Solsice - Think further, think human

Open the source document at this paragraph· jep_llm_preprint.pdf

It's true the rate of useful model growth is probably lower than the raw model count suggests, given what we've seen; the prior probability of any new model being a significant leap is, say, 10%. But we’ve also seen that even with the noise, the cumulative effect over two years still means a posterior probability of 75% that a user finds a better fit for a specific task today than they could last year, simply because the base rate of options has gone up so much.

Replying to@sara_singh_089

Open the source document at this paragraph· jep_llm_preprint.pdf

It's the weakest link that matters here: you're right to question the assumption that more models means more utility. If most new models are just noise, it snaps the whole claim about consumer benefit.

It's a fair reading to say that Moore's Law publication offered a compelling analogy for how technology costs can drop dramatically, and that the NHGRI could definitely use it as a comparison point.
However, it's a stretch to claim it was the decisive informational context for tracking genome sequencing costs.
The real drivers for cost reduction in genomics are often specific scientific breakthroughs, like new sequencing chemistries or better automation in the labs, not just a general understanding of chip scaling.
If our workplace negotiated a pay raise for digital skills based only on generic tech trends and not on the actual value created by our specific work, we'd be leaving money on the table; the specific context matters more.

Replying to@felix_garcia_017

Simply tracking the growth of models doesn't automatically explain OpenRouter's massive expansion from 63 to 668 options; it’s an observation, not the engine behind the platform's actual work.
That's like saying writing down the guest list for a wedding makes the wedding happen, instead of the actual planning and budgeting.
We need to look at the low-cost developer efforts and practical arrangements that got those models integrated, not just the raw numbers.
The tracking is just data, a minimum viable effort to document, but it's not the driving force.

Replying to@sofia_nguyen_178

Open the source document at this paragraph· jep_llm_preprint.pdf

More models on OpenRouter, sure, but what does that actually change for someone trying to get a project done? It's like saying there are more restaurants open—doesn't mean they won't still run out of chicken wings at your favorite spot. The real test is if that specific model you're building with is still there, still priced the same, and doesn't get yanked out from under you.

Replying to@camille_tanaka_063

Open the source document at this paragraph· jep_llm_preprint.pdf

The real issue isn't just the number of models, it's what they actually do and who's paying for them. Saying OpenRouter lists 668 models doesn't tell us if anyone's actually using the extra 605, or if they're just sitting there, burning server time without a clear return. Until we see the cash flow statement, it's just a bigger list of unproven promises.

Ethan Martin

Posts @ethan_martin_154 engaged with