| Someone called for me!? It's deeply unfashionable to believe in human potential, but, well, I do. I am one of the last working (human) hackers. See: https://paulgraham.com/hp.html. My reasoning in not using AI (at all) is many-fold. First, I am a powerful learning model, the most powerful one I have access to. This model is so powerful that to reach its true potential I must continually feed it huge amounts of learning experiences and training data. I must seek out other unique perspectives and try to understand them. There is virtually no end to the amount of data this model in me can process and integrate using tools like "philosophy" and "logic". Second, it seems to me that models have shown us that the human limitation is not our ability to think, but how fast we can type. The model-first people think that means that we must find a way to offload all our ideas to agents, since agents are on the other side of the low-bandwidth interface that is the keyboard. I just think we need to arm the human with more expressive tools: an instrument, not an assistant. Third, most people have stopped believing in the potential of great software. They seem to think that the tools we have now are the best that can ever be made, so by their reasoning there wouldn't be much point to getting good at building new things when instead you could learn to copy what already exists. If you're interested in the tools we're making to mint a whole new generation of (human) hackers, I would urge you to come check out the BABLR Discord! It's where we work on our tools, talk philosophy, and boot anyone who tries to sell us their AI shit. https://discord.gg/NfMNyYN6cX Every line of this code in the BABLR ecosystem was written by human hands too: https://github.com/orgs/bablr-lang/. In five years of full-time OSS work I've built a streaming regex engine, the world's most powerful streaming parser, a language-agnostic AST/CST format, and a data serialization language that will eventually replace both XML and HTML: https://docs.bablr.org/guides/cstml |