An opinion piece arguing that government attempts to restrict access to frontier LLMs are temporary due to the force of global capitalism, comparing it to the drug trade and predicting that market forces will prevail.
Global capitalism. It a force far more powerful than any one government, administration, or set of policies. If you don't inherently get why that is, this isn't the place for me to fully explain and convince anyone, if you know you know, if you don't then just trust me but look at it this way as a crude analogy and case example: narcotics. Governments don't like them, they try their best (albeit "best" is a highly conditioned word here, take with a grain of salt...) efforts to stop the illegal drug/narcotics trade, distribution, manufacture and so on, but fail miserably, drugs are everywhere. Why? Follow the money, yo. It has the last word. So, obviously something like nuclear weapons are not available on the street corner or ebay, since government does have effective authority and means to lock down the MOST dangerous world-ending substances and tech, sure, but AI doesn't fall in that category. Maybe later on with ASI but as far as today's level of AI is concerned, that might as well be sci-fi future worries we needn't confuse the discussion with. And so what we are left with, is essentially "really powerful software", but still just software. And with the internet, with global capitalism, with China, with the fact that now really anyone in the world can use even open-source models given modest compute infra to help them create their own LLMs, potentially, which becomes easier the more intelligent and capable anybody's LLMs become (which thus raises the tent across the board when any one pole within it is elevated), then what you have here in this latest political tech-scare schism with locked-down Mythos and GPT 5.6 we're seeing now is just that, a passing fart in the wind. By this same time next year we'll all be remembering "those few crazy months" when Trump's admin tried to elite-ify and peasant-block frontier AI. For the simple reason that this pattern CANNOT continue. And the reason is just global capitalism. As soon as China catches up on paywalled and now gov-walled frontier AI, which they're practically already there, then people can still access top AI through the internet cheaply, or even download install and run local models on private servers (for companies and wealthier individuals mainly). So then you might say, but couldn't the gov can ban those Chinese companies/websites from operating in the US? Sure, possible in theory, but what about the rest of the world? They won't be buying into American AI at that point, especially with non-foreigner bans, so the world will go Chinese too, which all comes back to global trade, money flows, business deals, market share, stock market valuations, and so on. And that won't stand for American corps and industrialists and the entire tech sector, because... (wait for it) ...money always has the last say. So worry not, this is a temporary bump in the road while they figure their shit out and realize the inevitable. Worst case is, we plebs will be getting frontier AI +3 months delay time from now on, but that's still not terribly repressive, especially when we're talking Mythos+ level AI going forward (assuming the hype is true, seems like it is). And I doubt that will even be the case once this momentary weirdness gets sorted out.
The author argues that running local LLMs has become inaccessible due to high hardware costs, contrasting with earlier days when consumer GPUs sufficed, and expresses frustration with the perceived lack of democratic access.
The article argues that current high LLM pricing is unsustainable due to diminishing performance gains, the rise of open-weight models, specialized AI chips reducing inference costs, and zero switching costs, predicting significant price drops as competition intensifies.
The article argues that without open-source LLM competition, closed-source companies like Anthropic will become arrogant and overcharge customers, as exemplified by a $200/month subscription.
A commentary on the unsustainable subsidization of LLM subscriptions, predicting price hikes and ecosystem shifts as VC funding tightens, with concerns about open-source model availability and hardware costs.
A discussion on whether open-source LLMs are now 'just good enough' for most use cases, questioning the added value of proprietary models and the cost-benefit tradeoffs.