These days you can build all sorts of software by asking an AI for it. It's like a genie that gives you infinite wishes but only sometimes grants them, and it's not always obvious if your wish has been granted or not. Still, the AI genie gets the job done _often enough_ to be interesting, even useful. If you're a software engineer you can, in theory, build any software you desire to suit your own needs. In practice this has historically been very costly in both time and opportunity cost (you could have been earning good money). The time continues to shrink though, so I hypothesize that we're going to hit some sort of inflection point where A LOT of people can build their own custom software. For example, here are some things I've built that I actually use: - Heyo - Rapid voice transcription for mac, running locally - Heyo Mobile - Voice journaling for iOS (replaced a paid subscription to voicepal) - Full Text Tabs Forever - Search the entire text of documents you've browsed online. - Vibesidian - Write Obsidian plugins with AI... directly within Obsidian. - Blog publishing tools - I write my blog in my notes app and a script handles updating links to make it act like a dynamic site. - this is pretty simple stuff, but i didn't even have to write it. i just asked the AI genie and it was done. The point here is not that any of these are all that useful (although I find them to be), it's that it's now relatively cheap to build your own software, and the trend continues to get cheaper. # Implications (Speculation) In the future, software will have to: - have a network effect - have interesting data - solve a complex enough problem that AI can't write a solution for it - I would put apps like Figma into this category Not a complete list, but my point here is that software cannot win simply be existing. For simple enough software people won't pay for what they could build themself for a few dollars worth of LLM credits.