Skip to content
Aakash GuptaAakash Gupta

Everyone's Using Claude. This PM Tool Does More

This episode is a full walkthrough of how Meng To, founder of Design+Code, actually uses Codex day to day. Plan mode, a 20-agent fleet, the taste skill that makes AI design look senior level, and a digital twin built from ten old photos, plus his take on why some PMs are surviving this year's layoffs. Full Writeup: [URL-VERIFY] Transcript: [URL-VERIFY] --- Timestamps: 0:00 - Intro 2:44 - The best AI design tools right now 4:05 - What Codex actually is and how to set it up 8:43 - Inside Meng's real Codex project folders 13:42 - Plugins vs skills vs computer use 22:00 - Ads 27:03 - Building your project folder system 30:57 - The permission tiers, from default to full access 32:55 - Plan mode in action, building a slide from real data 42:03 - The screenshot shortcut that changes everything 47:32 - The taste skill that makes AI design look senior-level 55:07 - Building your own AI digital twin 1:05:31 - How some PMs are surviving the layoffs 1:12:40 - Outro --- 🏆 Thanks to our sponsor: Arise: Trace, evaluate, and fix your AI agents before broken behavior ships to users - https://arise.com --- Key Takeaways: 1. Codex isn't a chatbot, it's a fleet operator - Meng runs 20 agents at once while stepping away from his desk entirely. Each one works on something different, slides, charts, brainstorms, while he does something else completely. 2. Plan mode isn't optional - Skipping it means paying twice, once to build the wrong thing, once to undo it. Codex returns a full breakdown, architecture, steps, and open questions, before touching anything. 3. A screenshot beats a paragraph every time - It shows the AI what you mean instead of what you think you mean. A simple two-key shortcut drops any window straight into the chat as context. 4. The taste skill is the real differentiator - Without it, AI design defaults to generic. With it, the output looks like something a senior designer with years of experience actually made. 5. Trust is earned in tiers - Read only first, then supervised access, then full access. Skipping straight to full access before learning where the AI tends to fail is how people get burned. 6. HTML beats Figma for speed - Every extra tool is a login, a subscription, and a context switch the AI can't do for you. Keep your blast radius small. 7. UGC won because audiences are tired of corporate polish - A synthetic version of you, used honestly, reads as more human than a generic message. Ten old photos is all it takes to build a digital twin. 8. Technical PMs aren't surviving layoffs because they write code - Meng hasn't written a single line in six months. They're surviving because they're fluent enough to direct a fleet of agents and catch a wrong output before it ships. 9. Meng builds his own tools when nothing off the shelf fits - His own video editor, his own SaaS templates, his own design brainstorming app. The tool built for your exact workflow beats the popular one every time. 10. The bar isn't five star anymore - Five star is just the floor everyone clears by default now. The real question is what six, seven, all the way to eleven star looks like, because that ceiling rises exactly as fast as the floor does. --- 👨‍💻 Where to find Meng To: LinkedIn: https://www.linkedin.com/in/mengto Design+Code: https://designcode.io Aura: https://aura.build 👨‍💻 Where to find Aakash: Twitter/X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com #ProductManagement #AIDesign #Codex --- About Product Growth: The world's largest podcast focused solely on product + growth, with over 200K+ listeners. Subscribe and turn on notifications. --- YouTube Tags: Codex, OpenAI Codex, AI design, Claude Code, product management, AI PM, design with AI, Design+Code, Meng To, AI agents, plan mode, AI avatar, digital twin, PM career, AI tools for PMs

Meng ToguestAakash Guptahost
Jun 23, 20261h 14mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Codex-first PM workflow for AI design, slides, and career survival

  1. Meng To argues Codex is a “ChatGPT on steroids” because it is project- and folder-based, can operate on local files, and can use tools (including “computer use”) to execute real work rather than just chat.
  2. A practical setup is emphasized: keep work local in structured project folders, use Obsidian as a knowledge layer to organize the flood of AI-generated docs, and add voice input (WhisperFlow) to increase throughput.
  3. The episode differentiates plugins vs skills vs computer use, framing plugins as deep integrations, skills as user-configurable reusable instructions/recipes, and computer use as a universal automation layer that can click through any UI.
  4. Live demos show Codex planning-first workflows (Plan mode), permission tiers, generating slides from real local data, iterating with screenshots as high-signal context, and using a “taste” skill to raise design quality.
  5. Career-wise, Meng claims layoffs disproportionately hit non-technical PMs, and “technical” now means understanding AI/tooling jargon and orchestrating an agent fleet to deliver above-baseline quality—potentially as a founder with AI handling the “janitor work.”

IDEAS WORTH REMEMBERING

5 ideas

Treat “projects” as folders to make AI work persistent and reusable.

Meng’s core workflow is folder-based: each Codex project points at a local directory so artifacts (markdown, images, code, slides) remain accessible outside the chat and can be organized like real work products.

Use Obsidian as the organizational layer for AI-generated documentation.

Codex may generate many .md docs, but you won’t want to hunt them in chat history; Obsidian provides a navigable knowledge base (“brain tree”) over the same local folders.

Start with Plan mode to prevent premature building and reduce rework.

He repeatedly recommends asking the agent to plan first (MVP, architecture, steps, deliverables), then asking clarifying questions before authorizing execution—especially for app builds and presentations.

Choose automation depth: plugins for integrations, skills for reusable guidance, computer use for universal UI control.

Plugins are “team-built” integrations (Gmail, Calendar, Linear), skills are lighter-weight user-defined recipes (copywriting, SwiftUI, Tailwind, “taste”), and computer use can operate any app even without an integration.

Permissions and token settings are product decisions, not just preferences.

Meng frames default vs full access and low/medium/high compute as levers balancing trust, cost, and accuracy; he uses higher settings with more tokens to run many concurrent agents.

WORDS WORTH SAVING

5 quotes

Technical is this. This is what I mean by technical. You're no longer the, the bureaucracy player, the politics player in the company. Nobody needs that anymore, because AI can kind of juggle around that.

Meng To

You have an army of agents. So when I say an, an agent, like this is an agent, this is an agent, this is an agent, and they're all working simon- simultaneously.

Meng To

So Codex, which for those who don't know, is basically the ChatGPT but 10X, like on stero- steroids basically.

Meng To

HTML is gonna be the most controlled, the fastest, and it's all gonna happen in Codex.

Meng To

Eventually, you will get fired because AI will replace everyone at some point, or we will find new ways to, to do our job.

Meng To

Codex vs Claude/Cursor/Open-source agentsLocal-first project folders and context managementObsidian as a knowledge base for AI-generated artifactsPlugins vs skills vs computer usePermission tiers, token budgets, and speed/quality tradeoffsPlan mode and planning-before-building workflowsSlides/design generation: image drafts, taste skill, HTML exportScreenshot shortcut and mobile Codex controlling desktopDigital twin/avatars for UGC and internal commsPM-to-founder shift amid layoffs and rising quality bar

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.