Lots of critiques here! Something missing in this discussion is people asking _why_ it is that they're doing this. The people who work there aren't stupid!
I think this is a disconnect between people who think that large companies are static entities with established products vs. large companies that still operate like a startup and are trying to grow. When you're building your business from $0 in revenue, you don't know what will work! You try different things, you [launch over and over again](https://www.ycombinator.com/library/6i-how-to-launch-again-a...)...all in hopes of something that works, sticks, and starts to grow.
In every example here, I see OpenAI trying something new, hoping it will grow, and shutting it down after it doesn't. Sora is the pre-eminent example of this. They make news, but you don't talk about the things they launch that successfully grow!
OpenAI isn't shutting down Codex or ChatGPT, because those were launches that they did that actually worked! When you go look at the tweets and communication from OpenAI employees when ChatGPT launched, nobody was sure that it would work. But it did. And if they hadn't launched, we would have never known how valuable it was.
All that is to say...you don't know what will work until you launch. Most things fail, and it's correct to shut them down. But focusing on the products that haven't worked instead of the products that have gets you more clicks, but actually depresses innovation by making future launches less likely.
People are much more willing to give the benefit of the doubt on things like that when the flagbearers of your industry aren't running around sucking all of the oxygen out of the system and telling people things are "solved": that your product will obsolete them in the next 6-12 months.
We get it. They say that stuff to raise money, make sales and keep the party going. But don't expect too much sympathy when the strategy falters a bit.
Sora was losing 15M a day and it was running at least 3 months, so that's a total 1.3 Billion. That's a pretty expensive experiment. It sounds like a company with lots of VC to burn and no discipline. Even Jensen Huang accused them of lack of discipline in business approach.
Yeah. $1.3B isn't scrappy startup pivots, it's the sort of money Google/Meta/Microsoft/Yahoo/Salesforce burn on strategic acquisitions. And those entities absolutely get and deserve the sneers when they "sunset" the product 6-18 months later having concluded that it wasn't even showing enough signs of being a market they should bother with keeping the lights on.
At least Sora was novel and technically impressive, I guess.
Ah, but when you're little, if each misstep annoys a few early users who hit a dead-end with your project, the longer-term reputational damage is trivial. You've still got 99.999% of the TAM (total addressable market) that is ready to be charmed by something new, with no negative vibes in their mind.
As you get bigger, serious numbers of people get annoyed at dealing with a company that keeps inviting us into the Roach Motel of doomed products and features. Big case in point was Google's spree, a few years back, in terms of launching big new services/features that soon afterward got shut down. Great training ground for ambitious PMs; miserable user experience.
Somewhere between the death of Google+ and the demise of Google Hangouts, even folks like me began thinking: Why should I engage with new Google stuff if it's likely to be blown up in a few years, leaving me with buried IP from whatever I tried to do?
I'm not sure your criticism is quite fair. I think everyone here is willing to cut more slack to the underdog. But when your company represents an outsized chunk of the digital economy and employs 10k+ people, and only then says "sooo, let's try to build some sort of a profitable product here", I can see why people are rolling their eyes.
OpenAI also burned a lot of goodwill by pretending to be a nonprofit foundation focused on the betterment of mankind and then executing one of the most spectacular rugpulls in modern history. So yeah, people will be giving them a hard time even if it turns out that the valuation is justified.
Companies with non-stupid people can still do stupid things.
I think the issue with the experimentation is that they still don't have an obvious golden goose yet. Google has been able to fuck around with experiments because search/ads are always still there to carry the team and provide an infinite money spigot, even if the experiments mostly fail. But OpenAI doesn't really have an equivalent for that.
Or perhaps they lucked into chatGPT and the true prowess of the product function is being laid bare. They've had many failed projects now: that shopping stuff, a web browser, Sora.. the success they are having from Codex wasn't an original idea but a classic-case of copying. Have they run out of steam?
Very much possible. What has come out of Meta organically besides Facebook? Its valuation comes predominantly from the assets they acquired + investments yet to be made that build on the acquired assets.
Google is similarly iffy with product development. OAI is better still, marginally, in terms of delivering a more polished experience.
Also I do regard them stupid, simply because they are not following wisdom that was shared decades ago by someone with an incredible batting average when it comes to innovation: start with the customer experience and work backwards to the technology.
> When you're building your business from $0 in revenue, you don't know what will work!
When you announce a post money valuation of $852 billion, you should probably be a bit better at figuring out what works, though. You're not a scrappy startup any more, even if you like cosplaying as one.
They very nearly gave Elon Musk a controlling interest in the company. Their justification for not doing so was entirely vibes based. "Stupid" is a broad categorization, someone can be smart in some areas and do dumb things. You shouldn't let your personal appraisal for someones talent color the actual results they produce.
Something missing in this discussion is people asking _why_ it is that they're doing this. The people who work there aren't stupid!
They have infinite amounts of investor money to burn and no obvious way forward. TFA's line about "spaghetti at the wall" pretty much summed up what happens in that situation.
And in terms of "the people who work there aren't stupid", you can have technically talented people who are very good at their specific thing and hopeless at anything else, a friend of mine once summed it up as "the dumbest smart people I ever met". This is why you need skilled management to let them do their thing but also steer them in the right direction as they're doing it. From the descriptions of OpenAI it's kinda rudderless apart from the one-man hype machine at the top.
Ben Thompson interviewed UA's CEO on Starlink a few months ago.
Scott said: "It took time to negotiate, because we wanted to own the consumer data, and at the beginning, Starlink did, so that was hard, and then, the other thing was I wanted to let my big competitors in the United States finish their deals with other providers and get locked in so that we would — eventually, everyone’s going to have Starlink."
Brilliant. Just brilliant. Ensured that UA would be first (of the 3 major US carriers) to Starlink and that everyone else had to wait until their existing agreements multi-year expired before switching. UA's best CEO in decades!
I've definitely thought about substituting a nonstop flight for a 1-stop flight on UA regional jets just to get Starlink on the entire route. The annoying this is I live by a UA hub and UA doesn't fly regional planes between UA hubs.
So the best I've been able to do is a regional flight to a UA hub near me, and then a non-regional flight back to my home airport. Which is honestly probably not worth it. And it's definitely not worth doing a two-stop trip so I'm really excited for them to roll it out on their mainline jets!
Interesting, because I’d choose not to fly with an entire carrier because they have chosen to implement Starlink and support a Nazi pedophile, but horses for courses.
if you want that kind of control i think you should just try buff or opencode instead of the native Claude Code. You're getting an Anthropic engineer's opinionated interface right now, instead of a more customizable one
"Can we make the change to change the button color from red to blue?"
Literally, this is a yes or no question. But the AI will interpret this as me _wanting_ to complete that task and will go ahead and do it for me. And they'll be correct--I _do_ want the task completed! But that's not what I communicated when I literally wrote down my thoughts into a written sentence.
I wonder what the second order effects are of AIs not taking us literally is. Maybe this link??
I don't find that an unreasonable interpretation. Absent that paragraph of explained thought process, I could very well read it the agent's way. That's not a defect in the agent, that's linguistic ambiguity.
I mean humans communicate the same way. We don't interpret the words literally and neither does the LLM. We think about what one is trying to communicate to the other.
For example If you ask someone "can you tell me what time it is?", the literal answer is either "yes"/"no". If you ask an LLM that question it will tell you the time, because it understands that the user wants to know the time.
very fair! wild to think about though. It's both more human but also less.
I would say this behavior now no longer passes the Turing test for me--if I asked a human a question about code I wouldn't expect them to return the code changes; i would expect the yes/no answer.
I love this writeup--it's one of the refreshing looks into how startup innovation happens on-the-ground. We're inundated with new products and startups so often that it's easy to forget that the people working on the product are taking a bet with no promise of future payoff. In this case, it didn't work out, despite the team putting in their hard work, sweat, and clearly lots of stress.
Startups are not for the weak but the process detailed here is how we've gotten some of the most transformative and innovative products in technology. Props on attempting this unique idea; very sad that it didn't work out, but sometimes the market just can't support certain ideas!
If you know anything about tech, you will know that tech as an industry is highly deflationary--billionares use the same iPhones as you do! (in contrast, they don't drive the same cars you do)
This boils down to the fact that chip fabs have massive fixed costs and near-zero marginal costs, and these chips power all of tech. So the more chips they can produce for a given fab, the more profit they can make, meaning that companies are incentivized to sell as many products as possible for as low a price as possible.
We're supply constrained in the short-term because demand for these AI tools is so high that TSMC and other chip manufacturers can't keep up. But long term, supply/demand will equalize and tech will continue its deflationary trend. Sure, the frontier will always require the best possible chips, but AI coding is highly competitive, and competition drives price decreases. So prices may stay high right now, but it seems unlikely to me that this will stay true long-term.
All four of the author's steelmanned arguments at the end for a price decrease seem likely to come true already: competition is intense (OAI brags about how much cheaper they are compared to Claude), OAI subsidizes open-source influencers already, companies' earnings calls all call for more investment in fabs, and we're already close to saturating all of the benchmarks used for RL!
It is still exactly the same iPhone tech-wise, just with a custom "case".
I wouldn't go as far to call it "brain disease" though: in a sense, it is OK for someone well off to spend on expensive products (made by less rich), so things would equalize at least a bit.
Just like we in IT might happily pay 3% of our salary on slightly better shoes, and someone else would claim we have a "brain disease" because you can get perfectly good shoes for 5x less money.
That iPhone is 15x more (I don't know the exact price, sorry)? Same order of magnitude.
3% of your monthly salary for $200-$500 shoes (I see plenty amateur runners getting carbon sole shoes in this price range, for instance), when you could get a pair for less than $50.
Thanks you for adding a slight to your "understanding" statement. You can try figuring out the point I was making (wealth is relative, even "commoners" like us overspend from someone else's perspective).
I would say most of Europe talks and thinks in monthly (and even net after-tax salary), so hopefully that helps in the future when someone does not adapt to US gross annual salary figures for you.
This is something that is often cited as trueism, but there is no natural laws which makes this a necessary true. There is plenty of room for black swans in market laws. So much so that the term black swan is probably better known in the field of economy then any other field.
Competition may drive down the price of LLMs, however there is a greater then zero probability that it won‘t, and if it won’t, your whole counter-argument falls apart.
I can't think of any high volume/consumer electronics/computer technology that has not been driven down in price over time. So based on historical precedent, I think your "greater than zero probability" might be only a tiny bit greater than zero.
But other that comes to mind are MRI scanners, superconductors, quantum computers.
I think in general this market law is subject to selection bias. The technology which does decrease in price will become commonplace and easy to find, whereas the technology which doesn’t risks becoming obscure and maybe even removed from consumer markets.
EDIT: just to clarify, the point about black swans is that the prediction is always close to 0 probability of the existence of black swans, until we actually observe one, then the probability is suddenly exactly 1. If LLMs are a black swan for this market law, most people will assign a close to 0 probability ... until they don’t.
Oops my bad formatting - instead of "high volume/consumer electronics/computer technology" I meant to say "high volume consumer electronics/computer technology" which would have ruled out those examples. But your other point is true, there are always shortages of MRI scanners.
I was a bit confused as to how everything works until I read it in detail. Really cool tools, but I think one thing that would help in the introduction is: saying explicitly that the generated .md document is for you (the user) to read through, observe the output of the CLI call, and ensure that the output matches what you would expect.
It's basically an automated test, but at a higher abstraction level and with manual verification--using CLI tools rather than a test harness. Really great work!
Oh this is smart! Reading where Slack stores the local data in your filesystem instead of using their API/MCP (which they charge for).
Very clever; similar to OpenAI launching Atlas when websites start blocking bot requests--just build your own browser so your bot becomes an actual user.
I built this exact solution months ago, digging up Slack’s local storage but it failed because they had encryption on the db and the keys weren’t my account keys. Curious to see how they did it
Water allocation in the American West has been a mess ever since the beginning, when Prior Appropriation was decided as the way to claim water rights. Essentially, the first person to put a claim of water into "beneficial use" gets those rights.
This is why you see California with such a large share of the Colorado River's water rights, even though it "touches" the river the least: they were the earliest fast-growing state to "use" that water. And that's why you see so many water-hungry crops being grown in the West--the owners have the rights already, and to them, if they don't use it, they'll lose it.
So any agreement here needs to make a compromise between states, the federal government, prior settled law, and owners with effectively "free" water that don't want it taken away from them.
It's a complicated issue, but one step would be to force private owners of water rights to list their rights on an open market (right now some owners of water rights, like the Imperial Irrigation District can choose to never sell them). At least that way you can start the conversation somewhere.
(In fact, John Wesley Powell, namesake of Lake Powell, argued strongly against "prior appropriation" before the area was even settled, and instead argued against a collective approach to the limited and volatile amount of freshwater. He did not succeed.)
> force private owners of water rights to list their rights on an open market
You don't need to force them, they've done it for decades to the extent it is allowed. I've owned titled water rights in Nevada. They are worth something but not nearly as much as many people likely assume.
Nevada has additional complications due to the structure of the aquifers. It is difficult/impossible to move water from where it is to where it may be needed.
> complicated due to the structure of the aquifers
Guess - you're referring not to the aquifers themselves, but to the shape of the watersheds. Especially to the "water doesn't naturally flow along roller coaster tracks" topography of the https://en.wikipedia.org/wiki/Basin_and_Range_Province
Yes, the watersheds but also the prevalence of mountaintop aquifers that don't go beyond their part of the range. These can be extremely fragmented. You'll have ample water for the taking in one area and none a few kilometers away. AFAIK, you can still acquire revocable water rights to some of this without much fuss since it doesn't interact with any official watershed. It just isn't located anywhere remotely convenient or useful.
Sometimes the valleys have good wells but that isn't guaranteed due to the geology of Nevada. Lots of brine, sulfur, hydrocarbons, hot springs, etc. You never know what you are going to get and the fresh water eventually mixes into this underground.
You do not even need to force water rights owners to list their rights on the open market. They already want to, it is just illegal for them to sell their water rights (separate from their land) to entitys with more productive uses of the water.
That's not a solution, either. See William Mulholland buying up the water rights in the Owens Valley to feed Los Angles, thus turning Owens Lake into toxic dust that is costing $1 billion and counting to manage. Mulholland is long dead, and we're just getting started paying for that.
Water rights in the West are hard, and we've known that since John Wesley Powell was in charge, as a nearby commenter explained. The Colorado was divided up during an unusually wet year a long time ago, and rising demand and falling supply have only made things worse ever since.
I think this is a disconnect between people who think that large companies are static entities with established products vs. large companies that still operate like a startup and are trying to grow. When you're building your business from $0 in revenue, you don't know what will work! You try different things, you [launch over and over again](https://www.ycombinator.com/library/6i-how-to-launch-again-a...)...all in hopes of something that works, sticks, and starts to grow.
In every example here, I see OpenAI trying something new, hoping it will grow, and shutting it down after it doesn't. Sora is the pre-eminent example of this. They make news, but you don't talk about the things they launch that successfully grow!
OpenAI isn't shutting down Codex or ChatGPT, because those were launches that they did that actually worked! When you go look at the tweets and communication from OpenAI employees when ChatGPT launched, nobody was sure that it would work. But it did. And if they hadn't launched, we would have never known how valuable it was.
All that is to say...you don't know what will work until you launch. Most things fail, and it's correct to shut them down. But focusing on the products that haven't worked instead of the products that have gets you more clicks, but actually depresses innovation by making future launches less likely.
reply