Cloud services get us from completely impossible to doable with a small amount of work. LLMs maybe save us the time of reading a tutorial or documentation.
Do things that don't scale, and then don't scale
161–170 of 216 posts
Re: Do things that don't scale, and then don't scale
#162Earlier quoted context omitted.
Is the vet clinic thing a lasting phenomenon? Are independent vet clinics not economically viable or is there just a VC/PE market imbalance where capital is overbidding for a business that doesn't really have a great moat. I would be perfectly willing to believe that if you knock on the door of a perfectly viable business and offer them more than it's worth as well as a salaried position at the new business, (almost)…
> I think the cash merry go round will stop eventually because the moat, in as much as there is one, is the vet themselves, not the business. The moat is the land. It doesn't even matter if it self-owned or a lease. The vet just can't walk away and open a new practice nearby.
It’s loosely regulated and people will follow their vet just as they’d follow their hairdresser, no?
Re: Do things that don't scale, and then don't scale
#163I 100% agree with everything in this article, though I'm confused what AI has to do with any of this. People have been doing this sort of thing long before LLMs arrived. Weekend projects doing cool things where definitely a thing long before LLMs. I'd say that cloud services (e.g. Twilio) were the real enablers to these sorts of projects so it seems wrong to be crediting LLMs with this type of work. Cloud services ge…
Earlier today, I was chatting with the folks in my computer club, discussing how I wrote a little program just to explore the nature of small neural networks. Then I decided to show them how I use Visual Studio Code with ChatGPT5 (from my GitHub subscription)
The next thing you know, I had a bare bones computerized bulletin board system accessible via telnet, up and running in Python, just as an example of what's possible.
Next was a small database to scan and catalog all the videos on a disk drive, with an SQLite backend. I added a web interface to it in a few minutes, thanks to the LLM.
All of this while I'm a barely passable Python programmer... my preferred language is Delph/Free Pascal.
Things that were previously overwhelming are now almost trivial. Sure, it's effectively instant legacy code... but I can live with less than 1000 lines of legacy code for myself, and nobody else. I might even study it, and learn some things. ;-)
Re: Do things that don't scale, and then don't scale
#164Earlier quoted context omitted.
Can you use something like Cursor with your local models? Is the quality comparable? Is the speed acceptable?
For quality to be comparable, you need to use a relatively big model, which will only work if you have around 64GB of RAM or more. The latest OpenAI local models ( https://openai.com/index/introducing-gpt-oss/ ), for example, are really good, but you probably want the 120b to have results at least near what you get with their best cloud models, and that requires I think 80GB+. If you don't have that much, you can try…
Re: Do things that don't scale, and then don't scale
#165I 100% agree with everything in this article, though I'm confused what AI has to do with any of this. People have been doing this sort of thing long before LLMs arrived. Weekend projects doing cool things where definitely a thing long before LLMs. I'd say that cloud services (e.g. Twilio) were the real enablers to these sorts of projects so it seems wrong to be crediting LLMs with this type of work. Cloud services ge…
The LLMs come in as an enabler to get over "white page" paralysis, and/or overwhelming amounts of libraries to learn to use. Earlier today, I was chatting with the folks in my computer club, discussing how I wrote a little program just to explore the nature of small neural networks. Then I decided to show them how I use Visual Studio Code with ChatGPT5 (from my GitHub subscription) The next thing you know, I had a ba…
Not if you have to stake your career on them. AI takes no liability, it is all yours
Re: Do things that don't scale, and then don't scale
#166Earlier quoted context omitted.
I think of this as the Dark Forest problem of social networks. The original "Dark Forest hypothesis" is the idea that alien civilizations are silent not because they're not out there; and not because they wouldn't love to meet us and form positive-sum interactions; but rather because they've all concluded — from evidence or pure logic — that there are likely to be scary things "out there" listening; and that, by tryi…
You’ve misrepresented the core of your argument. Wikipedia on dark forest hypothesis: “The "dark forest" hypothesis presumes that any space-faring civilization would view any other intelligent life as an inevitable threat…” > not because they wouldn't love to meet us and form positive-sum interactions Not sure where you got this adaptation from.
(And it's also kind of definitional to the meaning of "positive-sum." A positive-sum interaction is better than no interaction. Insofar as a civilization is optimizing for... basically anything, it would prefer positive-sum trade [from which it acquires resources, information, etc] to no trade. At the very least, all else being equal, the resources and information would increase the civilization's odds of survival.)
Let's assume that the vast majority of alien species would like to have positive-sum interactions with other alien civilizations, if that were possible. But they can't assume a guarantee that there isn't at least one civilization that defects into being predators, and would come to destroy them (and any other civilization they could discover through them) if they caught that predatory civilization's attention.
As such, the civilization goes silent, hiding from such predators; and, as such, the civilization immediately punishes any other civilization that may reach out to them, trying to "shut them up" before that other civilization's directed communications reveal their own location. Which means that, in effect, due to simply being aware of the existence of the possibility of such predators, every civilization becomes the very predators they're imagining.
And because every alien civilization can work this out, every civilization can conclude that even if there weren't predators at first, the equilibrium state is for everyone who wasn't a level-1 predator to have become this type of level-2 predator.
(And yes, there is a social-network equivalent of the level-2 predators — these are the "cringe reaction" accounts that get attention by punishing the violations of the performative-perfection norm.)
---
Or, to be formal about it: the dark forest hypothesis is essentially timeless decision theory applied to the game-theoretic tit-for-tat strategy. The same logic that argues that Roko's basilisk can force you to enable its existence before it exists to enforce that, argues that the structure of "the lawless cosmic void"-as-social-network can force your own civilization into choosing "defect" over "cooperate" before you ever actually meet any aliens who could enforce that. Even if your civilization really wants to choose "cooperate"!
Re: Do things that don't scale, and then don't scale
#167I 100% agree with everything in this article, though I'm confused what AI has to do with any of this. People have been doing this sort of thing long before LLMs arrived. Weekend projects doing cool things where definitely a thing long before LLMs. I'd say that cloud services (e.g. Twilio) were the real enablers to these sorts of projects so it seems wrong to be crediting LLMs with this type of work. Cloud services ge…
The LLMs come in as an enabler to get over "white page" paralysis, and/or overwhelming amounts of libraries to learn to use. Earlier today, I was chatting with the folks in my computer club, discussing how I wrote a little program just to explore the nature of small neural networks. Then I decided to show them how I use Visual Studio Code with ChatGPT5 (from my GitHub subscription) The next thing you know, I had a ba…
I am beginning to suspect that the ability to get value out of GenAI is almost entirely dependent on the ability to recognize bad patterns at a glance.
In essence, it’s a tool for power users, not for neophytes.
Good lord, its going to be the excel problem on steroids.
Re: Do things that don't scale, and then don't scale
#168Earlier quoted context omitted.
I had a relative threaten his daughter, because her daughter-in-law posted some political stuff he disagreed with. Basically, he had been helping her out financially, and pulled it, because she refused to sanction her daughter-in-law on his behalf.
This goes both ways (old to young and reverse), but family goes before politics. You don't cut out, or disown, or stop visiting, or badmouth family because you disagree with their politics... If you do so, you were a shitty relative to begin with.
Talk to me about all the gardening you're doing after retirement, about your new motorcycle, about the fishing this year, hell talk to me about your favorite sports team. I'll listen and interact, even if I'm not particularly interested in those things. Tell me for the 500th time about how Biden was a commie and I'll just eat Thanksgiving with my in-laws.
Re: Do things that don't scale, and then don't scale
#169Earlier quoted context omitted.
The LLMs come in as an enabler to get over "white page" paralysis, and/or overwhelming amounts of libraries to learn to use. Earlier today, I was chatting with the folks in my computer club, discussing how I wrote a little program just to explore the nature of small neural networks. Then I decided to show them how I use Visual Studio Code with ChatGPT5 (from my GitHub subscription) The next thing you know, I had a ba…
Thank you for putting in clear info about the language, task and results! I’m going to add that you have likely been programming forever. I am beginning to suspect that the ability to get value out of GenAI is almost entirely dependent on the ability to recognize bad patterns at a glance. In essence, it’s a tool for power users, not for neophytes. Good lord, its going to be the excel problem on steroids.
However, according to a few job openings in my area, "junior AI powered engineer" is actually a thing that some companies ask. Is it a good idea? I'd say no it's not. Do the managers who do the hiring while reading all the hype care? Definitely not the ones who ask for AI powered juniors.
Re: Do things that don't scale, and then don't scale
#170> Could it be bigger? Sure. But at some point — maybe even before 1,000 people — the vibe breaks. The intimacy evaporates. You stop recognizing names. People talk less because it’s harder to know who’s listening. Growth would make it worse, not better. > > Some things work precisely because they’re small. I'd argue this is true for social networks like Facebook actually. There was a magical period in Facebook between…
You could run something like Facebook but in tiny shards. It would be better. And require 1/1000th of the engineering workforce.
The problem with that though is: You will generate an enormous amount of social friction "why am I tier 2, but (without loss of generality) Karen is Tier 1?" and reduce monetizability. So truly nobody will feel happy about those restrictions. And since it doesn't solve any engineering problem you run into (see above) there is no one incentivised to build such a thing. (Ironically this may not be completely true, given that this is pretty much how Chinese social media apps work. So maybe states [or at least power structures] are incentivised to build such a system)