Stop Losing AI Work. Ultimate Backup Guide for the Non-Techy.
The video demonstrates how to back up and version-control local AI knowledge base folders using cloud storage services and GitHub, then explains how to access these folders remotely from mobile devices using Claude's remote control feature in VS Code. The presenter argues that combining cloud backup with Git versioning gives users the ability to roll back AI agent configurations and track incremental changes over time.
Summary
The presenter begins by referencing a previous video where he built a Personal Knowledge Assistant (PKA) demo folder using Claude, featuring an inbox, a team of AI agents, and a custom interface. The current video focuses on two goals: ensuring the local folder is properly backed up and making it accessible from multiple devices.
For basic backup, the presenter explains that cloud storage services like iCloud, Dropbox, Google Drive, and OneDrive can sync local folders automatically. He demonstrates his own setup where his PKA system is already synced with iCloud, while business folders for his team are stored in Dropbox. He highlights Dropbox's selective sync feature, which allows him to mark large files (like 4+ TB of video editing assets) as 'online only' to conserve local storage while still maintaining cloud backup.
The presenter then introduces GitHub as a second, complementary layer of protection on top of cloud storage. The key distinction he draws is that GitHub provides versioning — a timestamped history of every committed change — rather than just a backup copy. This allows users to ask Claude to look backwards through the commit history to identify when something broke and extract working configurations from the past. He emphasizes this is not simply about rolling back and losing recent work, but about having a full track record of incremental changes that Claude can intelligently navigate.
He demonstrates the Git workflow inside VS Code, showing how changed files appear in the source control panel and how Claude can execute 'commit and push' commands on the user's behalf through natural language. He also walks through setting up a new local Git repository for his personal PKA demo from scratch, using Claude in the terminal to run 'git init,' configure a .gitignore file, and then sync the repository to a newly created private GitHub account.
On the topic of cross-device access, the presenter experiments with accessing the GitHub repository directly through Claude's mobile app and Claude Code, but finds the pull request and merge workflow too cumbersome. He instead recommends using VS Code's built-in '/remote' command to activate a remote control session, which mirrors the local Claude terminal session directly to the mobile Claude app. This allows him to interact with his local folder from his phone without any GitHub merge complexity.
The video closes with a summary of the recommended stack: a cloud service for backup and multi-device sync, a local or remote Git repository for versioning, VS Code as the all-in-one interface, and the remote control feature for mobile access. The presenter teases a follow-up video on building persistent, self-improving memory for AI agent teams within these local folder structures.
Key Insights
- The presenter argues that GitHub versioning is fundamentally different from cloud backup because it allows Claude to navigate backwards through commit history and compare past working configurations with a broken current state — effectively using the AI to debug itself across time rather than simply restoring an old snapshot.
- The presenter demonstrates that Claude can execute the entire Git workflow — committing changes, pushing to GitHub, and initializing new repositories — through plain natural language commands inside VS Code, removing the need for users to learn Git syntax directly.
- The presenter finds that accessing a GitHub repository directly through Claude's mobile app creates a confusing pull request and merge workflow, and concludes this approach is too cumbersome compared to using VS Code's '/remote' command to mirror a local terminal session to the mobile app.
- The presenter explains that GitHub is not suitable for storing large media files like images or video assets — his Dropbox folder holds over 4 TB of video editing data — and that Git repos should only track agent configurations, interfaces, and knowledge base text files, not binary or large-format content.
- The presenter describes running multiple terminal sessions in VS Code simultaneously, each in a separate remote control session, which can then be independently accessed and switched between on the Claude mobile app to manage different local folders or AI agent teams from a phone.
Topics
Transcript
[0:00] In a recent video, I showed you how we built this PKA demo folder here. So, this is essentially just a folder that we created using Claude where we ended up created a personal knowledge base with an inbox with a team of agents and all this and even having an interface just built for ourselves this way. So, this is the current state of it. And if you haven't seen the video, feel free to watch it afterwards. It's there's no need to watch it to what I'm showing you today. But if you want to get to the [0:30] same stage, you can watch the previous video. I will link it in the description below. And today,…
Full transcript available for MurmurCast members
Sign Up to AccessMore from ICOR with Tom | AI Productivity
I stopped doing my weekly review. The AI does it now.
The speaker has replaced traditional weekly reviews and manual note-linking with an AI-powered system built on a local folder structure called a 'scaffold.' This approach eliminates the time-consuming review ritual while enabling the AI to automatically connect notes and surface meaningful patterns that humans would miss.
I Made Opus and Fable Grade Each Other. Opus Admitted It Lost.
The creator demonstrates real-world AI applications in their productivity business using Claude Fable and Opus, showing how Fable excels at comprehensive analysis of complex systems while comparing its performance to Opus on specific tasks. The key finding is that Fable's holistic understanding justifies its higher cost for business-critical audits and optimizations.
Our AI invented the numbers. We caught it.
The hosts discuss the importance of double-checking AI outputs and validating information, emphasizing that this is not a new problem but rather a longstanding practice required with any information source. They explore how AI tools can be customized to individual workflows and contexts, arguing that understanding foundational methodology matters more than specific tools.
An AI Just Handed Me a Fake $67B Statistic
The speaker shares how an AI confidently generated a completely fabricated $67.4 billion statistic while researching AI hallucinations, then explains his systematic approach to combating AI misinformation: using dual independent search engines to cross-check answers and identify where hallucinations occur.
AI Is Quietly Taking Your Memory
AI companies are racing to own users' context and memory through integrated tools like Claude in Slack and ChatGPT's auto-memory features, creating a lock-in trap. The speaker argues users should own their context (expensive, irreplaceable) in local plain text files while renting the AI model (cheap, replaceable), and demonstrates his personal knowledge assistant system built on markdown files that work with any AI model.