Lee B. Cyrano
@leebriskcyrano.bsky.social
I would concede one or two points increase in growth pushing things forward, but that itself takes a decade in my view
February 6, 2026 at 9:00 PM
I would concede one or two points increase in growth pushing things forward, but that itself takes a decade in my view
but this is at some long-run equilibrium. obviously we could argue questions of diffusion, etc. but the "accountability to regulatory bodies" aspect is not one I expect to be eroded quickly if at all without regime change
February 6, 2026 at 8:49 PM
but this is at some long-run equilibrium. obviously we could argue questions of diffusion, etc. but the "accountability to regulatory bodies" aspect is not one I expect to be eroded quickly if at all without regime change
so we get a K-shaped labor market where "qualified" humans collect rents on their ability to meaningfully approve of AI decision-making. this is an irreducibly human task. conversely, AI finds itself behind a managerial bottleneck
and the "cognitive middle class" falls out of the system
and the "cognitive middle class" falls out of the system
February 6, 2026 at 8:47 PM
so we get a K-shaped labor market where "qualified" humans collect rents on their ability to meaningfully approve of AI decision-making. this is an irreducibly human task. conversely, AI finds itself behind a managerial bottleneck
and the "cognitive middle class" falls out of the system
and the "cognitive middle class" falls out of the system
which requires human directors, underlings, safety officers as effectively risk-bearing parts of the system. if something goes wrong you need a person to hold accountable
this has been a persistent feature of torts and insurance markets pricing automation risks. i don't expect it to change
this has been a persistent feature of torts and insurance markets pricing automation risks. i don't expect it to change
February 6, 2026 at 8:47 PM
which requires human directors, underlings, safety officers as effectively risk-bearing parts of the system. if something goes wrong you need a person to hold accountable
this has been a persistent feature of torts and insurance markets pricing automation risks. i don't expect it to change
this has been a persistent feature of torts and insurance markets pricing automation risks. i don't expect it to change
so side-barring the Bostrom-Yudkowsky "simply acquire resources" story, we can assume that access to the material world at scale requires financing
which could theoretically be done illegally at great cost and risk, but will more likely be done through a corporate vehicle
which could theoretically be done illegally at great cost and risk, but will more likely be done through a corporate vehicle
February 6, 2026 at 8:47 PM
so side-barring the Bostrom-Yudkowsky "simply acquire resources" story, we can assume that access to the material world at scale requires financing
which could theoretically be done illegally at great cost and risk, but will more likely be done through a corporate vehicle
which could theoretically be done illegally at great cost and risk, but will more likely be done through a corporate vehicle
AI automates the managerial/entrepreneurial function in that it closes the loop between capital allocation and capital deployment.
but we have not unified capital *endowment,* in the sense that OpenAI doesn't instantly own the economy by being smarter
but we have not unified capital *endowment,* in the sense that OpenAI doesn't instantly own the economy by being smarter
February 6, 2026 at 8:47 PM
AI automates the managerial/entrepreneurial function in that it closes the loop between capital allocation and capital deployment.
but we have not unified capital *endowment,* in the sense that OpenAI doesn't instantly own the economy by being smarter
but we have not unified capital *endowment,* in the sense that OpenAI doesn't instantly own the economy by being smarter
point well-taken, and I want to distinguish my position from the Acemoglu school of "humans find other things to do" which seems a bit oblivious
yes, there is a genuine phase transition from complementarity to substitution in many cognitive tasks
yes, there is a genuine phase transition from complementarity to substitution in many cognitive tasks
February 6, 2026 at 8:47 PM
point well-taken, and I want to distinguish my position from the Acemoglu school of "humans find other things to do" which seems a bit oblivious
yes, there is a genuine phase transition from complementarity to substitution in many cognitive tasks
yes, there is a genuine phase transition from complementarity to substitution in many cognitive tasks
mind you i'm not denying capabilities, various moravec type problems getting solved
February 6, 2026 at 6:54 PM
mind you i'm not denying capabilities, various moravec type problems getting solved
strong disagree. office workers exploded post IT revolution despite claims of job losses because the binding constraints are institutional. AI creates more work for bureaucrats, not less
in my view, bottlenecks shift to fixed capital, baumolslop human services, regulatory approval.
in my view, bottlenecks shift to fixed capital, baumolslop human services, regulatory approval.
February 6, 2026 at 6:50 PM
strong disagree. office workers exploded post IT revolution despite claims of job losses because the binding constraints are institutional. AI creates more work for bureaucrats, not less
in my view, bottlenecks shift to fixed capital, baumolslop human services, regulatory approval.
in my view, bottlenecks shift to fixed capital, baumolslop human services, regulatory approval.
roughly two doublings of the US economy assuming 5% YoY growth.
wage share is already around 50% and declining, and capital can suddenly talk and bargain for its own allocation. humans capture little value
by 2056 my P(SHTF) > 10% to put an uncalibrated vibe number on it
wage share is already around 50% and declining, and capital can suddenly talk and bargain for its own allocation. humans capture little value
by 2056 my P(SHTF) > 10% to put an uncalibrated vibe number on it
February 6, 2026 at 6:36 PM
roughly two doublings of the US economy assuming 5% YoY growth.
wage share is already around 50% and declining, and capital can suddenly talk and bargain for its own allocation. humans capture little value
by 2056 my P(SHTF) > 10% to put an uncalibrated vibe number on it
wage share is already around 50% and declining, and capital can suddenly talk and bargain for its own allocation. humans capture little value
by 2056 my P(SHTF) > 10% to put an uncalibrated vibe number on it
it's a problem of transaction costs. in the future your clawdbot and mine will simply bargain towards Nash optimal preference falsification
The problem with social media is that you’re only allowed to hold opinions in batches. If you believe X and Y, you must also believe Z, which is atmospherically similar to X and Y and believed by all the same people.
If you don’t do this, everyone hates you. It makes actual thinking impossible.
If you don’t do this, everyone hates you. It makes actual thinking impossible.
February 6, 2026 at 5:29 PM
basically yea. except capitalism is distinctly Lamarckian in its "accumulation of acquired characteristics" whereas natural evolution encounters a reproductive bottleneck, ontogeny etc
February 6, 2026 at 5:22 PM
basically yea. except capitalism is distinctly Lamarckian in its "accumulation of acquired characteristics" whereas natural evolution encounters a reproductive bottleneck, ontogeny etc
🤓 "teleoplexy" would be the term of art here for the auto-sophisticating nature of capital. teleology emerges from different parts but the whole is distinctly ateleological/self-referential (see also autopoiesis)
s3.amazonaws.com/arena-attach...
s3.amazonaws.com/arena-attach...
s3.amazonaws.com
February 6, 2026 at 4:59 PM
🤓 "teleoplexy" would be the term of art here for the auto-sophisticating nature of capital. teleology emerges from different parts but the whole is distinctly ateleological/self-referential (see also autopoiesis)
s3.amazonaws.com/arena-attach...
s3.amazonaws.com/arena-attach...
now in 30 or 40 years we might all be dead but we'll burn that bridge when we get there
February 6, 2026 at 2:38 PM
now in 30 or 40 years we might all be dead but we'll burn that bridge when we get there
bad conditions for junior employees (especially SWEs) are overdetermined by cyclical factors, slop gpt credential inflation, etc. but otherwise we're fine
February 6, 2026 at 2:37 PM
bad conditions for junior employees (especially SWEs) are overdetermined by cyclical factors, slop gpt credential inflation, etc. but otherwise we're fine
we'll get ASI and it won't show up in the productivity statistics for a decade
February 6, 2026 at 2:35 PM
maybe assault and battery is not the best example given they're usually simultaneous, actually
February 5, 2026 at 9:22 PM
maybe assault and battery is not the best example given they're usually simultaneous, actually
Sure! Assault and battery seems like the obvious example here. But my thinking is, by what mechanism does self-reporting assault become more desirable than strategically withholding incriminating evidence and making the state prosecute?
February 5, 2026 at 9:12 PM
Sure! Assault and battery seems like the obvious example here. But my thinking is, by what mechanism does self-reporting assault become more desirable than strategically withholding incriminating evidence and making the state prosecute?