How We Do AI - Two Professionals. One Mindset.
Tom and Paco showcase their newly upgraded membership platform built with Claude Code and the new Claude Design tool, demonstrating how AI is deeply embedded in their daily business operations. They walk through the redesigned iCore learning platform featuring structured lesson navigation, downloadable resources, and an AI community manager. The episode also touches on Opus 4.7 improvements, Claude Design's slide deck and UI prototyping capabilities, and a broader philosophy about AI as a productivity partner rather than a replacement.
Summary
Tom and Paco open the session by announcing major changes to their 'my iCore' membership platform, which they built entirely using Claude Code starting in June of the previous year. They quickly pivot to discussing the newly released Claude Design tool and Anthropic's Opus 4.7 model, sharing their real-world experiences using both in their business.
Tom demonstrates Claude Design live, explaining how it functions as a separate web application tied to a Claude subscription. He walks through how he fed in their GitHub repository and brand guidelines to create a design system, then used it to generate UI components like a new commenting system, course navigation layouts, and a slide deck — all matching their existing branding automatically. He highlights the 'handoff to Claude Code' feature, which generates a command that Claude Code uses to access the design files directly within the actual codebase, eliminating the need to manually link design assets.
The platform overhaul is shown in detail: the new iCore Journey view presents lesson quality scores horizontally across all courses, allowing members to spot weak areas at a glance and jump directly to relevant lessons. Each lesson now has a structured layout with a magic slide, TLDR, cheat sheet, full 30,000-word guide broken into sections, and downloadable ZIP files for workshop-style courses. Tom contrasts this with typical platforms like Circle and Mighty Networks, which he says cannot handle the volume and asset variety their content requires.
Paco describes using Claude Design to prepare a presentation for his company's annual convention, noting how Claude's accumulated context — stored conversations, notes, images, and thoughts gathered throughout the year — can now be combined with design preferences to generate polished slide decks without needing a professional designer. Both hosts emphasize that this reduces dependency on external collaborators and removes friction from the creative process.
The hosts address two community questions: one about Andrej Karpathy's LLM wiki approach to PKM, which Tom argues lacks a defined methodology compared to their iCore system, and one about an AI context file, which Tom redirects to the iCore MCP server — a tool members can connect to Claude Desktop that gives the AI access to their full iCore framework, tool stack, learning resources, and journey progress.
They discuss their AI team structure, which includes named agents like 'Checks' (a community manager agent) and 'Felix' (a front-end developer agent), orchestrated through a system Tom uses daily inside his iCore setup. They note that Checks responds to community questions based on knowledge from all courses and past coaching sessions, and can cite specific timestamps from recorded sessions.
Paco closes with a personal reflection on how a well-configured AI system eliminates the professional solitude that entrepreneurs and executives often feel, arguing that a deeply contextualized AI acts as a genuine thinking partner that improves decision-making. Both hosts confirm the platform has reached 4,000 members, with nearly 1,000 joining in the current month.
Key Insights
- Tom argues that Claude Design's 'handoff to Claude Code' feature is a breakthrough because it allows Claude Code to access design files directly within the actual codebase via a single command, eliminating the manual linking of design assets and producing much better context-aware code changes out of the box.
- Tom claims that Opus 4.7 represents a dramatic improvement over previous weeks, where token limits were hit quickly and output quality was poor during coding sessions — stating he now runs it on 'high effort' mode for extensive code changes without hitting token issues.
- Paco argues that a well-configured AI system eliminates the professional solitude felt by entrepreneurs and CEOs, claiming his own decision-making has measurably improved because the AI acts as a partner that knows his full context — effectively functioning as an extension of himself.
- Tom contends that standard membership platforms like Circle and Mighty Networks structurally fail when content volume and asset variety are high, and that building a custom Claude Code application gave them the flexibility to create lesson layouts that those platforms cannot support.
- Tom responds to Andrej Karpathy's LLM wiki approach by arguing it lacks a defined methodology or system, and that simply creating index folders connected to Obsidian does not constitute a meaningful advancement over what anyone can already do — contrasting it with their iCore MCP server which provides the AI with structured access to a member's full learning journey and tool stack.
Topics
Transcript
[0:02] Hello everybody to another week with AI well and productivity with a from a productive week at least for the members in our membership they might have seen a huge change since yesterday. And this is just the beginning we rolled out a whole new uh learning experience. Well, we are we are no longer doing the proper the typical podcast, right? So I can share the screen and share actually. So we dive right into this. But we also want to talk about Opus 4.7 and Claude design [0:35] studio and all these things. Um we will dive into this for sure today too and obviously also whoever is inside the live call. Let's see. Okay. I didn't…
Full transcript available for MurmurCast members
Sign Up to AccessMore from ICOR with Tom | AI Productivity
Obsidian is the tool I teach in. Keep the one you have.
The creator addresses criticism about switching to Obsidian after previously promoting other tools, clarifying that his tool-agnostic ICOR methodology remains unchanged and that Obsidian is simply an interface layer for his folder-based system. He emphasizes that the underlying folder structure, AI automation, and productivity principles are identical regardless of which tool visualizes the content.
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.