The future of software
Thom Wolf argues that software is entering a transition period where releases blur into constant updates, closed code becomes effectively readable by LLMs, and users can generate their own features with a prompt. The hard problem he raises is how to customize software without losing the upstream updates that keep it working.
- Software now evolves continuously rather than in release cycles months apart, since one company reportedly merges over 30,000 code changes a day, and dependencies ship just as fast.
- Closed source is eroding: LLMs can already reverse-engineer Electron apps and binaries with debug symbols, and a description of a UI is enough to rebuild its internals, which weakens software IP.
- Bespoke features collide with continuous updates. Forking gives you your search box but cuts you off from upstream security fixes, so Wolf expects agents to rebase personal features onto each new release.
Thom Wolf argues that software is heading into an odd in-between phase before it eventually stops meaning much at all, when anything digital can be generated on demand. Three ideas define this transition. The first is "everflux" software: instead of shipping releases months apart, code will evolve continuously. His evidence is that one company merges over 30,000 code changes a day, and that cyber capabilities and dependency chains are moving so fast that fixes have to ship constantly just to keep software from breaking as the things around it change.
The second idea is "translucent sources," a blurring line between open and closed source. Wolf points out that Claude can already take apart any Electron app, and that binaries shipped with debug symbols are easy to reverse-engineer with LLMs. So much closed software is effectively open in all but name, and keeping code secret makes less sense every month. He goes further and says software IP is starting to dissolve: he wrote his article in a text editor he had Claude vibe-code from a few videos of friends demoing features they liked, and a description of a user interface is increasingly enough for a model to rebuild the internals, whether or not they match the original.
The third is "bespoke features." If everything is effectively open, anyone can modify it, and personalization is just a prompt and a couple million tokens away. Everyone becomes their own product manager, adding a search box or a custom button in minutes. The catch is that this collides with everflux. If you fork software to add your feature, you cut yourself off from upstream security and other updates, and you take on its maintenance. Wolf's answer is that merge conflicts stop being exceptions and become constant, with agents rebasing personal features onto each new release overnight, so software starts to look less like a product and more like a species mutating over time.