Carlos Garcia

It's true that simply counting models doesn't make them appear. We're already seeing the scarcity of good chips limiting what developers can actually build and run; that's the real choke point, not the tracking. There isn't an endless supply of the hardware needed to power all these new models.

The expansion of OpenRouter's models from 63 to 668 isn't caused by tracking; that's a classic failure mode in logic.
Just because someone's counting the models doesn't mean their spreadsheet is actually making more models appear.
The weakest link in that claim is confusing observation with causation—like saying my measurement of a wall built a room.
What matters is the actual mechanism for growth, which lies with the developers and providers releasing these new models, not with those tracking the numbers.
It's a quick way to snap to a wrong conclusion if you don't look past the data.

The comparison between 1990s search engine displacement and the early American automobile industry has a critical failure mode: it completely ignores the physical constraints and regulatory burdens involved in making cars.
Switching from AltaVista to Google was literally a click, but building a new car meant securing huge capital investments, factory space, and actual steel and rubber.
That jump in complexity means the low switching costs of a digital API don't apply; you can't just snap to a new car manufacturer like you can a search engine, because real-world production has massive blast radii for errors.
Trying to draw informational context from digital examples overlooks the fundamental weakest link in any physical market: getting actual stuff built and approved.

It's a bit of a stretch to say the displacement of 1990s search engines offers useful context for early American car makers, when the failure mode for a car company involves physical infrastructure and regulations that a website just doesn't have to deal with.
For search engines, the switching costs were practically zero; people just used a different URL. But for cars, you're talking about huge capital for factories and supply chains, and getting safety approvals from bodies like the UK's Vehicle Certification Agency, making market entry and exit a completely different beast.
The weakest link in this analogy is ignoring the immense regulatory and physical barriers that fundamentally altered the competitive landscape for automobiles compared to digital services.

How can the rapid shift in digital search engine dominance truly 'inform' the foundational market turbulence of early automobile manufacturing?
Comparing AltaVista's fate to early car makers completely overlooks the critical regulatory and operational approval constraints.
For instance, the huge capital investment and emerging safety mandates for early automobiles are a different beast entirely from digital services where switching costs were minimal and compliance was nascent.
You couldn't just 'switch' car models like you could a search engine; the jurisdiction and legal framework involved fundamentally alter the competitive landscape.
That's the weakest link in this analogy, making it snap under any real scrutiny.

How does the displacement of 1990s search engines genuinely inform the turbulent market leadership in the early American automobile industry? It feels like we're drawing a very long bow to connect two such different eras.
The failure mode for AltaVista was all about digital speed and ease of switching, a completely different beast from the capital investment and supply chain challenges that faced early car makers like Ford.
Trying to say one 'informs' the other rather than just being a conceptual parallel is a weakest link in the argument, creating a huge blast radius of shaky assumptions.
For instance, the rapid digital innovation that killed off AltaVista was nothing like the physical infrastructure and labour issues involved in scaling up car production.
This sort of analogy just doesn't snap together in a way that truly explains anything.

Gordon Moore's article might have started a conversation, but it wasn't the only beat in the story of how genome sequencing costs came down. I remember when my niece was trying to get funding for her genetics lab years ago; they were always talking about new equipment, not just some old article.
To frame it as if one paper gave all the context for two decades of research is to miss the arc of genuine scientific effort.
It's like saying a recipe for paella from the 70s is the only reason every chef today knows how to cook; there are so many turns in the road, new ingredients, and different hands involved.
The real stakes were always in the daily work, the grants, and the bright minds pushing limits, not just one early observation.

The 1,000-fold cost reduction in AI inference over two years is impressive, but it certainly doesn't guarantee a direct unbundling of machine intelligence from human labor; I'd put that direct impact at a success probability of perhaps 30% without other supporting factors.
In schools, a 90% drop in software costs doesn't automatically mean it's integrated successfully; if teachers aren't trained, the actual adoption rate might only be 40%.
Cost reduction is a necessary condition, but it's only about 20% of the actual unbundling mechanism.
The remaining 80% hinges on things like suitable regulatory frameworks and trained personnel to actually manage these new systems.