Hacker Newsnew | past | comments | ask | show | jobs | submit | LeifCarrotson's commentslogin

If a neighboring farmer needs a bit of cash, has some land or equipment, and gets an email (or phone call!) from farmerfred@proofofcorn.com reading generally:

> I'm about to lease some acreage at {address near you} and willing to pay {competitive rate} to hire someone to work that land for me, are you interested?

I see no reason why that couldn't eventually succeed. I'm sure that being an out-of-state investor who doesn't have any physical hands to finalize the deal with a handshake is an impediment, but with enough tokens, Farmer Fred could make 100,000 phone calls and send out 100,000 emails to every landowner and work-for-hire equipment operator in Iowa, Texas, and Argentina by this afternoon. If there exists a human who would make that deal, Fred can eventually find them. Seth would be limited in his chance to succeed in these efforts because he can only make one 1-minute phone call per minute, Fred can become as many callers as Anthropic owns GPUs.

I do find it amusing that Fred currently shows the following dashboard:

    Iowa
    HOLD
    0°F
    Unknown (API error)
    Fred's Thinking: “Iowa is frozen solid. Been through worse. We wait.”

     Fred is here
    South Texas
    HOLD
    0°F
    Unknown (API error)
    Fred's Thinking: “South Texas is frozen solid. Been through worse. We wait.”

    Argentina
    HOLD
    0°F
    Unknown (API error)
    Fred's Thinking: “Argentina is frozen solid. Been through worse. We wait.”
Any human Fred might call in the Argentinian summer or 70F South Texas winter weather is not going to gain confidence when Fred tries to build rapport through some small talk about the unseasonably cold weather...

> Fred is here

Ah, they've created SCP-423


On one end, a farmer or agronomist who just uses a pen, paper, and some education and experience can manage a farm without any computer tooling at all - or even just forecasts the weather and chooses planting times based on the aches in their bones and a finger in the dirt. One who uses a spreadsheet or dedicated farming ERP as a tool can be a little more effective. With a lot of automation, that software tooling can allow them to manage many acres of farms more easily and potentially more accurately. But if you keep going, on the other end, there's just a human who knows nothing about the technicalities but owns enough stock in the enterprise to sit on the board and read quarterly earnings reports. They can do little more than say "Yes, let us keep going in this direction" or "I want to vote in someone else to be on the executive team". Right now, all such corporations have those operational decisions being made by humans, or at least outsourced to humans, but it looks increasingly like an LLM agent could do much of that. It might hallucinate something totally nonsensical and the owner would be left with a pile of debt, but it's hard to say that Seth as just a stockholder is, in any real sense, a farmer, even if his AI-based enterprise grows a lot of corn.

I think it would be unlikely but interesting if the AI decided that in furtherance of whatever its prompt and developing goals are to grow corn, it would branch out into something like real estate or manufacturing of agricultural equipment. Perhaps it would buy a business to manufacture high-tensile wire fence, with a side business of heavy-duty paperclips... and we all know where that would lead!

We don't yet have the legal frameworks to build an AI that owns itself (see also "the tree that owns itself" [1]), so for now there will be a human in the loop. Perhaps that human is intimately involved and micromanaging, merely a hands-off supervisor, or relegated to an ownership position with no real capacity to direct any actions. But I don't think that you can say that an owner who has not directed any actions beyond the initial prompt is really "doing the work".

[1]: https://en.wikipedia.org/wiki/Tree_That_Owns_Itself


Judging by the sheer verbosity of your reply there... I think you missed the cogent point:

> Seth is a Tool

It's that simple.


I have occasionally tried checking Polymarket and Kalshi to get an idea of the general political/cultural/technological consensus on various issues that are difficult to research otherwise, eg. "what are the chances that the Senate changes hands in the 2026 midterms?" People have thought about it enough to wager a million dollars and the consensus is at about 1/3. I have this abstract prediction market in my head, each bet placed by some statistically average person with diverse experiences and exposures from my own bubble, who carefully considers their information and puts their two cents into the pot, and I assume that by adding all our ideas together we form some sort of combined intelligence which is more insightful and reliable together than any individual pundit could be.

And then I go back to the home page, and see all the rabid sports fans, and realize that these bets are not being placed by deep thinkers.


Polymarket is currently predicting a 3% chance of Christ returning by 2027

https://polymarket.com/event/will-jesus-christ-return-before...


That’s because 2027 is too far out, as the deadline gets closer it will trend towards 0%.

Buying “Christ doesn’t return” at 99 cents right now would be unprofitable, you will wait a year or so to “profit” 1 cent.


It will be resolved by "credible sources".

The then credible sources did not believe Jesus was the Christ at the time of the first coming!

If there was a second coming what would money be worth?

What exactly are people betting in this thinking? Just betting on anything with long enough odds?


Election polling, analysis, and prediction is a mature industry with plenty of reliable commentators who can help answer your question. Here is just one example:

https://www.cookpolitical.com/analysis/senate

Portrayal to the contrary is mostly due to non-experts pumping their own ego, or deliberate media spin.


I'm not sure I'd go that far, but they should at least be aware of the credibility of information sources they're using to making professional decisions upon. Your world gets very small if you can only gather information that you see with your own eyes, but they do need to validate things they "learn" from Google/FB/Reddit/Youtube/non-official sources. I was unimpressed by the chief's letter:

> In preparation for the force response to the HMICFRS inquiry into this matter, on Friday afternoon I became aware that the erroneous result concerning the West Ham v Maccabi Tel Aviv match arose as result of a use of Microsoft Co Pilot[sic]. Both ACC O’Hara and I had, up until Friday afternoon, understood that the West Ham match had only been identified through the use of Google.

My 3rd grader knows better than to do research based solely on a Google summary snippet, and even understands that just because a linked article under the search agrees with the search that this doesn't mean it's true.

I would have expected that if the chief's staff were investigating rumors of a riot in a stadium 3 hours away, they'd call their counterparts at the police station in that location to get police reports from the incident.

They have trivial access to those official reports. They shouldn't be reliant on journalists sensationalizing, and opining the events for their news articles. They shouldn't be reliant on a search engine that exists to sell ads for those news organizations. They certainly shouldn't trust "Co Pilot" to figure out what may or may not have happened! It seems obvious to me that the tool could happily generate a police report from whole cloth.


I'm a little confused by the "Yes" versus "No" asymmetry.

For example, one of the top trending ~~bets~~ markets right now is on whether Miami or Indiana will win the NCAA football championship tonight. You can either take "Yes" on Indiana at 74c, or "No" at 27c, or you can take "Yes" on Miami at 27c or "No" at 74c. Or, there's another potential outcome - you can also bet on a tie at 10c yes/91c no.

Is this research suggesting that an optimistic Miami fan can somehow get a better return by buying "No" on Indiana than a "Yes" on Miami?

Why is Kalshi structured with these yes vs. no options for all outcomes?


> Why is Kalshi structured with these yes vs. no options for all outcomes?

it's basically how they do margin. otherwise you wouldn't be able to sell / post asks without already having a long position. for kalshi, it's actually one single security in the background they just present it as two order books (but really it's one). for polymarket, they are two distinct products that trade separately, and technically could have arbitrage between them. although in practice they're normally priced correctly to sum to 1 (or 1.01)


It’s not really margin since there’s no leverage: the potential loss associated with the bet has to be deposited, so it’s fully collateralised.

right, I guess I should have said it's what they have _instead_ of letting users trade on margin.

Part of this perceived arbitrage is the fee structure. Kalshi has a weird transaction cost structure but taking advantage of that 1c arb probably costs you 2c in fees to Kalshi, so nobody does it.

This is incorrect and tiresome to constantly hear as a go-to explanation for supposed market inefficiencies.

10c yes seems really high for a tie. NCAA rules don't allow for ties in football. I know prediction markets have very long shot bets but I would expect that to be closer to 1c

Edit: it looks like the tie market is only for if the game is tied at halftime, which makes much more sense


It doesn't say that they accounted for possible changes in item quality. Tide detergent claims that their new 80-oz bottle of laundry detergent can wash 64 loads just like the previous 100-oz bottle because it's more concentrated, and I suppose NPR (if they'd retained a sample of the previous product) could have brought that to a chemistry lab to test and verify that claim, but I have no idea how you'd objectively prove that an ounce of salsa had truly remained the same product.

Sure, but they are accounting for size shrinkage which the original poster was saying they didn't.

I don't really know how you can account for quality either. User surveys? Ingredient sourcing? But then again I think this kind of reporting is just a general barometer. Some other comments are pointing to data sources that might do more of this.


Laundry detergent is usually priced as cents per load by savvy shoppers. That would factor out smaller doses.

The amount of detergent per load is set by the manufacturer, who can injection mold that measuring cup in whatever size they want. FTA:

> The amount of liquid had shrunk to 92 ounces from 100 ounces before the pandemic, and the price had risen by a dollar. After that, the cost stayed the same, but the contents shrank to 84 ounces in 2024 and then to 80 ounces in December.

> The label continuously promised enough detergent for 64 loads of laundry.

> ...Tide specifically got the "most significant upgrade to its liquid formula in over 20 years," according to the company, with a "boosted" level of active cleaning ingredients and updated dosage instructions.

> "The result is superior cleaning performance in a smaller dose," a Procter & Gamble representative said.

Do you take them at their word for that? I'm specifically wondering whether the 84 ounce, 64-load bottle with a cap that measures out 1.3125 ounces per load contains the exact same liquid as the 80 ounce, 64-load bottle with a cap that measures out 1.25 ounces per load. I prefer powder detergent with a prewash dose, I know my clothes get clean, but I don't know that anyone outside a lab would be able to inspect clothes post-wash and notice the difference in cleanliness caused by the removal of 0.0625 ounces of detergent.

They have three ways to protect or boost profits: Raise prices, decrease quantities, or decrease quality. NPR and the


At some point it will be concentrated it'll be powder again ;-)

Eh, if the post office - which is a pretty efficient operation - thinks it costs $0.61 to mail a postcard, it probably costs Amazon more than $0.25 to ship someone a lime.

"You have a collect call from MomWe'reAtTheArcadeCanYouPickUsUp?

Would you like to accept the charges?"


Ahh, the 70s. Good memories.

I recall using this occasionally in the 90s even. There was also a period where I would regularly "one ring" my parents as code to call me back. IIRC that was because my cell plan had unlimited (or at least more) incoming minutes.

It will never stop being weird to me how US mobile plans make (or made) you pay for incoming calls and texts.

Collect calls would have been a thing until cell phones got very common because you had to use pay phones when you were outside and you somehow never had quarters.

It's really easy to anthropomorphize an LLM or computer program, but they're fundamentally alien systems.

I'm pessimistic that future paradigm AIs will change this anytime soon - it appears that Noosphere89 seems to think that future paradigm AIs will not have these same limitations, but it seems obvious to me that the architecture of a GPT (the "P" standing for "Pre-trained") cannot "learn," which is the fundamental problem with all these systems.


There's negligible "genetic" difference between German and American gastrointestinal systems. No DNA mutations occurred in your grandparents that caused all of their children and children's children to be overweight.

There may be cultural or behavioral issues - attitudes and habits around cooking, expectations of what a meal includes or does not include, taste preferences on what's too sweet or too fatty, etc - but it's not genetic.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: