Uploaded August 2026 | Updated November 2026, 3 hours ago
π Theo put out a video saying you should turn off Claude Code's memory, and it triggered me enough to record a response. My argument: the real problem with all AI is the gap between what's in your head and what the AI knows you want. I walk through how LifeOS closes that gap with the Ideal State Artifact, one document that works as both the build spec and the testing harness, and show the skill system and blogging pipeline that run on it.
π¬ This is a response to Theo's video β he's actually my favorite AI creator right now and I watch everything he puts out, which is exactly why I cared enough to push back on this one:
π₯ Turn off Claude Code's Memory: https://ul.live/yttheomemory
π₯ Theo's channel: https://ul.live/yttheo
π₯ I also talk about Matt Pocock's work a lot in this one β he's doing amazing stuff on the development side, and his Wayfinder and GrillMe skills are excellent (I've incorporated ideas from both):
π₯ Matt's channel: https://ul.live/ytmattpocock
π AI Hero: https://ul.live/ytaihero
π§ What I talk about:
π They Don't Have the Plot: Why specs, PRDs, and Ralph loops fail β when you break the work into pieces, the AI can't see what you're actually trying to build.
π The Ideal State Artifact: One document holding the stated goal, the ideal state, and individual testable claims. The spec is also the testing harness.
β³ Come Back in Six Months: Pull up the ISA for a production app like Surface or Human3 and keep working β no digging through old sessions and scattered plan files.
π Writer's Blindness: Why you can't articulate everything you know, and how the Discuss skill pulls it out of your head and into the artifact.
π§° The Skill System: Concepts, workflows, and deterministic CLI tools β about 150 small skills that encode exactly how I want things done, from blogging to booking a restaurant.
π§ Memory Systems Reframed: Theo is right that auto-written memory files decay, and wrong about memory. The whole harness is the memory system β it merges your brain with your AI.
π Everything I've written on this idea:
π The thesis in one site β Define It In Text:
https://ul.live/ytdiit
π₯ Announcing PAI 5.0 (the LifeOS release this video shows):
https://ul.live/ytpai5
π Building a Personal AI Infrastructure:
https://ul.live/ytpaiinfra
π Bitter Lesson Engineering:
https://ul.live/ytble
π AI is Mostly Prompting:
https://ul.live/ytmostlyprompting
π How to Talk to AI:
https://ul.live/yttalktoai
π How to Rate the AI We're All Chasing:
https://ul.live/ytrateai
π Claude Code is the Biggest AI Development Since ChatGPT:
https://ul.live/ytccagi
π We're All Building a Single Digital Assistant:
https://ul.live/ytsingleda
π Stay connected
π¬ Subscribe to the newsletter at:
https://ul.live/ytnl
π¬ Join the UL community at:
https://ul.live/ytcommunity
π£ Follow on X:
https://ul.live/ytx
πΌ Follow on LinkedIn:
https://ul.live/ytli
π Theo put out a video saying you should turn off Claude Code's memory, and it triggered me enough to record a response. My argument: the real problem with all AI is the gap between what's in your head and what the AI knows you want. I walk through how LifeOS closes that gap with the Ideal State Artifact, one document that works as both the build spec and the testing harness, and show the skill system and blogging pipeline that run on it.
π¬ This is a response to Theo's video β he's actually my favorite AI creator right now and I watch everything he puts out, which is exactly why I cared enough to push back on this one:
π₯ Turn off Claude Code's Memory: https://ul.live/yttheomemory
π₯ Theo's channel: https://ul.live/yttheo
π₯ I also talk about Matt Pocock's work a lot in this one β he's doing amazing stuff on the development side, and his Wayfinder and GrillMe skills are excellent (I've incorporated ideas from both):
π₯ Matt's channel: https://ul.live/ytmattpocock
π AI Hero: https://ul.live/ytaihero
π§ What I talk about:
π They Don't Have the Plot: Why specs, PRDs, and Ralph loops fail β when you break the work into pieces, the AI can't see what you're actually trying to build.
π The Ideal State Artifact: One document holding the stated goal, the ideal state, and individual testable claims. The spec is also the testing harness.
β³ Come Back in Six Months: Pull up the ISA for a production app like Surface or Human3 and keep working β no digging through old sessions and scattered plan files.
π Writer's Blindness: Why you can't articulate everything you know, and how the Discuss skill pulls it out of your head and into the artifact.
π§° The Skill System: Concepts, workflows, and deterministic CLI tools β about 150 small skills that encode exactly how I want things done, from blogging to booking a restaurant.
π§ Memory Systems Reframed: Theo is right that auto-written memory files decay, and wrong about memory. The whole harness is the memory system β it merges your brain with your AI.
π Everything I've written on this idea:
π The thesis in one site β Define It In Text:
https://ul.live/ytdiit
π₯ Announcing PAI 5.0 (the LifeOS release this video shows):
https://ul.live/ytpai5
π Building a Personal AI Infrastructure:
https://ul.live/ytpaiinfra
π Bitter Lesson Engineering:
https://ul.live/ytble
π AI is Mostly Prompting:
https://ul.live/ytmostlyprompting
π How to Talk to AI:
https://ul.live/yttalktoai
π How to Rate the AI We're All Chasing:
https://ul.live/ytrateai
π Claude Code is the Biggest AI Development Since ChatGPT:
https://ul.live/ytccagi
π We're All Building a Single Digital Assistant:
https://ul.live/ytsingleda
π Stay connected
π¬ Subscribe to the newsletter at:
https://ul.live/ytnl
π¬ Join the UL community at:
https://ul.live/ytcommunity
π£ Follow on X:
https://ul.live/ytx
πΌ Follow on LinkedIn:
https://ul.live/ytli


