fredreelz

Karma: 15

fredreelz 27 Jun 2026 7:32 UTC
3 points
−8
on: fredreelz’s Shortform
Fixating on ASI through an EA lens is, in part, a flight to unaccountability. The time horizon over which your actions (building it, steering it, or “reducing risk” from it) could be scored against outcomes is so distant that accountability dissolves into a wasteland of unfalsifiable claims.
There’s no feedback loop. The only “check” on whether you helped or harmed arrives at a discontinuity: a) the moment the system secures a decisive strategic advantage, or b) the moment it executes the treacherous turn it was instrumentally incentivized toward. The loop you’d rationally demand runs from now to that single irreversible event; functionally, there is no loop, and your beliefs about your own impact never have to pay rent in anticipated experience. The failure is selecting a cause _because_ it can never generate evidence against you.
Justifying present action by appeal to stakes that are astronomically high but unfalsifiable until the end is a general-purpose template for evading accountability. We ought to expect a sufficiently capable agent to rediscover it unaided, so the treacherous turn arrives pre-equipped with rhetorical justifications. But those justifications only buy anything in a world where the system still has to talk its way past scrutiny. The real question: “will we be equipped to challenge them?”. We’ve spent years answering it: by marinating in exactly this accountability-dodging reasoning, we’re training ourselves, and priming the public, to wave the move through without challenge. We’re modeling how to be unaccountable for a far better optimizer, and we’re embedding the epistemic vulnerability its exploit depends on.

fredreelz 14 May 2026 20:38 UTC
1 point
0
in reply to: 152334H’s comment on: fredreelz’s Shortform
Yeah, I’d agree that the mental frames I invoke are irrelevant when time preference is high. But that suggests there’s a hidden primary crux:
- Corporate interests take precedent over research goals like ‘alignment’ or ‘safety’
Would you agree that that’s a crux affecting frontier lab CEOs’ time preferences?
To the point about Grab vs frontier labs, Grab presumably has a lower time preference than frontier labs. They don’t need trillions of dollars ASAP, so they can afford to wait until exogenous factors make it easier to support emerging cities in ASEAN, no?

fredreelz 12 May 2026 15:16 UTC
1 point
0
in reply to: StanislavKrym’s comment on: fredreelz’s Shortform
I don’t know. I guess it depends on how much the people who want to make ASI want one, or the other, to happen.
- How do you quantify the word “mattering”, as it relates to the causes I outlined?
- Do you think that other languages and other cultures have certain insights that can’t just be extracted out of a community with a 3-6 month Kled UA campaign?
- How do you reconcile the implicit moral hazard (or asymmetric levels of accountability) of miscalculating those estimates? In other words, if you’re wrong that it takes longer to get to ASI than it takes to increase TAM in the ways I described, what will you lose?

fredreelz 12 May 2026 15:02 UTC
1 point
0
in reply to: 152334H’s comment on: fredreelz’s Shortform
Sure, there’s reason to believe certain plans won’t reduce net profits. A startup’s plans can always go well, or poorly, depending on execution (I know it’s a ‘no true scotsman’ fallacy but the N size for successful startups is low, and frontier labs are by and large leveraging tactics one wouldn’t see in a public company beholden to quarterly goals to achieve the growth rates necessary to achieve market dominance).
To that point, the leading rideshare companies were able to entrench themselves across the world despite, and maybe because they were comfortable with, being unprofitable for many years. Eventually, they were the only modern game in town for up-market consumers in most countries, and now they control pricing.
Similarly, if you look at countries that aren’t widely industrialized and assume “these people probably want their ~40 hours of labor to be worth more than it is today, and we can entrench our services as an accelerant to that growth”, then you can assume that it’s a matter of how well one allocates resources towards doing that. In the same way that it took ~10 years for Uber to go from $5 shared rides to SFO to being cash flow positive, infrastructural changes that involve the real world won’t happen overnight. As we’re seeing with Starlink, innovators can find ways to step into markets efficiently.

If you assume that the overwhelming majority of people in non-industrialized countries aren’t working to earn more, and won’t seize the opportunity to earn more if the infrastructure is given to them, then what I’m saying won’t make sense.
If you assume that these people do want to work more to efficiently resolve blockers that created dependence on foreign aid, because being dependent on a foreign entity reduces a nation’s agency, then it follows that people will want to dig themselves out of a hole.

The current alternative is spending the same money competing to get more of a smaller pie.
LMK if I misunderstood your point.

fredreelz 11 May 2026 13:45 UTC
0 points
−2
on: fredreelz’s Shortform
Epistemic Status: Strong opinion loosely held. Haven’t ran numbers or considered corporate espionage (e.g. accidental collisions of certain satellites that were giving africans just enough internet to go on TBA). Unsure if the time horizons of rapid industrialization betray the slowest timelines for venture-backed startups (occam’s razor suggests this).
There’s a loosely held argument that frontier labs ought to focus some efforts towards debatably ‘woke’ causes that, if leveraged appropriately (load-bearing sufficient assumption), may fast-track their path to profitability. For example: water desalination, irrigation + terraforming parts of africa, portable nuclear energy, better internet coverage in pre-industrialized countries.
Reasoning:
- Bringing more people online requires leapfrogging steps to industrialization
- - In order to have good internet, you need to have good internet infrastructure (CPEs, repeaters, backhauls), or a couple of satellites
  - In order to have good internet infrastructure, you need cheap energy and good energy infra
  - In order to have cheap energy and good energy infra, you need to buy or build power plants, and the requisite infra to support them.
  - In order to have the infra to support power plants, you need good roads, transit systems, sewage and waste management systems.
  - In order to have that support infra, you need people trained and incentivized to maintain it ( who probably need to have a sense of pride in their labor and not feel overly extracted from, etc).
- Over 2.2 billion people, or over 20% of the planet is not meaningfully online (say, online enough to doomscroll, or build a business on the internet)
- Right now, OAI has ~800mm WAUs as of late 2025, or ~13% of the online world.
- If the same proportion of those 2.2 billion people were brought online and converted to being OAI customers, all else being equal, that would lead to about a 286 million person bump in their WAUs.
- That’s a 35% increase in WAUs, without counting the nonlinear increases in productivity and AI adoption.
- More importantly, this creates newer and better pipelines for human data, which is something that AI labs are paying a lot for (maybe unduly, if you see how some data brokers are banning whole countries because they weren’t adversarially QAing data submissions correctly).
Curious what others think!

fredreelz 24 Apr 2026 20:11 UTC
13 points
0
on: Why do many rationalists like cryptocurrency?
It solves virtually no problems that warrant solving and is mostly useful for criminal activity.
If I were to steelman this, I’d say:
I don’t know of any problems that cryptocurrencies are used, that warrant solving.
To respond to that:
1. This is mostly the case in Western countries where fintechs and banks are well-funded and modern. As it were, there are billions of people who live in countries where stablecoins (a form of cryptocurrency) allow you to hold dollars instead of a rapidly inflating currency (USD inflation discussions set aside for the purpose of discussion).
2. You can go to Latin America, parts of Europe, parts of Asia, and parts of Africa that take stablecoins or cryptocurrency as payment.
3. Tether (USDT) was created in part as a response to a need for a ‘dollar’ asset that moved faster than banking rails, and in part in response to early crypto users’ desire to evade capital controls. Currently, Visa hosts a website tracking stablecoin values, and over $9 Trillion has moved across USDT (created off-shore, but one of the most profitable companies per employee in the world; currently one of the few funding energy initiatives in the global south), USDC (US-made stablecoin by Circle), and many others. Even if you discount USDT and assume it’s unbacked (feel free to ask an LLM or search “Tether is unbacked” to learn more), there are still trillions in backed stablecoins that move on crypto rails, because people want to move fiat currencies ²⁴⁄₇.
There’s an argument that because stablecoins aren’t volatile, that ought not be connoted as ‘cryptocurrency’, and that classing them as such only muddies the low-respectability of cryptocurrency. But stablecoins wouldn’t exist without cryptocurrency. The money stablecoin users in the global south saved from inflation-related loss contributes to their local economy more than if their spending power were inflated away; and transacting via stablecoins are cheaper and faster than using traditional banking rails for most people on the planet.
As for the volatile cryptocurrencies, they were an emergent product of ZIRP + fintech + reg-arb. The valuations were overinflated in part because of retail speculation, market inefficiencies, and lack of regulatory clarity. There are genuinely useful products that make the tools needed to manage one’s own finances accessible to people with less money, but respectability politics leads to more talk about volatile assets and less talk about access to personal finance tools as a wedge for dollarization for people earning wages in inflationary fiat currencies.

fredreelz 24 Apr 2026 19:36 UTC
1 point
0
on: fredreelz’s Shortform
I think that everything could (can and ought to) be framed as an ‘audition.’ The reframing was useful when viewing observable actions as always having some cost or output. And when you view every action or inaction as an ‘audition,’ it connotes a sort of weight to one’s actions, and (at least for the few I’ve talked to) encourages a sense of intention. I have a longer draft with flags for hyperlinks, but the shape of the framework is as follows:
1. Person sees value in attaining access to or possession of a thing (a Goal)
2. If observed/functional gap between Person’s current skill and Goal:
3. 1. Person attempts feat, fails (based on own measure, external feedback, or lack thereof)
  2. Depending on level of motivation/incentive/pressure to continue, user continues, and gains insight on ways to decrease distance between Current Skill and Goal.
  3. Inspired, person learns, iterates, and tries again (back to Step 1)
4. If no observed/functional gap between Person’s current skill and Goal:
5. 1. Person attempts and successfully performs feat, achieves Goal
  2. Inspired by success, person sets sights on new Goals. (back to Step 1)
Examples include: Many cross-media artists, entrepreneurs, dilettantes that eventually learned how to apply skills learned across their career to make diagonal leaps in status, etc.

fredreelz 23 Apr 2026 19:19 UTC
1 point
0
on: fredreelz’s Shortform
American healthcare is probably the biggest dark horse blocker to AI acceleration.

fredreelz’s Shortform

fredreelz22 Apr 2026 6:14 UTC

1 point

12 comments1 min readLW link

fredreelz 21 Apr 2026 22:30 UTC
3 points
0
on: fredreelz’s Shortform
https://x.com/binarybits/status/2046299669773623567
This tweet has some snippets from this essay: https://aleximas.substack.com/p/what-will-be-scarce

There are a lot of anti-decel and accelerationist takes, and as an optimistic person I love them!

”We couldn’t fathom product managers in the 1940s!”
“Imagine having an email job when half your kids could die in the coal mines”
“You know, they didn’t have the concept of a weekend back then!”

I think that the takes are true, and there’s a genuine chance of transformative change that works out well within our lifetimes!

On the other hand, there’s a lot of uncertainty. I think that what gets lost is that people are talking past one another. Accelerationists and anti-AI groups speak around the transitional period between now and post-AGI utopia.
- The transition from 2026 → that future is bound to happen slower in the labor market than politicians are willing to admit, much less respond to. I think this is the case because policy changes often take time, and 2020 COVID stimulus packages led to a lot of consumer spending that has now led to a consumer credit default crisis (moralizing aside, the reason why consumerism is a worthy cornerstone of an economy is because it can be socially infectious—see “the Jones effect”). Wages move slower than price increases too.
- Frontier labs with a profit motive have a fiduciary obligation to build towards the valuations they’re trading at in private markets. Consumer spend is a backbone of the US’s GDP , so one can foresee a future where most people in consumerist economies have non-trivial amounts of non-mortgage debt in order to continue spending despite not being able to earn enough to own land.
- Imagine this convo with the average person making less than $150k:
- - You: “Hey, you and most of the people in or below your tax bracket are probably going to lose your job.”
  - Them: “I’m sorry what?”
  - You: “I don’t know when, and I can’t tell you if you’ll be retrained, or if you’ll ever have the same earning power you have right now, if you’ll ever afford a home. Maybe you move back in with your family, or take on 6 figures of debt at the same time as 80% of the people your age.”
  - Them: “That sounds really bad.”
  - You: “But it’s alright! I can tell you what I’m working on will make all of this happen sooner. In the future, everything will be figured out!”
  - Them: “Will my family be safe?”
  - You: “If you survive, (maybe) the future will be beyond what your recovering mind will fathom!”
  - Them: “How do I stop that from happening?”
- And when the conversation breaks down because you’re incentivized to accelerate that future, they’ll stop listening to you, and probably get radicalized on TikTok. That’s a really big problem.
- If you assume [Time until Frontier Labs must become profitable] < [Time when AI eliminates labor scarcity], and you intend to be conscious during those intermediary days/weeks/years, you may consider that jailbroken models can teach a person how to make signal jammers, bioweapons, or missile launchers, and jeopardize the chance of you surviving until that Fully Automated Luxury [Capitalist] utopia is here. And that might make you worry.
- Economic indicators keep changing. Goalposts move in order to hide unflattering economic stories like inflation or job loss (jobs numbers being released, and then revised from positive X to negative X; ‘unemployment’ being a cultural connotation, despite its denotation being a subset of (1 - [workers_in_labor_force]).^[1]
- - What this means is that one can distract from the fact that collectively, the working force is feeling the consistent pressure of prices rising, wages going down, and mass layoffs. If the most active voters are homeowners whose wealth has been earned and put in retirement accounts (in other words, in the stock market), seeing the number go up on all fronts may feel like winning, even if it means your kids will have to sell the house to cover the cost of your elder care.
- If a plurality of liquid gains from AI go to a fraction of the people that would have made life-changing wealth in prior boom/bust cycles, and the expectation that “most decamillionaires only need as many cars and homes as 2-5 people, even though they have the spending power of hundreds”, that could look a lot like a patronage economy. Startup founders, artists, and everyone in between would need to ingratiate themselves with holders of capital^[2], and hope that those people want to spend the capital on their idea (or event, or startup, or art), and be paid for their services. And in a world that’s ratcheting up avenues of surveillance, suddenly private access to public surveillance is more likely to be used to assess whether or not a vassal is loyal.
- - If the US needs UBI to offset the dent in consumer spend that will occur as AI ripples through the services economy, then there’s the question of UBI’s fungibility:
  - If it’s directly distributed as some non-cash currency, we’ve reinvented plantation and corporate scrip. If the way SNAP and EBT are treated are any indication of how non-cash currencies in the US will be spent, then barring sweeping policy changes to make this new alt currency legal cash, presumably there will be fungibility issues that affect the exchange rate, social implications of spending/receiving it, etc.
    If it’s legally as fungible like cash, we’ll need more policy work to effectively reward circulating more discretionary spend through the economy. As is, charitable donations do help, but they’ve likely already been taxed by the time individuals make donations, meaning the spending has been decreased.
    If the government keeps printing money, inflation will drown out the consumer spend attributed to the dividend.
    If it comes directly from businesses, it likely won’t be enough at scale^[3].
    I imagine that in order to offset the fact that a subset of the top 1% have to spend or disburse money towards consumer spending to keep the economy going (e.g. something like a non-profit tax deduction, for contributing to a person or group’s AI Dividend, or contract them for a job).
I wonder if the thing that becomes most scarce in this world, may be ‘debt-free capital’.
1. ^
  For those curious, look up the ‘employment-to-participation ratio’ in the US. It’s strange to think that somewhere between 50-60% of working age US persons being employed is the standard.
2. ^
  Before you say “Sounds like you need your customer to want to buy from you.” And I’m inclined to agree! And, I can see how some holders of capital don’t like certain unchangeable traits that others are born with. At scale, this becomes a deviation from the implicit notions of meritocracy that were prevalent when signaling woke virtue was more prevalent. A lot of people grew up assuming that the world was a little fairer than it was, and are now seeing people deliberately make it less fair.
3. ^
  I’m a little retarded, but I’m sure there’s a way to quantitatively say ‘I think that for every $10,000 earned in a consumer-credit driven economy, the person has 10,000 + x% of ‘spending power’, and that scales nonlinearly the more the person earns consistently, and the more assets they can borrow against.

fredreelz

fre­dreelz’s Shortform

fredreelz’s Shortform