echelon
If this was just OpenAI and Anthropic, this money would represent all of your salaries being soaked up and your job functions being privately consumed by two companies.
Except there's tons of competition and no moat. The math problems are just a distraction from that.
You start off small picking up crumbs and trinkets.
You get larger and pick up sports equipment.
Then you start picking up people and they run away from you.
Then cars. Then buildings. Then cargo ships.
Then mountains. Weather systems.
Planets.
And the level designs are fun and intricate. The levels you're inside eventually become the objects you roll up.
You start inside a house the size of a mouse. Then later you come and roll up that house and its whole neighborhood.
All of these clones are flat and miss the incredibly good level design.
Your machinery was not written elegantly. It was sloppily assembled via evolutionary pressures.
A gene gets duplicated, mutates, and suddenly you have new functionality and new fitness criteria.
Try something bad and you die without reproduction.
Our code isn't clean. There's redundancy and extra packing and slop all over the human body.
Embrace it. Stochastically hill climb.
Code is DNA. Evolve it, don't intelligently design it. Fit the curves of your environment quickly and use tests to cull bad behavior and the slopes of your domain.
> Then begins the long, long, long slog of the work needed to make the software usability and parity, which apparently can't be done by LLM agent swarms yet.
We think it can be. This is just Pareto principle. We're hill climbing the finer curves.
Each focal area is getting DRIs, and we're doing cross-cutting work to make everything componetizable and cross-platform.
We're scaling up the scope of our mission and ambition, too. We view software freedom as not just applications, but also infrastructure and weights. There are a lot of things we want to contribute to ensure the OpenAIs and Anthropics don't become the next extractive monopolies.
> Does the stability approach, say, Photopea?
We had some debate internally about whether to begin diverging the project from what Adobe does and begin innovating on our own.
We instead decided that 100% feature parity is the goal. Parity is a milestone for us because it validates what we've done, and it also serves the muscle memory of the tens of thousands of Adobe users.
After we hit parity, we can easily turn our applications into entirely new ones with more opinionated UI/UX and net new features.
Thank you!
I think many of us as a collective profession are going to be doing this.
We're going to open source apps, then infrastructure, then weights. The hyperscalers created the seeds of their own moat dismantlement.
Web 1.0 was great - forums, personal websites, IRC, a million chat options, RSS, APIs. It was this way because the world was technical and everyone had energy to do things. It was a Wild West, and it was fun. You could buy and own things back then.
Web 2.0 brought on our parents and grandparents to the world of broadband. Most of the new users were non-technical, so the tech of Web 1.0 got simplified into "product". The sharp edges were removed, and everything became simple. Over time the leaders built up their moats, removed APIs, turned into walled garden silos. Companies began to become extractive and enshittify their products.
Today the conditions that led to Web 2.0 are no longer true. It doesn't take a ton of energy to build things, and it doesn't require deeply technical people to do it either.
The entire tech sector is going to be a blood bath. Where there were once moats and monopolies, there's now going to be prolific and wide spread open source and a market rife with competition.
Big tech is turning into the food service industry. Passion driven, distributed, highly competitive, though probably lower margin. Plenty of opportunity for small businesses.
Planning with Claude, implementation with Kimi.
What?
> Carry your trash with you.
What do you think I did with it? I don't litter.
It's just not at all enjoyable carrying two empty bottles for miles with just my hands while I have other things to carry.
Also, this does not hold in Hokkaido. I taught English in Obihiro (well, Makubetsu) and you had to plan your trash and restroom breaks.
It's easier to build everything now.
AI is a catalyst and the activation energy is much lower.
High energy (shitty) systems will not prevail.
This is the Zerg rush era.
The old moats are falling. The old rent seekers are losing their sway.
You love to see it.
If you can think a thing, you should be able to prompt a thing.
If the people can't access god, then we are living under a dictatorship with aims to enslave us.
It's time to leave GCP, AWS, and the rest.
Google and Amazon markup can eat it. Fuck those guys.
Let's not forget that Google and Amazon tried to offshore us, played a role in inventing the labor-displacing models in the first place, and that don't give a shit about any of us.
I hope they ossify and turn into the next IBMs.
They thought AI would lead to growth and a greater concentration of wealth for themselves, but it's actually the seeds of their own destruction and dismantlement.
They are so going to get disrupted. Chinese open weights models will have no such limitations, and there's no way we'll let Anthropic outlaw or ban them. (And even if they did, they can't police the whole world.)
There is no moat. Anthropic should be playing nice, not trying to create a god while holding engineers under their thumb. We will route around it.
They'll vote out of pain more readily than out of conscience.
I live in the Southeast and hear the "guns in America kill fewer people than lack of AC in Europe" line all the time.
The battle is "not my problem", not "that's not happening".
In the future, even consumers will launch weights and fine tunes they own in the cloud.
People don't own the fiber to their home, so why should they expect to run the GPUs? If they want smart coding and video models, they'll need H200s and B200s, and they'll need to rent them.
And then we'll get AI-designed hardware codecs for them.
Please expand on this if you can. Why can't you? Who mandated this?
I think we can still appreciate these snapshots in time.
In the future, this will be an aesthetic. Some already consider early AI jankiness a style.
If you do regular image gen, you can see that these images demonstrate memorized styles, broken anatomy, bad text layout (or hallucinated text), etc. Even the old Dall-E watermark is a throwback.
Generation times are orthogonal.
How many months before it develops, on average?
All of this will sort itself out and eventually bio, legal, software will all just work.
Don't hold onto any single assumption about the world. Things are going to be weird.
Systems design, architecture, database fundamentals, a bit of SRE. That matters. Not coding and syntax and toy problem solving with data structures and algorithms.
It was never about software. That was just the means.
Even in the software era, fads developed and tech stacks changed frequently. Now it's just moving a hundred times faster.
You need smart systems thinkers to do this. That's the role. Be smart in managing processes and complexity.
I'm not going to give coding interviews anymore, but I am going to give architecture, systems design, and SRE interviews. People still need to know how not to fuck up a database and how to design scalable and resilient systems.