This comment always comes up but isn’t this literally irreducibly complex in the sense of Kolmogorov complexity?
I’m out of my depth here but it does seem prompts have a “smell” to their complexity that means we can use a classifier to mate them to a model with better success than random noise, but still don’t essentially all of the savings associated with systems like this just come because there is some “complexity arbitrage” where there is enough of a token cost differential between models that you always save money by trying the cheap model first and then trying a hard model if it turns out to be hard?
Everyone seems quite upset/perplexed by the Gemini naming scheme but I quite like it. And hey, if you find yourself lost, the periodic table is probably the most ubiquitous ordered list in the world and serves to clear things up pretty fast. This will likely not be strictly true as models release over time, but is a good rough indication of what the team thinks we should expect of the models, in the same way the other labs run their naming.
I also like the personalities implied by these, with Anthropic saying “human literary achievement is neat”, OpenAI saying “space is neat” and Google being like “science is neat”.
Edit: I was coming back to say “I wish we could talk about stuff more substantive than the naming scheme” but (1) my comment may be objectively incorrect, and (2) I have to apologize that I haven’t helped the situation by… commenting on the naming scheme.
They're in alphabetical order, almost like Ubuntu release names. There was a leak about a Gemini Oxygen model, which would further support alphabetical order -- skipping names they don't want to use.
So, names don't align with capability but instead "release freshness". This is unlike every other provider but it is what Google does (did?) with Android releases I believe (dessert names).
> And hey, if you find yourself lost, the periodic table is probably the most ubiquitous ordered list in the world and serves to clear things up pretty fast.
But the way google ranks the models by capability doesn't follow periodic table order.
Hydrogen - 1
Carbon - 6
Neon - 10
Argon - 18
It could very well be based on their density (g/cm^3 in standard conditions):
Hydrogen - 0.0000899
Neon - 0.000900
Argon - 0.001784
Carbon - 2.267
This would make more sense than assuming it is based on their periodic table order as it more accurately maps to how Google markets their capabilities, but is too esoteric to really be useful for clearing up confusion for normal users, IMO.
The periodic table isn't really helping. Hydrogen is the lightest, that much makes sense. Beyond that it doesn't seem to work. The elemental ordering is H C Ne Ar. The cost ordering is H Ne Ar C (with Ar and C being the same per token but I gather that C thinks more and thus costs more). The capability ordering seems to be H Ar C Ne. Argon is the highest on the table but is neither the most expensive nor the smartest.
What will actually "hurt retirements" (delay them) is if this bubble bursts like the dotcom one did. Which seems strictly less likely without the most valuable company in the world relentlessly increasing leverage on its single investment thesis.
I think so but it may be a very short term effect, I think their pace has roughly matched the pre-speed up pace of the frontier labs (the 1.7-2x speed up noted on by METR), and I’d guess we see a similar speed up from the Chinese labs once their internal models hit an equivalent level enabling the same speed up.
Following that logic the gap widens a little bit here but then will steady again. Before this there has been a general catch-up trend too which probably also persists leaving us in a slow closing of the gap as the net effect?
If the speed up is the reason, then they will never catch up (with traditional means). Because it means RSI predictions are real, and we will see an explosive growth in capabilities (far beyond exponential).
If compute is the bottleneck we’ll encounter in the coming year, then you would expect the gap to increase even more. The US and its companies has three times as much AI compute as the rest of the world combined.
I’m actually expecting it to expand. There is so much new production being built right now by every company due to the shortage, and China doesn’t have a single EUV production, and only expect a prototype one in two years.
The past few days I’ve been finding myself asking “seriously, where is it”… feeling that the world has somehow changed because Qwen 4 27B was later than I expected.
My qualitative take to support this is that Gemini seems to have the strongest spatial reasoning that translates nicely into CAD and engineering domain challenges. Less clear, but I also suspect a better world understanding of materials and their interactions in an intuitive way. I have this sense that Claude can tell me everything there is to know about how to spec a gasket, but it’s like Gemini actually knows what a gasket feels like and can intuit what it’s going to do.
My partner runs an internal computational mechanics benchmark that hasn’t tracked progress with model advancements (flat/no progress since early 2026) until suddenly saturating from argon and carbon. My take is that it’s due to spatial reasoning (most of the challenges are programming that require a lot of shape rotation type thinking), but we don’t know just yet why Gemini is doing so much better.
4.5 hours for any terrestrial based satcom solution is insane. At this point, the battery tech alone with these power requirements is stuff that Garmin, Viasat and Honeywell could only dream of obtaining in the same quantities.
Distance alone isn’t the real issue (newer Pixel phones can talk to GEO, which is 100 times farther out than Starlink in LEO!), data rate over distance is.
It's not inherently because of space, it's because phased array antennas are just very power hungry. Most of the link budget is coming from the satellites huge antenna, the mini has a comparatively tiny aperture
wow this would be a neat twist if true - we all defaulted to the negative explanation of having killed them all where a much more positive explanation might be that we're no longer luring them to their deaths at the leading edge of automobiles?
I guess it is most likely some combination of the above, but if there's any evidence it would be a neat reframing
I am not making any claims about the relative prevalence of insects day vs. night.
I can say with absolute and total confidence that the phenomenon of insect splatter was a common - if not frequent - daytime experience. I've not experienced this phenomenon in decades.
It’s effectively crowdsourcing urban planning, I’m in support.
I honestly think if we invited this type of empowerment/interaction in the physical world a bit more it would probably be good for spurring creativity in society (I think we do a pretty good job of it in software). It’s just a nice reminder that everything is just made up by someone else, likely no smarter than you.
Also I can’t believe this was 10 years ago, I remember being super involved in SPUR in SF (and getting coffee there almost every day) and going to talks about this kind of thing
It's admittedly appealing because it feels far closer to ordinary citizens having an influence, but I can't help feel like this is a failure on the part of the professionals whose responsibility the planning and implementation of this is.
Have we really gotten to the point where we can't trust public infrastructure to be properly implemented, and so the only solution is to take it into our own hands? What are the tax dollars paying for then?
I think in many places we have gotten to the point where the planning, processes, and public approval process for some of these simple projects is more costly in time and money than just building it as a temporary experiment and revisiting it in a year or two to measure the outcome.
Not that there shouldn't still be some kind of standards and community approval, but I think we can afford to set that at a lower bar for certain categories of easily reversible projects.
And hopefully incorporating more of these ideas can teach us to be more efficient on transit infrastructure projects in general, where the US often significantly trails other countries on cost and implementation time.
Changes are stymied by bureaucracy put in place due to fear of litigation. The primary reason that things are studied, studied again, given community input, and checked for environmental impact is that if a city or municipality gets sued due to a road change, the onus is on them to demonstrate that they didn't create an unsafe condition. For example in cases where cars crash and are liable for property damage, car owners try to sue the city to claim that the area was unsafe. Pointing to a massive paper trail like this removes liability from a set of city or municipal concerns.
A municipal or state government will not be held liable for the consequences of shitty infrastructure unless there's like 9/11 levels of "the system was blinking red and you ignored it" and even then it's coin toss at best. Like maybe if they kill a ton of people there might be a lawsuit that sticks that's sooooo unlikely for the scope of the kind of thing we're talking about here.
The bureaucracy is put in place because none of the people putting it in place are paying for it, headcount is power, control over resources is power, etc, etc, and all the other perverse incentives you see in government workplaces. And of course once the gravy train is running all these parties have an incentive to keep it running.
I mean, this is the "bullshit" that our DOT told us when talking about guerilla urbanism at a real city meeting. I agree that it's frustrating, but it is what we were told. I don't know how to "disprove" a defensive posture. And yes, our city goverment has been sued in cases on the grounds of poor street design (though they've only had to pay out once in 20 years.)
> The bureaucracy is put in place because none of the people putting it in place are paying for it, headcount is power, control over resources is power, etc, etc, and all the other perverse incentives you see in government workplaces. And of course once the gravy train is running all these parties have an incentive to keep it running.
Do you have any real experience in city or local government?
>. I don't know how to "disprove" a defensive posture. And yes, our city government has been sued in cases on the grounds of poor street design (though they've only had to pay out once in 20 years.)
You ask them how much money they'd pay out to lawsuits, how much money they were paying out before adopting those policies. You ask them how much these policies cost (I bet they haven't even calculated it because they'd rather not have that number exist). Etc. Etc You make them spell out the math for you.
>And yes, our city goverment has been sued in cases on the grounds of poor street design (though they've only had to pay out once in 20 years.)
Suing the .gov is often the only way to get them to shape up and do the thing they should have in the first place. Getting sued doesn't mean much. It's all about the results.
The fear might be real in the minds of the decision makers but that doesn't mean it's well founded or the policies that arise out of it are not wasteful.
>Do you have any real experience in city or local government?
> The fear might be real in the minds of the decision makers but that doesn't mean it's well founded or the policies that arise out of it are not wasteful.
I mean I 100% agree with this by the way. We tried to tell the DOT to look at its own stats. The one lawsuit they actually paid out in was from street design before the formation of the DOT itself and the fact that they fell behind on the paving schedule. If you actually amortize the payout over 30 years (I think the city has had one of these roughly every 30 years) it ends up being some minuscule part of the budget. But they still shut us down.
We kind of have reached that point. You can look at the cost overruns and public consultation before every new project as failures of the system to deliver what people want.
I’m out of my depth here but it does seem prompts have a “smell” to their complexity that means we can use a classifier to mate them to a model with better success than random noise, but still don’t essentially all of the savings associated with systems like this just come because there is some “complexity arbitrage” where there is enough of a token cost differential between models that you always save money by trying the cheap model first and then trying a hard model if it turns out to be hard?
reply