Claude Reads My Handwriting and Sketches Now
The creator demonstrates a workflow for capturing handwritten notes and sketches using PDF Expert on iPad, syncing via iCloud, and having Claude AI read and act on the annotated PDF files. The system connects to a local 'myPKM' folder where an AI orchestrator named Larry processes inputs and outputs. The workflow eliminates the need for complex automations by relying on simple PDF files as the universal format.
Summary
The video presents a personal knowledge management (PKM) system built around a single local folder and Claude AI, with a focus on integrating handwritten notes and sketches into this system. The creator explains that after testing various tools including reMarkable, Apple Notes, Notability, and GoodNotes, he settled on PDF Expert as the simplest solution because it natively works with PDF files, eliminating the need for conversion steps or complex automations.
The workflow is structured around two key folders inside PDF Expert synced via iCloud: a 'Team Inbox' where the user deposits handwritten notes or sketches for the AI to process, and a 'Needs Feedback' folder where the AI places documents for the user to review and annotate. The AI orchestrator, named Larry, monitors these folders and acts on their contents when instructed, reading PDFs directly and interpreting handwriting and sketches without any OCR preprocessing.
Two primary use cases are demonstrated. First, the creator shared an AI-generated plan as a PDF into the Needs Feedback folder, annotated it with checkboxes and handwritten notes using his Apple Pencil, and then had Claude read the annotations to understand his decisions. Claude successfully identified 15 annotations across 6 pages using PDF metadata, correctly interpreted most handwritten feedback, and updated its behavior via a new 'Handwritten Collaboration Loop' SOP stored in the team knowledge base. Second, the creator sketched a rough layout for a membership profile page and dropped it into the Team Inbox. Claude correctly interpreted the sketch — identifying the header, avatar placement, bio section, and activity panels — and generated a polished HTML/PDF mockup that was then sent back for review and further annotation.
The creator emphasizes that this approach scales well, noting he has run up to 12 parallel AI sessions inside the same folder. He argues that expectations for AI handwriting accuracy should be calibrated realistically, comparing it favorably to human assistants working with poor handwriting. The system is described as personal and private, not meant for team sharing, and represents the culmination of a journey through tools like Evernote, Notion, Obsidian, and Heptabase, all now replaced by a local folder and Claude.
Key Insights
- The creator rejected reMarkable, Apple Notes, Notability, and GoodNotes specifically because they do not natively store content as PDF files, requiring conversion steps or third-party sync hacks that introduce fragility into the workflow.
- The creator found that Claude itself performed OCR on handwritten PDFs without any preprocessing by PDF Expert, meaning the AI handled handwriting recognition natively as part of reading the file.
- The creator argues that AI handwriting accuracy expectations should be lowered to match realistic human assistant benchmarks, noting that a human assistant given the same poorly-written, context-free annotated PDF would likely not perform better than Claude did.
- Claude detected that the annotated PDF had increased in file size compared to the original, using this metadata signal to infer that annotations had been added before even reading the content — identifying 15 annotations across 6 pages from metadata alone.
- The creator describes running up to 12 parallel Claude sessions simultaneously inside the same local folder, handling tasks ranging from tax calculations to coding and support tickets, all coordinated through shared team knowledge files like SOPs and workstreams.
Topics
Transcript
[0:00] If you follow this channel, you know that my whole life and business runs through this one local folder and Claude, and that's it. And the the way that we set this up, I showed many times on this channel already. You can download the scaffold of this for free But there is now literally everything in this one folder And if you're into handwriting, then this video is for you because I really look deep. I tested it all to make it possible that my handwritten notes get frictionless into the single source of truth onto my local folder. [0:31] and there are many ways to do this. But I found really literally the simplest solution and the…
Full transcript available for MurmurCast members
Sign Up to AccessMore from ICOR with Tom | AI Productivity
Claude sucks. And here is why.
This video explores why Claude and other AI models produce inconsistent results, examining how model selection (Haiku vs Sonnet vs Opus vs Claude 3.5 Fable), effort levels, prompt clarity, and system context dramatically affect output quality. The speaker demonstrates these differences through a Venn diagram creation task and advocates for using organized folder systems with documented SOPs and code-based solutions to ensure consistent AI performance.
Claude can now prompt its own SESSIONS. I built the boss.
Claude now has the ability to interact with multiple sessions, allowing a streamlined orchestration of tasks through an agent called Larry. This integration can simplify task management and enhance productivity by enabling a single main session to oversee various sub-sessions or tasks.
I switched off Claude's memory. My folder does it better.
Tom explains why he disabled Claude's native memory feature and instead built a local folder-based system (myPKM) to maintain persistent context with AI agents. He shares an audit revealing systemic issues like contradictory documentation, forgotten rules, and parallel session conflicts, then outlines how to systematically fix these problems using the ICOR productivity methodology.
I do not download invoices anymore. My folder and Claude do it.
The speaker demonstrates how he uses AI (Claude) integrated with his local folder system to automate invoice collection for his accountant, downloading 36 invoices in 5 minutes with minimal manual intervention. He explains his AI team methodology, work streams, and how persistent local knowledge bases enable AI to handle recurring tasks independently without constant re-explanation.
Claude finished the work. Then I lost it in the chat.
A creator demonstrates a comprehensive local-folder-based system for managing AI-generated deliverables asynchronously, featuring an interactive browser interface that tracks decisions, changes, and conversations without relying on external databases or web applications. The system enables collaborative editing, visual annotations, and automatic context preservation across multiple parallel AI projects.