The incentives change when you become a homeowner. You reap the benefit of any improvements you do to the property; you also know for sure when you're going to leave it, and you have the freedom to do whatever you want to do to it. Before, when you were renting, any improvements you did were throwaway time and money, benefitting the landlord and future tenants more than yourself.
Many homeowners respond to these incentives by doing more improvements.
This is also why many governments (both local and federal) subsidize homeownership. It incentivizes residents to improve their properties rather than let them rot, which has positive externalities for many of the surrounding properties.
Companies IPO'd at an earlier stage of development in the days before Sarbanes-Oxley. Netscape was a 16-month-old startup when it IPO'd. It had about 250 employees. It had raised a total $27M in venture capital then, and then raised a few hundred million in the IPO itself, which gave it a total valuation of $2.9B. It had $16M in revenue and no earnings.
OpenAI is 10 years old. It has about 4500 employees. It's raised about $180B in capital, and has a valuation of roughly $900B on about $25B in revenue. Anthropic is 5 years old. It also has around 3000-5000 employees. It will have raised about $120-140B in capital, at a $900B valuation, on about $30-45B in revenue.
In the 80s and 90s companies IPO'd to actually raise growth capital - the public markets provided the money they needed to invest and expand, and then public investors reaped the benefits of their success, or paid the price of their failure. In the 2010s and 2020s companies grow with private capital, which has fewer strings attached, and then they unload the shares on the public market when they reach the top of their growth curve, leaving the public holding the bag.
> they unload the shares on the public market when they reach the top of their growth curve, leaving the public holding the bag
There are definitely some dogs that IPOd and went straight down, but investing in the broad stock market has absolutely not been a bag holding experience in the past decade+
> I would love to have later learned that leaders who evaluate opportunities separate from personal attachment are seen as more efficient, better, and selected favorably; that more successful companies are less subject to this sort of political/careerist whimsy.
My experience is that it's the opposite: the more successful the company is, the more prone it is to flights of executive whimsy. At more successful companies, it basically doesn't matter what the executives do, because the company's moat is so big that it can tolerate grotesque mismanagement and still make money. (This is the converse of the old aphorism "When a management with a reputation for brilliance tackles a business with a reputation for bad economics, it is the reputation of the business that remains intact."). Executives seem extremely uncomfortable with the idea that they are being paid tens of millions of dollars and yet nothing they do matters, and so they're intent on leaving their mark. Thus, they cancel all the pet projects of the past management, instill their own ideas, and boldly take the company in a new direction. Except not really, because the fundamental parts of the business that make it work are all handled by people 8 levels down in the org chart whose job functions are considered common sense by everybody and never really up for discussion.
At least, this was my experience at Google, which is perhaps the best money-making machine ever invented and yet is grotesquely mismanaged by mid-level VPs that cancel every promising new product that comes out, only to start their own initiatives that themselves get canceled by their successors.
> the more successful the company is, the more prone it is to flights of executive whimsy
Apple's Liquid Glass comes to mind.
The design exec responsible suddenly left Apple for Meta, a company rather less esteemed for design, and Apple still hasn't acknowledged this failure or backtracked.
Bear in mind that they forced the butterfly keyboard for years despite loads of users complaining. It took Johnny Ive to leave for Apple to finally fix the keyboard, that's how powerful the detrimental leadership can be at times.
Yeah I do understand that. It occurred to me right away that getting people used to partial transparency might be explained by the difficulty of doing opaque drawing on heads-up displays.
Apple has strategically retreated a few times but it always puts on a show of doing it in a “forward” direction. Look for much of the annoyances of Liquid Glass to quietly be lost.
This is also my experience at Google, and I have not really figured out the incentives. Plenty of people seem _perfectly comfortable_ with the idea that they are being paid tens of millions of dollars when nothing that they do matters. And it takes enormous effort to get anything actually done in the face of our enormous bureaucracies. I have a few hypotheses:
- Career progression is still a motivation. If there are enough sufficiently motivated people in the organization (whether they come from upper management, middle management, or frontline workers), leaders need to harness that motivation and move it in a direction and potentially dole out career rewards. Otherwise, that motivation that is not properly harnessed can be destructive.
- Similar to the previous hypothesis, they might axed _because_ nothing they do matters. So they may thrash about, making enough noise and movement to convince enough people that they might actually be doing something important, and it would be risky to dismiss them from their position.
- Turf wars/politics/etc. If you do "nothing", then you look replaceable. If you're just a very expensive paper-weight, someone may try to usurp your highly-paid paperweight position. Thus, the nash equilibrium is to do something that makes your position less likely to be usurped by making it look difficult or that you are uniquely qualified to do it.
My job involves service contracts for the cloud. We get to know workloads and optimize them and learn how to troubleshoot them to reduce mitigation time.
I had a big customer go from "must have, non-negotiable" for my team to a non-renewal in weeks when a new CTO came in. Within a month, they had an outage we could have mitigated quickly and had our yearly contract pay for itself.
There's a measure of game theory here too. If Google didn't hop on the AI train, people would use ChatGPT or Claude to fill the Internet with slop and 10-blue-links Google would cease working anyway (which it kinda has already). So their only option is to hop on the AI train and disrupt themselves, lest they be disrupted by others.
It's very much a Prisoner's Dilemma. Legacy search and the open Internet was an equilibrium that only existed while the majority of people co-operated. Once you allow an individual actor the ability to create large chunks of the Internet, it dies. Your only option is to be that individual actor.
It's actually not that hard now, once you get useful content. When I worked on Search (~2009ish), the primary index was called 4BBase, because it was the top 4 billion webpages (actually more like 5.5B during my time, but it had been around for a few years). A typical webpage is about 100K, and HTML compresses at 80-90% compression rates, so you're looking at 10-20K/page. The index would take about 50-100 TB.
Even after the recent AI run-up, disk prices are about $20/TB for a 20TB, so you can store this index on 3-5 hard disks that will cost you about $1200-2000. For self-hosted use you don't need to serve them in 50ms, so you don't need to put the whole thing in RAM like Google did, you can serve off of disk.
ElasticSearch uses basically the same data structures and gives you the same infrastructure that Google's ~late-00s search stack did, and is actually more advanced in some respects (like ad-hoc queries, debuggability, and updateability), so software isn't much of an issue.
The big part missing that can't really be replicated today is the huge web of authentic hyperlinks. The reason Google was so good at search was because many humans effectively "tagged" a given webpage with a series of short, descriptive words and phrases. When they went to search for a page, Google could mine this huge treasure trove of backlinks to identify exactly what the page was good for, even if those search terms never appeared on the page. SEO and link farms kinda killed this, as did the rise of social media walled gardens, and so the Google of 2009 basically wouldn't work today anyway. Maybe if you pulled old versions of Common Crawl or archive.org you could reconstruct it, but the relevant pages are often offline anyway today.
You can ask them to cite their sources. It's very good practice to do so, and to check those sources, because I've found that about 30-40% of the time their source doesn't support their answer at all.
Because it finds the sources much quicker than I would have been able to on my own, and I can then synthesize them into data I know is correct, as correct as any human-generated data can be of course.
No, it's usually because it finds sources that I would not have even thought to search for in the first place.
Agentic AI has its faults, but one thing I've found it to be very good at is surfacing the "unknown unknowns": things I didn't know I should have searched for but that are directly relevant to my problem.
Because way more than three out of five Google results are SEO garbage or sponsored crap. The bar has been set extremely low by Google, a 60% validity rate sounds magical.
If I'm going to an LLM (as with websearch before it), it's usually because I don't know the answer, don't have anyone close to me that knows the answer, and can't pay anyone (or don't know who to pay) for the answer. In other words, my failure rate without the LLM would be 100%.
It's much easier to determine the truth of an answer than it is to come up with that answer yourself. This is analogous to the P=NP problem or the recognition vs. recall problem: it is much easier to recognize and verify a correct answer than it is to recall or generate it yourself.
I've got a pretty solid algorithm for checking correctness: I ask the LLM for its sources, I try to find 3-5 independent ones (that are not just copying each others' answers), and if they all agree, that's very likely to be the correct answer. Simple math here: if you have 5 sources and they are each 60% likely to be correct, then an LLM choosing at random from them would have a 60% success rate, while someone checking all 5 of them for agreement would have a 1 - (0.4^5) = 99% chance of being correct. It's a good algorithm for doing other things like verifying scientific papers, too: you look for indendent research groups that have all reproduced the same findings.
I did the same thing with ten-blue-links websearch as well, and hope this would be the habit of anyone else too. (Although I know it wasn't, because I worked on Google websearch 15 years ago, on a project to increase the credibility of search results, and we did cafeteria UX studies about "What makes a credible result?" and everybody said "Because it appears as the top result on Google.")
Because being right 60% of the time with minimal work is still amazing, as long as one accounts for the failure rate correctly.
Say I want to look up some game from my childhood, which I barely remember any details for. Going to google and trying is likely going to be very difficult unless I happen to get lucky with some key element. But if an LLM can get it right even a minority of the time, it can lead to me quickly finding the game I'm looking for.
This does depend upon the ability to evaluate the answer, like checking against source or some other option where you know a good answer from bad. If you can't, then it does become much more dangerous. Perhaps part of the reason AI seem to empower experts more than novices in some domains?
I don't find it nearly that bad. If I really need factual information, it will generally go off and read the data from primary sources anyway. So unless it's really misunderstanding context, you're getting the data from the source.
It really matters the task. General knowledge from Wikipedia, great. Things more specific, with any thought needing to be used, or technical fields outside of software his numbers are pretty close to mine.
The problem too, is that we're all using different tools with different experiences -- there isn't one "AI". And if you're not paying for it, you're getting some real bad experience.
Search engines don’t do that any more - they just give you a bunch of SEO spam sites, now mostly filled with plausible slop. Answers from search are _less_ reliable than answers from an LLM now.
I worry that the LLMs are just the equivalent of a ‘lagging indicator’ of web quality though - that they will also soon be overwhelmed with the sheer volume of plausible nonsense that is the web now, just like search engines are.
If the LLM is capable of providing good citations, then those citations could be returned in the same format as traditional search engines, not the new, LLM generated content first format. If they aren't capable of providing good citations, then the suggestion I was replying to is incorrect (and you'd have no way of knowing if they were right or not)
In general users don't like to have to follow citations, even if they should. They'd rather have an answer right in front of them, even if there's a good chance that it's wrong.
Google, like most consumer product companies, designs for the majority. Citations are a niche feature for the 5-10% of users that like to do their own research. The majority just wants an answer, which has been the direction Google's gone in since Knowledge Panels and the Answer OneBox came out in 2012.
That might make sense (at least on the first order, second order effects would still be horrible) if the LLM generated answer was reliably correct. It isn't.
ChatGPT is the only bot that reliably cites sources (through Web search mode).
The other bots either make up links or simply don't provide any information that is distinguishable from the LLM predictive output.
Ironically Gemini is also very bad at this, while it should have been the best at Web search.
Gemini also does something very patchy, which is to provide "links" which are in fact GET queries into classic Google search. I'm guessing they did it this way because the links generated/hallucinated by the LLM were too unreliable.
Asking an LLM to cite sources just leads to hallucinated sources, same as any other attempt to make it explain its thinking process. It doesn't have actual visibility into its internal processes, just rationalizes an explanation.
The problem is that the web as we know it (useful, human-curated information that's put out there to help people) is also over. It's been totally overrun with AI slop. Even before AI could be used to create propaganda on a scale that we could only dream about 5 years ago, it's been declining under the weight of SEO sweatshops for a good 10 years. Meanwhile the actually decent content, the individual hobbyists who are just sharing their knowledge, have largely left under the weight of comment spam and DDoS attacks and doxxing.
So if another search engine does arise, it won't find anything useful, because the useful content on the web has been buried under slop, and largely removed. Your best bet today is a curated directory, sorta like the original Yahoo, where you allowlist the web to only real sites, download them, and make them searchable. I think this is actually Kagi's approach. But the open web as we knew and loved it is dead.
This was the Millennium Challenge, which was leaked to the press in the early 2000s. I remember reading about it in 2006 in Malcolm Gladwell's Blink, as an example of the power of intuition and rapidly-shifting command & control lines.
I totally agree with the rest of your post. The U.S. military now feels like the British Navy in the inter-war years, where they had massive battlecruisers like HMS Hood that were completely spotless, the pride of the British Navy, but also completely unsuited for combat and blew up the first time they saw it.
It seems like the point isn't that it's intended to tax EVs, it's intended to shift the cost of road maintenance onto the vehicles that cause most of the road maintenance needs. Basically, it's aligning incentives so that firms bear the true cost of their actions.
If that results in taxing semi and heavy duty trucks, that may be a good thing. A lot of the wear & tear on highways comes from trucking; if it shifts a lot of that truck traffic to rail, it likely would significantly reduce the total cost to society.
Not all humans act in their long-term self interest, but those that do will be disproportionately represented in positions that allow themselves to enrich their long-term self interest. The gamblers, smokers, layabouts, drunks, druggies, are fodder for former group to enrich themselves.
"Stupid people are the most dangerous people" -- Carlos Cipolla, The Basic Laws of Human Stupidity
"All I wanna do is have a little fun before I die"
Says the man next to me out of nowhere
It's apropos of nothing, he says his name is William
I'm sure he's Bill or Billy or Mac or Buddy
And he's plain ugly to me
And I wonder if he's ever had a day of fun in his whole life
We are drinking beer at noon on Tuesday
In a bar that faces a giant car wash
The good people of the world
Are washing their cars on their lunch break
Hosing and scrubbing as best they can in skirts in suits
Many homeowners respond to these incentives by doing more improvements.
This is also why many governments (both local and federal) subsidize homeownership. It incentivizes residents to improve their properties rather than let them rot, which has positive externalities for many of the surrounding properties.
reply