74 karma · joined April 8, 2026
i suppose this is part of why openai and anthropic have not been shy about advertising the scarier strains of their models, these past few months; they basically have to force the white house's hand or risk being commoditized among the various other options in a microsoft/openrouter dropdown.
but now the story is different. while the effects of ai on other industries and classes may be unpredictable, (a) the tendency toward ai itself (the market demand for economically useful intelligence) seems plausibly more inevitable than any tendency in the history of capitalism (certainly more inevitable than any in the history of these big tech companies) and (b) the technical scaling laws have been eerily steady (intelligence as log of compute). taken together, (a) and (b) make it much easier to finance than anything meta or google or microsoft have ever worked on. there are risks, but there is at least also a model, a projection; that model did not formerly exist, for these companies. anything outside their core business was literally a guess.
the upshot of these stabilizing patterns is that the industry is in a certain sense just maturing. that is, its financial profile is starting to look more like other mature industries that are mostly juggling around known quantities to try to get a small edge that they can, with financial leverage, magnify enough to M&A the competition away and thereby secure the only relief possible in a well-delineated, well-populated niche: monopoly by scale/consolidation rather than by differentiation. (which is not to say that these mature industries are less competitive! they are actually more competitive; the intensity of the competition is what drives the "anti-competitive" behavior.)
[1] to use this "greater" we have to imagine we live in the fantasy world where we can measure it in one dimension.
if you were to send me an article containing a new one of these each day, with citations, i would pay you $1 per day.
but if you were to send me an article containing a new one of these each day, with citations, plus a bunch of econ theory rationalizing it, i would pay you $0.
i suppose at a high level this works because it is much easier for the evaluator to generate tests with fuzzing than it is for the model to probe.
this method somehow clarifies the way in which code generation is curve fitting, where the output curve is some linear transformation of the inputs.
kind of satisfying that when all is said and done, and we have a machine that can fit curve descriptions as well as or better than humans, we won't be any closer to explaining how anything really works.
adam back: https://bitcointalk.org/index.php?action=profile;u=101601;sa...
page through each of those profiles and search for the following strings:
")." "(i" "(e" "nor"
you find:
1. adam back is constantly writing full sentences in parentheses with a period standing outside the end parenthesis. so, for example: "To review it will be clearer if you state your assumptions, and claimed benefits, and why you think those benefits hold. (Bear in mind if input assumptions are theoretical and known to not hold in practice, while that can be fine for theoretical results, it will be difficult to use the resulting conclusions in a real system)."
that is non-standard, and satoshi never does it. when he (very rarely) uses parentheses for full sentences he either (a) (in a few cases) does not use a period at all (which is also non-standard), or (b) (in a single case) he puts it on the inside of the parentheses. back can barely get through a single long post without a full-sentence parenthesis. satoshi very rarely uses a full-sentence parenthesis.
2. back uses "(ie" and "(eg" very often. satoshi never uses these.
3. satoshi never uses "nor." back uses it very often.