@willdepue I thought Chinese firms training on NVIDIA was well known, especially in Thailand where no firm could possibly afford the data centers, executives don't care about AI, yet they had an extensive build out (plus the US banned Thailand under Trump for suspicion of helping China).
Singapore is a melting pot of 3 countries. Australia is in bred. English adopters have a classic immigrant accent influenced by their native language. West coast girls are influenced by Instagram. America is the melting pot of the world. Where did each American accent come from?
I wonder to what degree did environments influence American accents. Language is affected by your surroundings (e.g. variety of words of colors in your language). I wonder if you can create a model of immigration patterns and ethnicity data to recreate the creation of an accent.
I don't know what happened to Codex, but GPT-6 made large projects unbearable and became unfocused. After 250% of weekly usage on 20x, I asked "how much is left", and it said "substantial amount". Same project done on Claude, varying models, and it was done within 30%.
@willdepue This feels incredibly optimistic that people will stop working, but the leverage of the few firms that overwhelmingly own the most cash to invest in AI won’t let people stop working. More and more people will simply be relegated to very low paying jobs.
Is it sad to say I’ve actually run out of big software ideas. I have both coding agent subscriptions, and I was going through my old notes to see what is there to actually implement, and I just don’t have anything worth wasting electricity/compute on. Nothing/no code solution is exciting to me anymore besides my current business.
The fun was figuring out the processes' problems. Code was that fun, but agents have made it less so. It's probably why running my current startup is fun; I don't spend 80% of my time coding. I spend it on the thing that made the coding fun which was problem solving.
@ivanburazin@andyhattemer SproutOS is built on top of @daytonaio, @neondatabase and our own serverless PaaS. The goal is to allow people to create personalized, tailored software in the cheapest age of AI.
LLMs are getting cheap enough that software doesn’t have to be one-size-fits-all. I built infrastructure to make personalized apps easy to make at https://t.co/e37u0zKYfU 🧵