Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.
Start free trial →Product Impact Podcast | Secrets to unlocking the value of AI · Product & Design · October 2026
A host had described a guest who said good AI design output means scripting rules for every component, so context becomes endless. Paz answers that piling on context is the wrong instinct; choosing and organising it is what drives output quality.
know it sounds ridiculous, but that's how I feel when they're sitting next to me and they're like, do you have sangria? And of course they all have sangria now because every tourist order is sangria.
…have different character lengths, different intonations of content, different intent and situational use of content. So the thing about context is effectively an infinite depth of rules. But the problem is usually we're so laissez-faire about them that we don't even recognize that they're rules.
Yeah, and we also assume that giving more context will be better. So, you know, you have a specific agent doing, I don't know, whatever, like looking for like health insurance policies. And you feel like if I give it more, it will be better. And that's not really how it works. You have to give it the right one at the right time. So organizing the context and making sure it's super clean, neat, and proper is like critical for having a good output on the other end. So it doesn't surprise me that they go so like nitty-gritty because that's what you had to do and it's a lot of work.
But this is where it gets tricky, right? Because if you overload the context window, the agent gets confused.
So first I wouldn't put everything in the context window. And those are also design decisions, by the way. You have to think: okay, is this something, is this instruction something maybe reusable that you don't need all the time, but you're going to use from time to time? So that will go into the…
You lived in the Bay Area for a really long time around a lot of, let's call them AI pill people.