The Skill That 10x’d My Claude Code Projects
The video introduces a Claude Code skill called 'Grill Me' that relentlessly interviews users to extract tacit knowledge from their heads into reusable AI context documents. The creator explains how this front-loaded knowledge extraction leads to higher-quality AI outputs faster than iterative trial-and-error. He shares his enhanced version of the original skill by Matt PCO, which adds automatic checkpointing to preserve Q&A sessions in markdown brainstorm files.
Summary
The video opens with the core premise that since everyone uses the same AI models, the differentiating factor in AI output quality is the context and personal knowledge you inject into the system. The creator argues that the hardest part of building effective AI workflows is extracting tacit knowledge from your own head — a challenge he compares to the painstaking discovery call process used when scoping client projects.
The featured skill, 'Grill Me,' was originally created by Matt PCO and consists of a surprisingly simple four-to-five sentence prompt that instructs Claude to interview the user relentlessly about a topic, walk down each branch of a design tree, provide recommended answers, ask questions one at a time, and explore the codebase when relevant. The creator emphasizes that a 'skill' doesn't need to be a complex automation — it can simply be a reusable prompt you don't want to retype every session.
The creator then describes his enhanced version of the skill, which adds automatic checkpointing after every question. His motivation was concern that in long grilling sessions (sometimes over an hour), earlier answers could be lost or misremembered as the context window fills. His version automatically creates a 'brainstorms' folder at the project root and writes a structured markdown file containing a summary, key decisions, a step-by-step Q&A log, and open flags for unresolved questions. It also proactively identifies gaps between the brainstorm session and existing skill/doc files, offering to update them.
The creator illustrates the value using a simple graph analogy: the traditional iterative approach might start a skill at 70% effectiveness and slowly climb to 95% over many iterations, whereas using Grill Me upfront can jump the starting point to around 90%, compressing the time to reach high performance. He frames this with the Lincoln axiom about spending most of your time sharpening the axe before chopping. The video closes with instructions on how to access both the original and enhanced versions of the skill.
Key Insights
- The creator argues that since everyone uses the same AI model, the real differentiator in output quality is the personal context, taste, and decisions you inject — not the model itself.
- The original 'Grill Me' skill by Matt PCO is only four to five sentences long, which the creator uses to make the point that a skill can simply be a reusable prompt rather than a complex automation.
- The creator enhanced the original skill with automatic checkpointing after every question because long grilling sessions (sometimes over an hour) risked earlier answers being misremembered as the context window filled up.
- The creator claims that using Grill Me upfront can launch a new skill at roughly 90% effectiveness from the start, compared to the traditional iterative approach that might begin at 70% and slowly improve over many cycles.
- The skill automatically flags knowledge gaps — things the user cannot explain well — and instructs them to gather that information from the relevant stakeholder and return to update the brainstorm document.
Topics
Transcript
[0:00] The toughest part about building good skills and building a good operating system is trying to get everything from your brain into your system. So, for example, what you're looking at here is after months and months of me building up all of the knowledge that lives inside of my AIOS. It's basically just the idea that if everyone's using the same model, so if everyone's using Claude Opus 4.8, then everyone's going to be using the same prompts and getting the same output because the model is fundamentally the same for everybody. So what really makes the difference is when you add context into that model and you give it your taste, your voice, your decisions, and that's…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Nate Herk | AI Automation
Fable 5 + Karpathy’s LLM Wiki is Basically Cheating
The video demonstrates how to build a personal LLM-powered knowledge base using Obsidian and Claude, inspired by Andrej Karpathy's approach. By ingesting various data sources (YouTube transcripts, PDFs, URLs) into an interconnected wiki structure, users can create a 'second brain' that helps AI agents understand context and relationships across knowledge domains.
How Claude is Creating a New Generation of Millionaires
Claude, an AI built by Anthropic, is enabling non-technical founders to build million-dollar software companies by automating development work. The transcript outlines why Claude is outpacing competitors, provides real examples like Vulcan, and offers a four-step framework for anyone to start building with Claude today.
How Anthropic Engineers Actually Prompt Fable 5
A guide to effectively prompting Claude Fable 5, covering six key techniques including providing context/intent, using negative prompting, letting the model act once ready, verifying outputs, avoiding reasoning requests, and keeping instructions concise. The speaker emphasizes that Fable 5 is expensive and should only be used 5-15% of the time, with pricing at $10 per million input tokens and $50 per million output tokens.
Stanford's Method Turns Claude Into a PHD Level Research Team
A researcher demonstrates the STORM method from Stanford, which uses five expert perspectives (practitioner, academic, skeptic, economist, historian) to create verified research reports. The method produces 25% more organized articles than competing approaches and is packaged as a reusable Claude skill that generates HTML briefings with peer-reviewed citations.
Is Claude Mythos Coming?
A YouTuber analyzes the brief appearance of 'Claude Mythos' on Anthropic's API, arguing it signals a marketing move rather than an imminent public launch. Despite competitive pressure from OpenAI and IPO timing, the creator believes Mythos capabilities will quietly fold into future Opus models rather than release publicly under that name. Three possible scenarios are outlined ranging from a limited gated release to Mythos remaining permanently restricted to vetted security partners.