I Built a Knowledge Engine in 30 Minutes
And my AI coworker did all the heavy lifting
This essay is a little different from my usual fare. Over the last 48 hours, I’ve been playing with Claude Cowork — the latest tool released by OpenAI competitor, Anthropic. If you haven’t heard of it, I’d suggest checking out the videos highlighted below. But to give you a sense of its power, I’m going to describe how I used it to scratch an enduring itch of mine.
In just 30 minutes, I built a tool that lets me automagically summarise YouTube videos, save them to a personal knowledge graph I can search and navigate, and then — after answering a couple of shaping questions about audience and tone — draft an essay from any slice of that library. Not configured a tool. Not customised a template. Built one, from scratch, as a Chrome extension. Here’s how Claude describes what we made together:
If you’re someone who watches a lot of YouTube, you know the frustration of investing 20 minutes into a video only to realize it wasn’t relevant to you. This Chrome extension gives you that judgment call in seconds. One click and you get a concise summary, the key takeaways, and a personalized relevance score — so you can make an informed decision about where to spend your attention before committing to a single minute of watch time.
But the real power is what accumulates over time. Every video you save becomes a node in your personal knowledge graph, automatically tagged and connected to related topics. Instead of insights disappearing into your watch history, they become searchable, navigable, and — crucially — synthesizable. When you’re ready to write about a topic you’ve been exploring, the extension pulls together everything you’ve learned across multiple videos and generates a first draft shaped to your audience, angle, and tone. It turns passive consumption into an active, compounding knowledge base.
So after watching the videos below, and ingesting them into my library, I asked my extension to synthesise what I’d learned into an essay. What follows is that output — lightly edited, but essentially machine-generated from my own curated knowledge base. Read it as a proof of concept: the tool writing about the tool that built it.
What Claude Cowork Actually Does (And Why It’s Different) - Tiago Forte
I Set Up Claude Cowork. My AI Now Remembers Everything. - ICOR with Tom
Claude Cowork Is Wild… Here’s How to Use It Properly - Skill Leap AI
When a Tool Becomes a Partner
Forget the distant, sci-fi future of artificial intelligence. The era of having a true digital partner—one that remembers your preferences, organizes your chaos, and executes complex projects from a simple prompt—is not on the horizon. It’s sitting on your desktop, waiting for you to assign it a folder. The recent launch of Claude Co-work represents a fundamental shift in how we interact with AI, moving it from a novelty chatbot to an integrated, persistent, and remarkably capable colleague. As several creators have demonstrated, this isn’t about asking questions anymore; it’s about delegating work.
From Chatbot to Colleague: The Mindset Shift
The most profound change Claude Co-work introduces is a complete reimagining of the user-AI relationship. As Tiago Forte astutely points out, this tool transforms AI from a “chat buddy” into a collaborative partner. The paradigm shifts from a transient conversation in a cloud-based chat window to a persistent, task-oriented collaboration within a shared workspace—a folder on your own computer. This local storage, as highlighted in both Forte’s and Tom from ICOR’s videos, is crucial. It means your AI’s understanding of you, your projects, and your workflows lives securely with you, fostering trust and enabling deep, context-aware assistance.
This shift requires a change in our own approach. It forces us to confront how much of our work was friction, not thinking. You can’t just throw disorganization at an AI and expect magic. As Tom emphasizes, “The key to effective AI assistance is having clear workflows and systems in place first.” AI is a multiplier, not a magician. Once you provide structure, however, the amplification is dramatic. Imagine an assistant that doesn’t just rename a file, but can autonomously categorize 2,900+ screenshots, as shown in Skill Leap AI’s demo. It’s this move from reactive Q&A to proactive project execution that defines the new era.
Building a Partner with a Memory
The magic of a true co-worker is continuity. A human assistant learns your habits, remembers your past projects, and applies that knowledge to new tasks. Claude Co-work replicates this through a simple mechanism: a local markdown file often called claude.md. This file becomes its memory. You don’t program it with code; you train it through conversation. Tom’s video illustrates this onboarding process, where the AI asks questions about your role, preferences, and goals, then updates its knowledge file. The result is a personalized agent that understands context. When you later ask it to “prepare a meeting brief for the Acme Corp call,” it already knows your job title, your typical briefing format, and can pull from your calendar and notes to build a comprehensive document.
Memory plus tool access unlocks real power. It’s not just about what the AI knows; it’s about what it can do with that knowledge. The demos across these videos paint a picture of a hyper-competent generalist:
The Personal Organizer: monitors a “scan inbox” folder and automatically files documents into categorized subfolders.
The Marketing Director: from one prompt, generates an entire campaign—ads, blog posts, newsletters, and a launch plan—while maintaining brand consistency by referencing your style guide.
The Research Analyst: web-searches for cheaper insurance alternatives, compares policies, and presents a clear breakdown.
The Executive Assistant: integrates with your calendar, optimizes your schedule, and can add events from its proposed plans.
The Practical, Accessible Future of Work
Perhaps the most compelling aspect of this shift is its accessibility. As Forte notes, this level of collaborative AI capability is now available for about $20 a month—a far cry from the enterprise-only tools of the recent past. While power users may gravitate toward higher tiers, the barrier to entry for a transformative productivity tool is remarkably low.
The implications are vast. Creativity isn’t being replaced. It’s being cleared of debris. This isn’t about removing human strategy or imagination; it’s about offloading the cognitive overhead and tedious execution that stifles them. It’s the difference between thinking about planning a family road trip and having your AI co-worker research destinations, draft an itinerary, and build a budget spreadsheet while you focus on what actually matters.
The future of knowledge work isn’t a dystopia of human obsolescence. It’s a partnership. A future where repetitive tasks fade into automation, information organizes itself, and ideas turn into tangible outputs with startling speed. The tools are here. They’re affordable. They work with the files already sitting on your drive.
Your co-worker is waiting.