2026-01-19
Why I love open source?
Disclaimer
Summary
- A lot of problems in society exist because its leaders are not willing to sacrifice short-term gains for long-terms gains.
- This is also true for both tech startups and Big Tech companies. Most closed source software is fragile and expensive to maintain. The people hired to maintain it will keep inventing new ways to make your life worse for their own short-term gain.
- Open source software is more likely to last long-term.
- Open source software is most useful when it contains less lines of codex and targets a software niche that isn't capex-heavy. AI, 3D and video are capex-heavy, so open source alone is less useful for improving governance there.
- Open source software is useful even if it is hard to use and therefore unpopular. It puts an upper bound on how much the closed source software is allowed to hurt its users, before they switch to the open source version. Otherwise the harm can be unbounded.
- If we build ASI, this entire post no long matters.
Main
I am still a huge fan of open source, as defined by Richard Stallman and the Free Software Foundation. "free as in free speech, not free as in fee beer". Here's some reasons why. (Maybe not exhaustive, haven't thought it through.)
Build things that last
- Short-term governance is usually bad
- Software is becoming part of every industry. Hence, how software is governed has major influence on how civilisation at large is governed. (The final boss version of this is ASI, but I mean this even in worlds where ASI is not built.)
- A lot of things in society are broken because the people doing governance are not trying to build things that last, but are building short-term things. Things could be institutions, cultures or products. People like Nehru or Lincoln built things that last centuries. But most politicians are only interested in getting a 5-year election seat. Most shareholders are only interested in optimising short-term returns.
- The competitive races in Silicon Valley are the ultimate version of this. Products that are good for the user short-term but screw them over long-term are so common, that people have invented a whole jargon for it.
- Even Big Tech thinks too short-term
- You might expect this to be true of random startups with limited runway. But even Big Tech companies fail to build things that last. Most leaders of Big Tech also lack long-term vision I think.
- Their censorship rules are not consistent, and change arbitarily based on moral fashions in society, who won the US election, and so on. I would have a lot of respect for a Big Tech leader who actually defended absolute free speech. I would have some respect for a Big Tech leader that did some censorship but atleast built a consistent process (like a judiciary system) for how these rules change. I have less respect for a leader who changes rules whenever they feel like for whatever stupid reason. They end up negatively impacting users, and don't even gain much benefit from it themselves in return. I agree with Paul Graham on this point.
- Big Tech leaders too compete to make maximally addictive apps, and lock you in hard enough that you can't switch to an alternative. This is again optimising short-term user engagement over building century-long loyalty with users.
- Open source software on the other hand seems more likely to last long-term, without short-term changes.
- A bash script from 20 years back will still run today. A onedrive backup from 10 years might not.
- A good litmus test: Suppose you backed up all your files (in various file formats) and all your apps, and all their dependencies on hard drive. Then you went to prison for 20 years. When you come out, will anything on that hard drive still run?
- And now remember it's not just your personal codebases that are built like this. All of civilisation is becoming increasingly dependent on short-term fragile stuff like this. Armies of devs need to be hired just to maintain the existing stuff, forget building anything new.
- FSF and Richard Stallman and his followers atleast have long-term vision. They are willing to commit to something and stick to it for 30 years. I think this is a better way of building society.
I don't think open source goes far enough. I think you should minimise lines of code.
- Ultimately cost of running software depends on cost of paying developers to write the code, and cost of compute and storage to run the code.
- For industries that use AI, video, 3D, gaming, etc - cost of the compute and storage will remain the dominant cost. Oligopoly is expected.
- For all other industries, cost of paying the developers is the dominant cost. How the software is written determines how much it costs to write, run, maintain, etc
- I agree with Moxie Marlinspike that the developers often act as an unelected priesthood who arbitrarily govern how their software is used. This is true even if the software is being written and hosted by a non-profit, and the codebase is open source, and anyone is free to host their own copy.
- Software is not a static thing you can write once and forget about afterwards. This is unlike books or movies, for example. Software has dependencies on lots of other software around the world. The codebase needs to be updated regularly and so does the infra hosting this.
- The best way to solve this IMO is to minimise lines of code. Every additional line of code in a codebase increases the number of hours of developer time that you need to pay until the End of Time, in order to keep updating and hosting that code.
Cost-of-exit
- In practice it doesn't seem like we are getting the full hardware, software and web stacks to become open source anytime soon. This will cost too much.
- FSF gets only $1M annual in donations AFAIK, which is not close to enough money to achieve these. Other orgs get paid in similar range, unless they cooperate with Big Tech.
- Python and Nodejs teams for example are more willing to cooperate with Big Tech and hence get paid more.
- I still think open source is useful on the margin. Even if your open source version is worse, it still puts a upper bound on how much the closed version can hurt its users.
- For example, Ubuntu is worse than Mac, and the fully open source Linux distro recommended by FSF is worse than Ubuntu. But the existence of these distros places some limits on how controlling a Mac can be on its users. Most open source software can still run on a Mac, and you aren't as locked into closed versions of them.
- With mobile OSes however, there is no usable open source mobile OS as of today. As a result, even basic commands like rsync to copy files will not run on an iphone. You need pay some app governed by Apple app store to do this. Apple then gets to decide if you're allowed to copy your files or not, and in fact Apple did attempt bringing in regulations for exactly this (scan files for CP before letting users copy them). Or apple might arbitarily decide to packet sniff your network traffic without telling you. (Also a real example, google it.)
Side Note: If we build ASI, none of this matters. As of 2026-01, AI is already too capex heavy for it to matter how the software is written. It will trend towards an oligopoly colluding with the govt even if it is open source.
Side Note: Eric Raymond's excellent analysis of status and ownership in open source culture
- To explain Raymond's point in simpler terms
- In open source communities, you get a lot of status if you publish good open source work. "Good work" is subjective and is hence judged by other peers who have already accumulated status previously in open source communities.
- Whoever is the first to attempt a project and do good work (as judged by above) gets to be the lead for that project by default. If someone else forks that project and tries to compete with it, this is seen as an attack on the status of the first person. If someone forks a project, it becomes harder for the community to realise that they didn't actually do most of the work, and hence don't deserve status for it.
- Raymond claims hackers are consistently lying (to themselves and to others) as to how much of their behaviour is driven by status, as opposed to self-actualisation. He says Ayn Rand and Nietzsche have correctly analysed how often behaviours claimed to be "altruism" and "self-actualisation" are actually status-seeking behaviours. (I agree Rand and Nietzsche are good at this. I'm unsure if most hackers really disagree on his point that status is what drives them.) He says some hackers are genuinely persuing self-actualisation now, but only after they first earned status in an open source community.
Subscribe
Enter email or phone number to subscribe. You will receive atmost one update per month
Comment
Enter comment