Sampling Parameters as Substances: An Honest Analogy A strange question that works: why does cannabis seem to make people more creative? It doesn't make anyone smarter — it changes filtering. The mind lets through associations it would normally reject. Distant ideas reach awareness more easily. That feels like creativity from the inside. Then someone asked a better question: so it's like turning up a model's temperature? Yes. And that opens a door. Every sampling parameter in a language model sits on the same trade-off — how boldly the model picks its next word — and each one has a rough equivalent in how a mind works. The one trade-off behind everything A language model doesn't "know" what to say. At every step it produces a probability for every token it could write next, then picks one. Pick the most likely word every time and the text is safe, correct, boring. It loops. Pick something unlikely often enough and the text is surprising — sometimes brilliant, sometimes nonsense. Every parameter below just moves the needle between those two poles. The LLM pharmacopoeia Max tokens is the odd one out — it's a budget, not a mood. It never touches which token gets picked, only how many are allowed. Why this matters now One caveat, because it's the most interesting part: the dial is disappearing. The newest reasoning models expose no temperature at all. The industry stopped turning the knob up and started spending the compute on thinking instead. The full article develops the analogy parameter by parameter, with a careful caveat up front: these are analogies, not neuroscience. No substance has a top_p value, and a model is not a mind with loose filters. The comparison earns its place because a person and a model both face a version of the same decision — choosing among thousands of possible next steps, and deciding how much to trust the obvious one. → Read the full post on mielony.com.