The people who own the content you decide to visit opted into it though, and they're the ones paying for it, right? No one accidentally sets up a CDN. I'm struggling to imagine workable economics where we have the rich services that we have today on the web but individual users get to dictate to sites they visit what their tech stack needs to be in real time. Imagine visiting your banks website and they're using Cloudflare, and you decide nope, I want the data delivered by CloudFront. Then the next person 20ms after you decides they don't like either of those, they really want the world to burn so they demand Azure.
That at least would bring back local bank branches at least. And movie rentals by mail.
If you've believed all the government data for the last 10 or 15 years I got a bridge for sale. It's one of those things like, yes, it's WAY more difficult than I think most people realize to arrive at numbers (be it employment, price indexes, etc), but it's also something that demonstrably can be done. It's like "space is hard", yes, but rocket companies still rightfully get flak if they can't perform.
Also, for what it's worth, it's a global phenomenon. China's numbers have always been almost wholly manufactured, but to be fair to them that was intentional. Across the West there's been a slow eroding of quality that they didn't intend necessarily, but has just been the effect of focusing on anything beyond pure merit in hiring and accuracy in measurement. Focus softened at exactly the time COVID made things more difficult, and the focus has only got worse. Ex: https://www.telegraph.co.uk/politics/2026/06/30/treasury-dit... Yeah that's going to work out great.
That said, I think GAO and CBO both do decent jobs at extremely thankless tasks. They both produce research and recommendations that are typically sound but get ignored by Congress 99.99% of the time.
Cost per task, GLM (due to its poor token efficiency) still lags. GPT 5.5 on low seems to wipe the floor with it. All the Chinese models still like to use tons of tokens. It shows up in time per task as well.
But, and this is the problem corporations face, that's the truth TODAY. They need to make decisions for the next quarter at minimum, for a year ideally. I don't envy IT leadership in this environment, there's no right answer along the lines of "no one ever got fired for buying IBM". Any choice today may potentially make you look silly in 90 days.
Now that enterprise customers are pay-as-you-go with tokens I suspect we'll see renewed interest in OpenAI and their focus on token efficiency. At least I hope so if the alternative is abandoning the tools entirely.
That site is useless though because thinking tokens (and caching) and the efficiency thereof aren't accounted for. GLM5.2 is promoted by every 50 Cent Party the PLA can muster on the internet but it falls short because of its extremely verbose thinking. Anthropic models have the same problem but starting from a much higher base of real intelligence.
Which is exactly why every credible comparison now represents cost associated with completing a task, not arbitrary input and output token costs.
I'm sure it also takes more compute effort to be at the frontier, rather than being able to distill and poach ideas from the frontier. No mistake that it's the same handful of labs taking turns at or near the frontier.
Right? For all the money paid by modern students is it unreasonable to expect some expert human interaction? The cost explosion has been centered outside the classroom, elsewhere in the org, so can't play that card.
Too overbearing though and you get... Mistral? A continent that hasn't been on the leading edge of anything (other that expansion of the regulatory state) for decades, and Europe feels it in their employment numbers.
Current French (8.2%), Spanish (10.3%) or even Swedish (8.6%) unemployment would count as a disastrous recession in the US. In the US we call 2007-09 the "Great Recession", which peaked at 10.0%, and that relatively brief time left a generational mark. That's a somewhat routine number by EU standards.
Not to mention you end up with bizarre effects. If the UK were admitted as the 51st state it'd immediately be the poorest. (Yes, some EU countries are wealthy, but they're also the size of US counties, if we cherry pick just Manhattan we could make some spectacular comparisons too)
So, it's a complex issue but the tradeoffs are absolutely tangible yet often dismissed.
Yeah, that's what we need, frontier intelligence models open in the wild that, if a jailbreak is reliably established, there's no possibility for anyone to ever patch at the API layer. Because there is no API layer.
Is it supposed to be? I think the point with some of these Macs is you get the capability in something the size of a heatsink from Intel's Netburst architecture era, or a Macbook light enough to stick in a backpack and take with you to lunch.
If you're talking about chaining together multiple GPUs you're talking about a different game -- I suspect, anyway. Seems like a high-spec Mac would be good for development and testing. Arrays of GPUs, better aimed at production use.
Based on DeepSWE, Opus 4.8 gets you more intelligent output at lower price (GLM's token inefficiency is really biting them). GPT5.5 even moreso. And I don't recall about Opus but GPT is much, much faster at getting you the answer (again, GLM's token inefficiency).
It's neat, I guess, that we can compare them against models released last year, but I care about my options today, and the pareto frontier is about as far away as it ever was.
Add on top of that the extra features OpenAI and Anthropic have in their apps and...
What's that got to do with the cost of a thing? Are tradesmen in Thailand entitled to Makita tools just because American plumbers can afford them? I'm struggling to understand the entitlement in some of the comments. And even though it doesn't matter I'd point out it's not like OpenAI or Anthropic are making enormous profits at the moment.
Eaaaaasy now, the Chinese labs aren't freedom fighters on behalf the common man. They're not non-profits, they're not neutral transnational organizations only dedicated to open source efforts.
They're Chinese companies offering open source models now as loss leaders to keep themselves in the game because they know virtually nobody, especially in the corporate world, would contract with them and give them access to their data. They might as well just send a Dropbox link of all their sensitive data directly to their Chinese competitors, same end effect.
They're also doing it as the digital equivalent of what they've done in other industrial sectors for decades. Undercut and flood the market and once you've killed or severely hindered your competition, then you have the market cornered. The moment they can afford to these open source releases will stop.
Then the world will be stuck, just the way the world is largely stuck on rare earths. Instead of being able to regulate the leading companies from DC and Brussels, they'll be stuck watching Beijing call the shots.
That world would likely always have guys like Mistral and Trinity, but it's an open question if they'll ever catch up to the frontier.
And then Beijing will enjoy access to the data (ask any multinational operating in China for more than 2 seconds how useful contracts and Chinas legal system is for protecting IP), and these companies will roll in the money, and the Chinese supply chain will grow up behind the labs.
So, let's not pretend they've got the moral high ground. No. That boot just isn't on your neck yet. They're playing the long game -- and they're good at it.
People seem to think it's all smoke and mirrors. IDK. My employer, in an industry as far removed from Silicon Valley as you can probably get, makes more and better use of it all the time. There's enormous amounts of work done every day in corporate America that amounts to "I need X, but X involves some data from legacy system Y and legacy system Z, and that's going to take me an hour to glue together because our entire enterprise runs on a system cobbled together over the past 50 years". You know what Codex/Cowork/etc can do really stunningly good these days? Take the files you provide, listen to what you want from them, ask clarifying questions as necessary, then write a script that programmatically does exactly what you ask for, checks its own work and then gives you the result.
We also realized that a project we thought was going to involve thousands of man-hours of very expensive, senior draftsperson labor to backport a feature into decades of CAD files we could, after years of procrastination, just forget! AI has got to the point where it can see and count what we need for us with greater accuracy, in our tests, than our very best humans, so we can just make the change going forward and let AI read the old files as needed.
There's incremental commercial adoption of that nature that I'm sure is happening slowly across the corporate spectrum, and that kind of thing is durable demand, not a bubble.
Note though that I'm talking about real revenue, etc. Not stock market bubbles. The feds been inflating that junk since the housing bust. We'll all pay for that eventually.
They could (and might as well) give Windows away for free. It's not an operating system for them, it's a sales funnel.
But that sales funnel needs to not exceed a certain maximum level of annoyance to the people traveling down it, or they'll look askance at other options.
I mean, not even all the Copilot silliness has been able to lead to The Year of the Linux Desktop, but Windows 11 as a whole I suspect has been good for Mac and iPad sales. And those are people that now aren't getting nagged by their operating systems to use OneNote, OneDrive, Word, Copilot etc at every turn.
Do you have HVAC? Maybe some of my difficultly understanding is being American I'm in a hotter, much more humid climate, so we've got central HVAC. A key feature of heat pumps isn't just that it lowers the temperature of the air, it also reduces humidity.
I've got a lot of exposure to new home construction here and can tell you I don't even know what "damp proofing" is, and our bathrooms don't need special paint. They're ventilated and we have HVAC. Beyond that, if homeowners take 30 minute showers with scalding hot water and the door closed then, well, the outcome is inevitable no matter what you do. Not just mold but you'll start damaging fixtures, etc.
It's hard to get old caulking clean and keep it mold free, just gotta recaulk regularly, but I'm somewhat skeptical of blaming paint.
This is a fairly well-trod path in economic policy circles, especially in Europe. You can either grandfather in buildings, perhaps with rules that line up with maintenance schedules anyway so that when something breaks anyway you replace it with the new standard (the HVAC world understands this well with refrigerants), or you sign yourself up for stunning, astronomical expense.
Not to mention, a lot of places around the world care about the look and character of historical locations. If a structure wasn't designed for central HVAC, for example, then there's often nowhere to hide the condenser units, air handle units or the ductwork. Same with insulation -- if that exterior wall wasn't intended to have it you've got a couple options and they both hurt.
Last of all, I'll mention labor. The type of skilled labor that can do any specific trade at all is relatively rare in the aftermath of the "college-or-bust" era, but the kind of labor you'd want for renovation work (fast, efficient, can tackle multiple different aspects at once without calling in different trades, and gets it right the first time to minimize disruption/call backs) is even more rare. To carry out some kind of massive renovation project at a national level even with infinite money you're talking about a generational timeframe just due to labor constraints.
If that ends up being true, GPT5.5 at 70 (and presumably Fable a bit ahead of that) is still in a different league, which was partly my point. To listen to online chatter, GLM5.2 is a tectonic shift in the landscape. In reality, it's just interesting. Probably safe to bet once the DeepSWE benches all get fully updated it won't even be on the pareto frontier.
I'm not accusing anyone specifically, but I've noticed Chinese bots swamping certain YouTube channels that, for example, cover US defense industry news. They'll downplay any and all technical advances, play up China's dominance, US cowardice, etc. All very transparent. I suspect some of the online conversation about open Chinese models is driven by that. How often do you see people talking about Mistral or Trinity? Never. Because they don't play that game.
Don't sleep on JPEG XL. It's used under the hood within DNG files (at least, it's an option, Adobe DNG Converter can leverage it, including by the CLI), DxO PureRAW leverages it in the latest versions. Apple Photos can view them, and I think it's been the compression methodology used inside their ProRAW DNGs for a while (which probably by default makes it one of the worlds most popular image compressors for RAW files). I've had a lot of success using it for various things. Had some issues surrounding metadata but that may be user error on my part.
That at least would bring back local bank branches at least. And movie rentals by mail.