But what if everything scales but what if no matter how complicated how obscure how mundane how niche what if everything I mean everything scales
Everything I mean everything doesn't scale.
Do things that don't scale, and then don't scale
51–60 of 216 posts
Re: Do things that don't scale, and then don't scale
#52It's only a startup when it can attain the hockey-stick growth. Otherwise it's just a sparkling hobby. Having a hobby is great! The biggest difference is that a startup is intended to make you a lot of money, and maybe change the way people do things, so you work on it full-time, and a hobby is intended to make your life immediately more enjoyable, and costs you money. A cargo ship and a pleasure boat have a number o…
Most businesses make enough money but not necessarily a lot of money. Slightly fewer make decent money. It's not that common for a business to make its founders "a lot" of money.
95% of businesses are classed as micro in the UK. Small and micro is 99%.
Re: Do things that don't scale, and then don't scale
#53Earlier 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…
>if you don't mind the politics of using that
what exactly are the "politics" of using DeepSeek? Feels weird to single out DeepSeek like that
Re: Do things that don't scale, and then don't scale
#54> 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…
I remember saying something on Facebook that wasn’t even that inflammatory but had a curse word and my grandma messaged me telling me to delete it. Instead I blocked grandma. Much happier just seeing grandma at holiday gatherings. I don’t think she even noticed honestly.
Re: Do things that don't scale, and then don't scale
#55Earlier quoted context omitted.
If the concern is about the inference cost -- we do have open-weight models that are getting more powerful, and hardware to run small-ish models cheap. I run agents using small local models in my MacBook.
Yes – but when people say "AI", most often they are talking about the latest OpenAI/Claude development.
Todays open models are the latest OpenAI/Claude X months ago.
Re: Do things that don't scale, and then don't scale
#56I use Claude Code to hack together a little webapp that allows me to make hex-maps for use in roleplaying games.
There are a lot of sites on the web that let you make a hex map. A lot of them are even free. Many of them have features that my little webapp doesn't have.
But mine works the way I want it. I wanted rivers and forest to be modifiers on top of the base terrain of the tile. I wanted to have a few different settlement icons. I wanted to have more variations of hills and mountains than most of the other sites do.
And if I'm ever missing a feature on this thing, I can just add it, rather than just sort of saying, "Oh well."
Because it's an app that's just for me, I don't need to worry about scale or security or monetization or anything.
It took me about an hour to two hours of my attention (spread out over two calendar days) to have the AI code it.
Re: Do things that don't scale, and then don't scale
#57This is a chord, this is another chord, this is a third chord.
Now start a band.
I dont think tech ever expected to have a "punk" phase but this is it. Do the math on how big your app has to be to make 240k a year (before tax). Now host that on something reasonable (not aws/cloud markup trash) and you can make a comfortable living. Payment processing apple/google store or square or ... CS, you dont need it any more! Accounting: software for that and a once a year with a professional. Payroll and insurance: there is a platform for that if you want to go into business with friends. Incorporation: tons of online tools will hand hold you through the process.
Your startup doesn't need to be a unicorn, it needs to pay for YOU.
Re: Do things that don't scale, and then don't scale
#58Earlier quoted context omitted.
You could run something like Facebook but in tiny shards. It would be better. And require 1/1000th of the engineering workforce.
I can see many way where you can only follow (and be followed) by 1000 people would be better in many way. An audience of 1000 isn't monetizable so the network wouldn't be poisoned by ads ("sponsored content" AKA "sponcon").
1,000 is way too many. Low hundreds max.
Re: Do things that don't scale, and then don't scale
#59> Then came the weirdness: bursts of Tor traffic, spammy signups I have a small hobby site - maybe a few hundred real users ever, and a handful of regulars. But the logs and users table are full of brute force and lousy sql injection attempts. Why does this happen? How is it economical?
Automated attacks, running on botnets or other breached servers.
Spraying 1-in-100000 chance attacks is very economical if you don't pay for compute or traffic.
Re: Do things that don't scale, and then don't scale
#60Earlier quoted context omitted.
You could run something like Facebook but in tiny shards. It would be better. And require 1/1000th of the engineering workforce.
I can see many way where you can only follow (and be followed) by 1000 people would be better in many way. An audience of 1000 isn't monetizable so the network wouldn't be poisoned by ads ("sponsored content" AKA "sponcon").
15 years later, we've reinvented Path: https://en.wikipedia.org/wiki/Path_(social_network)
> Path was a social networking-enabled photo sharing and messaging service for mobile devices that was launched on 14 November 2010. The service allowed users to share up to a total of 50 contacts with their close friends and family.