Episode 14 - AI Is Colliding With the Way Teams Build Products

In this episode of Not Brothers, Ryan and Mark dig into a major shift inside modern teams: AI now lets non-technical people prototype, build, and express product ideas in ways that used to require developers, designers, and long handoff cycles.That's powerful. It's also messy.The upside is real. AI can take someone from a vague idea to an interactive prototype incredibly quickly, compressing wireframing, design, and prototyping into a tighter feedback loop. Designers, strategists, and PMs can create something tangible enough for the team to test and improve.But a slick UI can create the illusion that something is "done" when there's no real infrastructure, no secure backend, and no maintainable architecture. AI is great at making something that feels real — but often it's a house of cards no responsible team can simply deploy.The team shares what Oodle has worked through internally: oversized pull requests, skipped requirements, one-off solutions, spaghetti code, missing docs, and unclear handoffs. The big theme — AI doesn't remove the need for product thinking, it makes it more important. Ryan's example: flexible custom fields in Cortex beat hard-coding for one client. And his best metaphor: AI will bore through a concrete wall with a spoon if you ask it to, so planning still matters.Rather than banning the tools — unrealistic, since "life finds a way" — Oodle builds guardrails: project instructions, agent rules, standards, and an internal assistant, Sheldon, that asks clarifying questions and turns vague bugs into actionable reports.Handoff quality matters too. If you build something with AI, you still own it: explain the problem, document intent, set success criteria, and make review easy. The workflow is also inverting — technical people now ask non-technical teammates to carry prototypes further before development takes over.The takeaway is simple: write things down. Clear writing, intent, and documentation are becoming core skills for turning ideas into real software.Chapters00:00 — The collision between technical and non-technical teams  01:11 — AI gives non-technical people a new way to express ideas  03:12 — Why “done” is harder to define now  05:34 — When a polished UI creates the illusion of progress  07:46 — Why AI prototypes often are not deployable  11:20 — Guardrails, standards, and responsible AI workflows  15:02 — Existing products make the collision more complicated  17:22 — Product design versus one-off feature requests  20:25 — AI will dig through concrete with a spoon  22:05 — The problem with huge AI-generated pull requests  24:47 — Why smaller chunks beat massive code drops  27:57 — Compression, cleanup, and maintainability  30:28 — Better pull requests need demos, screenshots, and context  36:01 — What open source is teaching us about AI-generated code  40:02 — Why banning AI is not the answer  42:31 — How agents can improve bug reports and feedback loops  45:17 — Don’t clean up everyone else’s AI mess  49:09 — The workflow has flipped for idea people  52:59 — Writing clearly is now a core AI-era skill  55:10 — Final takeaway: write it down