisis of a lack of creativity and inventiveness and it’s hard to decide to spend money on an update when it doesn’t seem like I have any good ideas of what to do with it after I’ve spent the cash!

I bet that if they can integrate LLMs _really_ well (I’m not sure the chat driven notebook thing is necessarily the way) it’ll be a massive upgrade.

I'm a non-mathematician and I used it for lots of novel stuff - GIS, visualisations of all kinds, machine learning. The Wolfram Community staff picks is a great introduction into the varied things you can do: https://community.wolfram.com/content?curTag=staff%20picks

I think it's an extreme example of not-invented-here syndrome. In many ways that leads to interesting novelties, and in others it leads to not having undo/redo until the 2010s.

That's exactly the same analogy I used to use, although I said "nuclear reactor included" - spaceship is better, it implies less danger and more expanded horizons!

Just before I stopped using Mathematica they came out with that headless kernel, and I had wondered if you could spin it up on a Kubernetes cluster or something.

I do notice that they have an "Application Server" for Kubernetes, which is pretty curious: https://github.com/WolframResearch/WAS-Kubernetes (though not updated in over a year)

One of the more interesting things about WL is that Stephen Wolfram is really a genuine daily user of the software his company makes and he's the final say on what ships and in what form. They used to livestream his meetings reviewing potential new features on Youtube, an interesting watch. It didn't make me want to work there but I did feel like he cared very much. Quite Jobsian, dare I say.

Man, I miss Wolfram Language. Once you've twisted your brain a little to grok its usage, it's such an incredibly high-value tool, especially for exploration and prototyping. I saw it more as a do-anything software tool for researchers rather than as a language aimed at programmers, so I put on a researcher hat and tried to forget everything I knew as a professional programmer, and had a few memorable seasons with it around 2016-2020. I remember calculating precisely which days of the year would cause the sunlight to pass through a window and some glass blocks in an internal wall, creating a beautiful light show indoors. It only took a couple of minutes to get a nice animated visualisation and a calendar.

Nowadays I'd probably just ask Claude to figure it out for me, but pre LLMs, WL was the highest value tool for thought in my toolbox.

(Edit: and they actually offer perpetual licenses!)

It looks like they've now added a Sonnet cap which is the same as the previous cap:

> Nov 24, 2025 update:

> We've increased your limits and removed the Opus cap, so you can use Opus 4.5

> up to your overall limit. Sonnet now has its own limit—it's set to match your

> previous overall limit, so you can use just as much as before. We may continue

> to adjust limits as we learn how usage patterns evolve over time.

Quite interesting. From their messaging in the blog post and elsewhere, I think they're betting on Opus being significantly smarter in the sense of 'needs fewer tokens to do the same job', and thus cheaper. I'm curious how this will go.

On my Max plan, Opus 4.5 is now the default model! Until now I used Sonnet 4.5 exclusively and never used Opus, even for planning - I'm shocked that this is so cheap (for them) that it can be the default now. I'm curious what this will mean for the daily/weekly limits.

A short run at a small toy app makes me feel like Opus 4.5 is a bit slower than Sonnet 4.5 was, but that could also just be the day-one load it's presumably under. I don't think Sonnet was holding me back much, but it's far too early to tell.

Very interesting! The one killer issue that jumps to mind is anti-cheat. I switched away from gaming on Linux via Proton to gaming on Windows because Battlefield 6's anti-cheat won't work under Proton. Many games are like this, particularly some of the most popular (Rainbow 6 Siege for instance). And BF6 made this decision only recently despite the growing number of Steam Deck players (and other players on linux - in fairness I don't think there would have been that many BF6 players on a handheld).

Edit: I specifically use a gaming-only PC. The hardware is used for nothing else. Hence, discussions of rootkits don't really bother me personally much and on balance I'd really rather see fewer cheaters in my games. I think it would be the same with any of these machines - anything Steam-branded is likely to be a 99% gaming machine and their users will only care that their games work, not about the mechanisms of the anti-cheat software.

My apologies, it was the Scots Wikipedia, careless of me. Link for context: https://en.wikipedia.org/wiki/Scots_Wikipedia#Controversy

Some cursory clicking about didn't reveal to me the actual corpus they used, only that it is several trillion tokens 'divided across the languages'. I'm curious mainly because Irish (among some other similarly endangered languages on the list) typically has any large corpus come from legal/governmental texts that are required to be translated. There must surely be only a relatively tiny amount of colloquial Irish in the corpus. It be interesting to see some evals in each language particularly with native speakers.

I think LLMs may be on the whole very positive for endangered languages such as Irish, but before it becomes positive I think there's an amount of danger to be navigated (see Scots Gaelic wikipedia drama for example)

In any case I think this is a great initiative.

I have to wonder if the 'reward' of higher limits for setting the browser as the default is in conflict with the Digital Markets Act (although as I understand it, OpenAI isn't currently seen as a 'gatekeeper' under that act - but I would be surprised if that didn't change in the coming years).

Oh well, I guess I will have to get my self-esteem from a different number on the internet!

I mean, this post is an hour old and the other only 5 minutes. Who's duping who here :)

Downloadable for MacOS here: https://chatgpt.com/atlas

From the description:

> Introducing our new browser, ChatGPT Atlas.

> Sam Altman, Will Ellsworth, Adam Fry, Ben Goodger, Ryan O’Rouke, Justin Rushing, and Pranav Vishnu introduce ChatGPT Atlas — our new browser. Now available globally on macOS. Windows, iOS, and Android are coming soon.

Wow, this is really an incredible bit of work! Very impressive stuff, I look forward to using it!

I'm not sure we diverge in this sense, I also think that's a good thing, if not the most amazing thing about modern programming! The democratisation of hyper-personal computing is incredible. Now anyone(-ish) can build exactly(-ish) what they need to create whatever they want to create.

What I suppose I'm unsure about is the value of publishing utility code when the cost of bespoke creation is so low. But there are a couple of interesting perspectives in the responses so now I have to go away and think about it some more :)

Very good point. I only thought about it in terms of the individual effort rather than the actual physical energy being used.

That's a very fair point in regards of the cost. I should have thought about that more.

The analogy is very interesting, but I think perhaps I didn't do a good job of articulating my point - I feel I was trying to say that everyone*, myself included, is a Photoshop user now, and we know how much work it is or isn't to achieve some goal with it (which was previously impossible or difficult when we all just had paintbrushes and easels). I'm not doing a good job of articulating my point now either unfortunately.

Perhaps I'm trying to say that we all* own photocopiers now? I'm not sure.

I appreciate you taking an opposing view. I will think about my opinion some more.