ai

My stance about AI is that they are tools that can be a significant aid for human activities. They might be environmentally expensive, but I'm hopeful this will improve over time, so I try not to worry about this too much. I try to stay optimistic and constructive; over time I have learned that negativity is generally not that helpful in life.

My belief is that these tools elevate the overall quality of art, science, engineering, and social life in general. They will help us achieve feats never possible before.

I also believe that the concept of intellectual monopoly (copyrights and patents) will be (and should be) incrementally eliminated. I wrote about this in @/intmon and @/progress but I plan to explore this again in a later post, so I won't go into this here.

AI slop doesn't worry me too much either. I expect that chain-of-trust tooling (like https://codeberg.org/robida/human.json) will reduce the quality concerns over time.

At the same time I try to avoid genai for generating the content. All the posts and code on this site are hand-entered by me. However, I do ask AI for review and hand-incorporate the findings I like. I wrote about this in @/aicritic. I created a custom CLI tool for this but I'll write about my tooling in a later post. I use various models of Gemini as my AI.

This significantly alters my style though. Just take any pre-2025 post (recognize them by the all-lowercase style) for comparison. 2025-06-16 is the exact date when my posts started to be AI-assisted.

The AI always complains about my clunky, incoherent, stream-of-consciousness style of writing so now I smooth things out a bit. The newer, AI-assisted posts should be easier to read. They have less personality (less sarcastic, less weird) but I hope the message is clearer at least. I even had a fully drafted post that I completely dropped after the AI explained to me that it made no sense and I expect this will happen again in the future.

I don't use AI to generate post ideas or write drafts at all. I have a huge backlog of stuff to process (see the big graph in @/mementomori) so I don't have the need to do so.

I develop code the same way: all code is written by me without distractions but I incorporate AI feedback during the review phase. No autocompletion or other assistance for me either. I wrote about this in @/autocompletion and I still work that way. Even path completion is disabled in my shell even though I use terminals as my interface to my computer.

I quite like https://codeberg.org/robida/human.json but I'm not adopting it yet on my site. At the time of writing v0.1.1 is still too basic, the extension is inefficient, and the protocol has too little adoption. I'll wait and reconsider later.

#c1 by aqisjqif-guest on 2026-05-11

The newer, AI-assisted posts should be easier to read. They have less personality (less sarcastic, less weird) but I hope the message is clearer at least.

I'm here for the sarcasm and weirdness! Don't change it!