Practical Philosophy of Systems

Last time we examined what happens when the person improving a system is too far from the consequences to know when improvement becomes damage. This month, we bring that question into the most immediate context we have: the tools we use every day to think, decide, and build.

Large language models are now embedded in how we write, how we design systems, and how we make decisions at work. The promise is leverage: faster output, broader knowledge, reduced friction. But every time we accept a generated answer, delegate a decision, or let a model draft what we would have written ourselves, something is happening to the feedback loop. The model has no stake in whether the output was actually right. It will not be there when the architecture it suggested proves unmanageable, when the email it drafted lands wrong, or when the decision it recommended turns out to be exactly the kind of thing that requires a human who would feel the consequence.

To understand this, we will look at: Under what conditions does delegating thinking to an LLM genuinely extend and improve human judgment, and when does it replace it with AI slop?

This is not a question about whether AI is good or bad. It is a systems question. The same one we have been asking all along: who bears the consequence, who holds the authority, and what happens to the feedback loop when the decision travels through something that has no skin in the game?

We are exploring:

• What are the actual conditions under which LLMs extend rather than replace judgment?
• What does a systems architect give up when they delegate architectural decisions to a model, and what do they gain?
• How do we stay the authors of our own thinking when the tool is increasingly better at producing the appearance of thought than we are?

This is not about rejecting the tools. It is about understanding what we are transferring when we use them, and whether we can afford to.

One ask: if your plans change, please update your RSVP, there is a waitlist, and a timely update genuinely gives someone else the chance to join.