My dev machine is an off the shelf Acer aspire laptop that I bought used, it comes with a 4GB RAM, I also added 4GB of swap memory. I use it to develop fullstack applications for servers, browsers and android. My stack is Rust, Javascript, Dart/Flutter, Postgresql and running Debian 13.
It works great, I mainly use 2 applications , Firefox and urxvt, using emacs -nw to write code, no LSP or code completion because they use too much RAM. I also use i3wm which very lightweight.
The only issue with it is that some keys on the keyboard don't work, so I mapped them to different keys, for example ctrl is mapped to caps lock, 0 is mapped to 3 on the numpad, enter is mapped to right shift etc. Because of this I can't type on other peoples computers, I would need to run my xmodmap script before I do that.
> The irony is rich that AI has turned Stallman's goal of software
freedom on its head. Having your computer do whatever your
imagination can think up -- that's real freedom.
The irony is that Stallman promoted free software and allot of code got published online for free, and now LLMs have trained on this code. Without Stallman there will be no vibe coding.
I don’t find that ironic. As the post says, AI gives people the ultimate freedom to materialize Stallmans world where everyone can modify the software they use any way they want. The fact that his movement helped that become a reality seems the opposite of irony.
> Coding agents often default to older patterns because LLM training data contains vast amounts of legacy code. This often leads them to generate bloated JavaScript for tasks that now have native, high-performance web platform solutions.
The main problem I see is that it will be extremely hard for new frameworks, languages to take off if everyone is using AI to code. It can lead to a future where all new programming languages and frameworks will be developed by LLM providers because only they, can train them with synthetic data.
I don't know about this concern for new frameworks
With LLMs improving, they will might need less example code to work with new frameworks decently
With LLM usage becoming ubiquitous for coding, framework developers will likely start making lots of curated, high-quality synthetic examples specifically for LLM training data usage as their main documentation effort
Open source should not be a threat just like Linux isn't really a threat to Mac and Windows. It all depends on who will get the network effects first (if there even is network effects in LLMs).
People pay for windows because of its strong network effects eg Microsoft Office and countless software that run on windows. Even though Linux is free most businesses today don't use it on the desktop.
Opensource also means do it yourself and most businesses don't have time or expertise to do that, but this very issue creates a market for 3rd party providers that use open models to compete against Anthropic. So hardware and compute availability becomes somewhat of a moat.
The business feels like like web hosting before cloud computing. Web hosting in itself does not really have a moat, but than again AWS appeared with their abstractions created lock-in and network effects and disrupted the hosting industry.
The network effect is way weaker for LLM's. Switching from AWS to GCP or Azure is at least a multi-month project for an engineering org, if not completely operationally impossible. Switching LLM providers is at worst asking your engineers to install a new coding agent client, and at best switching a URL and API token in your config file.
Definitionally, from the product they claim to sell, you could just ask Claude to help you migrate to an open source model (if you have the hardware or if you are ok with a subscription) with very little difficulty in the scope of a single conversation. The product they're selling is capable of helping you stop using it. And there's very little friction to use other products too, unlike linux vs mac vs windows, where program compatibility is not as trivial (although with WINE, Rosetta, and WSL, it is actually decently easy nowadays). In this case, you're just calling out to an API with a harness. And even their harness is capable of using open source models.
I feel like it's different from all other cases because it's the first product which can literally obsolete itself, lol. It can show you the door and walk you through it if you ask it to.
Regarding moats: it's been said by everyone else, but it doesn't seem like Anthropic has a moat. Everyone can easily switch from Claude to Codex, OpenCode, Pi, or any of the other zillions of harnesses. They can even have Claude do the migration for them if they want and it will oblige – that's what I did when I canceled my Claude subscription.
The idea that hardware and compute is their moat is interesting, but open/local models become more capable each week. As I write this, I have 3 instances of some no-name model working on 3 separate worktrees on my M1 Mac Studio. It was viable for me ages ago, and I'm not an AI enthusiast (I just happened to have bought a Mac several years ago that can run these models). I haven't paid Anthropic or OpenAI a cent since the early part of 2026.
But if you’re statisfied with just that throughput and model capability that a M1 can provide, than I don’t think you are a well monetizable client anyway? I agree with the no moat as I use a wide variety of models but to only use my M5 128 would definitely not statisfy my needs, so on top of that I pay.
I'm satisfied with it for the most part, which isn't to say that local models can't be slow or don't do dumb shit that show their comparative lack of intelligence sometimes. But when I need speed or intelligence, I just press ctrl+p in my harness and spend a buck on OpenRouter.
It's my belief that we will (or probably have) hit a point where the majority of us receive diminishing returns from the sota models for the work we do. I don't need Astra or Mythos just to write some crud apps or refactor react to svelte, for example.
I'm also local-only in my current development process, and have a couple local agents running right now. The problem with local right now is the price has gone through the roof. Even 3-6 months ago I would have said it's viable for a lot of people, but even getting the RAM for a local inference machine is a big investment for a lot of people trying to dip their toes into the space these days.
It's not about communism vs a free market, but about a change even more radical than that: leaving aside "made by a human" labeling, and even if there were no other changes to power structures or economies and no AI doom, what does a world look like if the cost of having something done falls to a few cents of electricity and its burden? Growing crops still takes a lot of land and other resources...
> what does a world look like if the cost of having something done falls to a few cents of electricity and its burden?
There will be an oversupply, demand will be there too, but no one can afford to buy since most jobs have been replaced by AI. So the price of goods and services should go down, or some form of UBI will have to be implemented which is just socialism for the masses. Its either that or total anarchy (also socialism).
> If humans remain in control of it, then how can it possibly lead to anything other than extreme oligarchy?
Depends on the distribution of such AI. For example, if everyone on Earth has their own, personal superhuman AI, it would not lead to extreme oligarchy. If only a handful of humans control superhuman AI, it would.
That is the classic problem of decentralized networks, if it is big enough there will be spam at which point you will need moderators to sieve it out. But I think AI can also be helpful here instead of having centralized moderators we can have AI (hopefully self-hosted/decentralize) moderate your network graph for bots/spam.
It works great, I mainly use 2 applications , Firefox and urxvt, using emacs -nw to write code, no LSP or code completion because they use too much RAM. I also use i3wm which very lightweight.
The only issue with it is that some keys on the keyboard don't work, so I mapped them to different keys, for example ctrl is mapped to caps lock, 0 is mapped to 3 on the numpad, enter is mapped to right shift etc. Because of this I can't type on other peoples computers, I would need to run my xmodmap script before I do that.
reply