Yeah - the whole Claude did it thing here reads to me like blaming a grammar issue on an LLM. Yes they have their issues but this isn't one of them - they're not deterministic they aren't going to make the exact same bug as a project you never heard of unless it's a specific technical shortcoming/training data gap/skill issue of the model. If you say "I want something exactly like xyz in every way" though then sure - but you're claiming you never asked for that? Not passing the sniff test here
Ai revenue or datacenter/compute revenue? There's a lot of circular financing right now and there's a pretty obvious bubble for sure but I haven't been able to figure out what exactly this guy's argument is after seeing it a few times recently. Like yeah most ai datacenter spend is those two companies - that's pretty standard monopoly (or monopsony for the cloud providers) dynamics. If you're saying some major percentage of all ai related spending is openai and anthropic spending money on compute - well that's not exactly right because Google etc are also spending money on building datacenters (that's not ai spend in this definition? WTF is ai revenue exactly?) - and that's just pretty much indicating it's a frothy market with two unprofitable companies at the center with huge cogs? We knew that already.
Feels like they want to make a clean headline grabbing argument about how "70% of all the spending is actually just these two companies" and are ending up with a really muddled headline that's just like yeah that's how monopoloies and duopolies work. When there's a lot more insidious circular complicated shenanigans going on that gets collapsed by this framing.
Tbh I appreciate knowing immediately that no effort was put into writing an article and that no effort is worth being put into reading it. I genuinely (hah) think it's good that different llms have distinct voices when writing. It doesn't bother me when using it for coding because ok that's just how Claude talks, and in the wild seeing folks present claudeslop as their own writing is a really nice really quick quality/effort signal. Adding another step to the prompt doesn't mean whoever generated this report actually out any more real effort into making sure this is any good
Shout-out to anthropic for having their models have such a strongly distinct writing style and personality that you can recognize their work instantly! It's quite nice to have such an immediate signal that if I were to proceed, I would spend orders of magnitude more time and effort reading the the text than the person claiming author credit spent writing or even reading it themselves.
There are some claudeisms here (it's not x, it's y) but it does read like they did at least one human pass on it. Or maybe this is opus 5 writing which is a bit less distinct idk. It doesn't feel as obviously generated slop as some of the stuff posted here with every other sentence a fragment and load bearing and all that nonsense
Liquid gasoline does not have anywhere near the energy of TNT, or even a battery for that matter. It contains basically zero releasable energy. It needs oxygen or another oxidizer to react with to actually release any energy. TNT and batteries' energy density calculations include the oxidizer and the oxidizer is in close proximity to the fuel (molecularly so in the case of TNT). If you 10x the "energy content" of gasoline it's still rate limited by access to oxygen. If you 10x the energy density of a battery (the type with the oxidizer contained within the battery, not a fuel cell or metal air battery) you've got 10x the energy ready to be released quickly if oxidizer and fuel mix in unfortunate ways
Jfc Claude's writing can be really intolerable sometimes. Trying to parse wtf this post is saying - I guess the author is trying to make a harness for local models, and tried to have codex/Claude code autonomously make the harness better, used some sort of vibe coded test suite for that goal that was broken, switched to a different benchmark later, and the conclusion is that occasionally you can use an AI to over fit a harness to a benchmark?
It's a good quote from Feynman. I think you'll find if you measure your paper against that honestly, you'll see where it continues to fall short, as far less falls out than goes in. All the counterexamples find themselves being dismissed by vague speculation, often not supported by further research, and that impulse towards dismissing things that don't nearly fit the theory means the theory is restricted from finding a path towards generalizing and capturing deeper truths - you end up with epicycles instead of universal equations
More directly the stuff about ranking different cultures on a scale of musical sophistication that ends in western harmonic music and dismissing obvious counterexamples as cultures that just haven't gotten there yet or just use other feature as ornamentation or just happened to not stumble on things because they were at the lowest level of musical sophistication is almost comically ethnocentric. Its an easy hammer to reach for when your counterarguments are not European (though in this case you're having to ignore an awful lot of European music and the huge influence of Greek music on the middle East) but yeesh
Awfully dismissive of non western tonal harmonic systems for a universal theory of music? The whole idea that musical traditions that use less harmonic structures just haven't unlocked all the structures available is kind of absurd - those musical traditions have been developing for the same amount of time - and there's a pretty huge blind spot for other traditions. The whole microtones just being ornamentations - no that's absolutely not the case. Trying to ignore/hand wave away all the counterexamples and force them into your system is how you get epicycles. If you're gonna do a universal theory of music it would be a good idea to engage deeply with more than one system. As a universal theory of harmonica maybe that's fine but it's sloppy and motivated reasoning
I would be stunned if there was a single line of hand written code in the entirety of Claude code lmao. Or if more than 20% of the code had been read by a person at any point. Why would you expect your ai code harness to not dogfood?
No what I'm saying is that this is a case where one person already does have a hell of a lot of influence over every downstream project already and that has consequences
Yes? Dynamic workflows are incredibly powerful - especially for massive rote migrations. They're also particularly nice when you want to stretch an expensive model like fable and you can have it orchestrate a bunch of sonnet agents or whatever and get decent results without having every token burned be a fable token. For my usage style I find the best thing to be about dynamic workflows to be a sort of context management win. The orchestrator agent, if it was implementing everything, would end up deep into its context window where performance degrades steeply, and if you ran into issues early on the session you'd have to be concerned about whether whatever reasoning traces are in context are poisoning your later outputs, while if you're using subagents once you steer the model whatever is doing implementation is starting from fresh context with a new prompt so you aren't forced to kill and restart your session all the time. The way workflows are implemented - scripting with clearly defined stages also mean that the workflow will terminate eventually instead of running forever.
Now with that said, opensi's ultra mode is absolute trash - they should just have stolen Claude code's implementation - and there are clearly modes like max reasoning where they'll burn double the tokens to get another 10th of a percent performance to win benchmarks, which you should basically never be using. They're not making those versions at the expense of more efficient reasoning levels though and I don't mind that they exist (as long as they don't get made the default mode - openai - fix your ultra mode already).
Then why are fable form anthropic and the 5.6 sol line from openai so much more token efficient than other models? I really don't get this take at all - they're supply constrained right now, and there's literally no economic incentive to make each response take more tokens for the same output when we're in a market as intensely defined by induced demand/jevons paradox as this one. People are hitting their limits. If they make each turn take less tokens and each session take less turns, people will make more sessions.
Unfortunately the one thing they seem to keep not letting their competitors our play them at is coming out with really damn good models. Not consistently mind you - openai had a really solid lead for a few months earlier this year when it looked like they were running away with it, but then we get fable 5 and opus 5 and people will put up with a lot to get that model quality
IMO firm wording against non generative uses of AI or lightweight uses like tab complete will go beyond just influencing good faith actors to reduce llm usage - it'll just influence them to not contribute instead. If you make it clear you don't even want llms used for debugging, if you're a good faith actor that makes use of even the most minimal extent of these very common and basic tools, you'd rather just not deal with the hassle. Either you don't deal with debian at all or you don't try and upstream your changes. The latter is probably the desired outcome to limit slop contributions, but when you have the Linus Torvalds making clear that llms are a tool and that he thinks banning their use is absurd when it comes to the kernel itself, youre gonna cut down on a lot more than slop.
The wording of "written with ai tools" implies generation, as do all the concerns about copyright etc. it would be beyond insane to attempt to ban asking a chat bot or an agent questions to navigate or understand a codebase, not in the least because that is completely unenforceable
Yeah there's a brand premium. Like literally with apples the ones with trademarked names can cost more. And the ones with trademarked names have an organization behind them that promote that apple variety and set fees etc for growing and selling those apples. And customers are willing to pay more because those apples usually taste way better (the group exists to stop growers from enshittifying the apple by selecting for yield over flavor like what happened to honey crisp). Those prices are being "manipulated" but that's not criminal behavior - it's not illegal and wouldn't make sense to try and make illegal. The frontier models also do have different coats and have "sales" (for personal plans, the amount of usage you can get on the 200 dollar plans is orders of magnitude more than you could get for a similar cost for any open model - you'd need the same capability at the same token efficiency at literally 1/40th the cost to be able to be cheaper) - to the extent that if there is illegal stuff going on it seems more likely to me that it's on the category of dumping/pricing unreasonably low to kill competitors in some anti-competitive way (though as I understand it it's not something courts tend to find as illegal) instead of price fixing
The idea that humans have undergone high selective pressure for genes related to immunity in some way during a recent 5000 year period in which massive urbanization lead to a truly unfathomable amount of deaths including staggering percentages of humans dying in childhood because of various infectious diseases, plagues, pandemics, etc, and that those changes involved large numbers of small changes across a large number of genes in complex regulatory networks, is probably the least surprising discovery in the history of the universe. It's good science and it's obviously important research to do, not in the least because it can shed light on evolutionary pressures that may be driving the rise in autoimmune diseases we see in the modern day, but the framing of the PR slop article around it is truly so stupid.
Maybe if it found selective pressures were still dominating and driving evolution in recent decades in developed nations with low childhood mortality that would be surprising. Though the framing of the article is almost that it would be surprising if mutations were happening at all which is obviously stupid - we haven't figured out how to reverse entropy so mutation rates haven't gone down, evolution by genetic drift is obviously continuing to occur, and with advances in medicine the number of folks surviving with mutations that would have been problematic otherwise is increasing.
Truly disappointing slop. Like obviously humans faced selective pressures during an era when like 8 in 10 kids died of plague or similar things, a biological factor where survival is correlated with genetic factors that influence immunity and disease survival.
Again, super useful science to see exactly what that influence was, but it would only be surprising if we learned we'd somehow managed to turn off entropy and statistics during the black death and we didn't face selective pressures.
Honestly the models are rled so hard on specific synthetic datasets and specific behaviors/personalities that I would be concerned that trying to change its behavior like this would hurt output quality. It's a tool, I don't care what garbage it generates or what it sounds like as long as it can do what I need it to do, and I don't get what I'm going to gain by having it burn reasoning tokens on word smithing it's responses to not "sound human" instead of on writing tests and reviewing code
The competition is between openai and anthropic, if there are price agreements between them that's absolutely price fixing. Or if there's collusion between the cloud providers to inflate compute. I would expect Amazon and gcp to both pay about the same in license fees to anthropic for their models though because they're paying for the same thing. If I buy an apple for a dollar at one store and an apple for a dollar at another store - maybe there's price fixing, or maybe that's just the cost of apples at the moment.
Did they actually ever cut the price on gpt 4? The oldest versions of it in the api still seem stupidly expensive? There were definitely price cuts as they introduced the turbo models and stuff, and new versions of each model might have gotten pricey cuts, but just because they're both called "gpt-4 something" doesn't mean they're the same under the hood or that they didn't change a bunch of stuff under the hood to make it cheaper to serve
I doubt there's any sort of criminal behavior there - the model is anthropic's up and anthropic probably charges a very expensive license fee that's the same for all of them, and their cogs on compute aren't going to be wildly different, so the main drivers of the cost are roughly the same and they're all offering customers the same end product so the prices would likely also be similar in the end
I doubt it's more profitable - ai overview is free/runs even when signed out in incognito, while the frontier models from openai and anthropic are nauseatingly expensive, especially for enterprise, and have a ton of users who are willing to cough up that money. Even if they aren't profitable because their cogs is even higher, they certainly have revenue with ai overview doesn't really have
I think they've been behind for a while - flash isn't even that competitive with glm 5.2 and from their hype around 3.5 flash at launch - that was certainly not intended to be the case
The tokens served number might include cache tokens which are a huge chunk of agentic token spend - and even with that the estimated burn rate for anthropic doesn't seem wildly off? They spend 1.25 bn per month on their deal with SpaceX alone.
They definitely cache results - I've searched and re-searched an identical query back to back a few times and seen identical results from overview. They are definitely throwing a stupid amount of compute towards these ai results nobody is paying for - changing punctuation and stuff does get you a different response - but they're not doing no caching.
Certainly what they're doing with their infrastructure is impressive but it's not super meaningful at the end of the day for a for profit company to be really impressively good at burning tens of billion dollars on a service nobody pays for while the same tech from their competitors is quickly becoming one of the largest spend categories for many software engineering teams
Whether or not the specific policy is good my preference is that changes to policy that have been in force for decades happen based on legislation and not the whims of 9 unelected people. We didn't get clear rules made my legislature, we lost an escape valve that allowed our regulatory apparatus to function while the gerontocracy on capital hill spun their wheels and left everything even murkier than it was before.