Why Human?

I suppose it’s worth clarifying the purpose of the disclaimer “by human, for human”, especially in writing about technology from a generally positive perspective. Why have a strict disclaimer against AI writing if I think AI has the ability to be a powerful and useful tool? AI does many things well, but it does not write for humans. The reason for this is a dependency on scaffolding and boilerplate. (As of the end of 2025 models at least.) In order to effectively complete their chain of thought, models have to put a great deal of language around the core information. While this has massive implications for contract and boilerplate document review, it leads to an unsatisfying information density for human readers. That doesn’t have to mean anything to you, but implication is that the best way to read AI writing is to have an AI summarize it for you.

Here is a common sequence of events. Someone has a thought, and they form a prompt to expand that thought. The LLM generates a wall of text that only barely expands the prompt despite being many times as long. The first person then posts that long text somewhere. Now a second person, not bothering to read something of such low information density, asks another LLM to summarize it. The LLM returns an only slightly more bloated version of the original prompt. That’s the ideal scenario where nothing is lost in translation. What the first person should have just written was the prompt, but there are many financial and cultural incentives against that.

This sets up what, if you’d like to flatter me, we can call the Condor Dilemma. If you know a block of writing is generated by an LLM, you are wasting time by not having an LLM summarizing it. If it is written (well) by a human, an LLM summarization is going to lose too many of the deeper points and constructions.

I write for my own interest, and not under any financial compulsions towards volume. This allows me to commit to human writing. Given that, I want to make it as clear as possible to my readers that they can read what I have to say, hopefully enjoy adequate prose, and not experience a creeping paranoia they are reading machine boilerplate. The best way to ensure the reader can read free of that suspicion is through the by human for human disclaimer. I suspect that some form of human disclaimer will become default in human centric spaces.