Claude Code New Features, Explained
Get Claude Code: https://startup-ideas-pod.link/claude-greg In this solo episode I lay out the exact system I use to turn Claude Code into what I call an AI employee. My premise is simple: give Claude the same things you would give a person joining your company (a workspace, memory, a brief, a clear ticket, eyes, review, a schedule, and permissions). I build the whole setup live around a real idea I found on ideabrowser.com, a missed-lead responder for med spas, and I share the specific prompts I use at each step. By the end, the product, the customer feedback, the docs, the demos, the reviews, and the recurring work all live inside one operating loop. I close with a seven-day plan you can run at your own pace. And a special thank you to Claude and Anthropic for supporting the podcast. Setup Claude Code to be your 24/7 employee: https://startup-ideas-pod.link/Claude-code Timestamps 00:00 – Intro: The AI Employee Map 05:47 – Step 1: Creating The Workspace In Claude Desktop 14:53 – Step 2: The Brief And Plan Mode 18:08 – Step 3: The Ticket And Defining Done 22:15 – Step 4: The Eyes And Desktop Preview 26:13 – Step 5: Review In Layers And The Diff View 29:34 – Step 6: The Schedule And Routines 34:32 – Step 7: Parallel Agents And Worktree Isolation 39:14 – Step 8: Permissions: Safe, Ask First, Human-Owned 41:15 – Step 9: Skills, Connectors, And Hooks 44:28 – The Seven-Day Plan 47:42 – Closing Thoughts Key Points • I treat Claude Code like a new hire: workspace, memory, brief, ticket, eyes, review, schedule, permissions. • The repo brain teaches Claude how I work and what good looks like. • Plan mode comes first: Claude reads the context, proposes an approach, and waits for my approval. • One ticket at a time means one task, one finish line, and one reviewable change. • The eyes matter: Claude opens the app in desktop preview, clicks the flow, checks the console, and reports what a buyer experiences. • Routines and permissions turn a chat tool into a 24/7 operator with clear boundaries. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/
Read Video · Transcrição & Insights
Este episódio tem transcrição completa + análises IA
Conta gratuita · sem cartão · 150 créditos no cadastro, suficiente para este episódio
- 📄 Transcrição completa com timestamps
- ✨ Resumo IA, palavras-chave e mapa mental
- 💡 Pontos-chave e citações
Linha do tempo do episódio
Introduction: most people use Claude Code wrong, and the fix is setting it up like an AI employee.
- Claude Code is described as one of the most powerful pieces of technology ever created, but most people misuse it because they never set it up properly.
- The host promises to reveal nine areas to master so Claude Code can spin up AI employees that work 24/7 and build AI-native companies.
- All prompts and 'sauce' are given away free in the episode, sponsored by Anthropic.
The map of an AI employee: workspace, memory, brief, ticket, eyes, review, schedule, and permissions.
- An AI employee needs the same basics as a human hire: a workspace (repo), memory (project context), a brief (plan mode), and a clear ticket.
- Vague prompts like 'make the app better' fail; specific tickets like 'add a waitlist form with a success state and check it in desktop preview' work far better.
- Claude Code now has 'eyes' — it can open the app, click through flows, and report what a customer would actually experience.
- Review, schedule (routines), and permissions complete the map, with humans owning trust decisions like dependencies, migrations, and payments.
Step one: creating the workspace and repo structure for a med spa missed-lead responder idea.
- The demo idea from ideabrows.com is a missed-lead responder for med spas, where inbound leads go cold because the business replies too late.
- The recommended repo structure is /app, /context, /customers, /spec, /demo, /routines plus three root markdown files: claude.md, roadmap.md, and review.md.
- A single setup prompt tells Claude the product, buyer, pain, promise, and goal, and asks it to request any missing context before writing.
- Claude asks high-leverage questions first, then scaffolds the entire optimized workspace automatically.
Conceitos-chave
- Claude Code— The core tool being taught throughout the episode as a platform for building AI employees.
- AI employee— The central metaphor and goal — configuring Claude Code to work like a full-time staff member.
- workspace repo— The first pillar: a structured repository where the product and context live.
Citações notáveis
I think there's a lot of people using claude code wrong. And they're using claude code wrong because it is one of the most powerful pieces of technology that have ever existed.
🔥— It asserts that most users are misusing a tool they think they already understand, setting up the entire reframe of the episode.
The mistake a lot of people make is they'll say, "Make the app better. Make this pop." These very vague prompts uh that are assigned to it.
💡— It names the exact habit most users have and shows why it forces Claude to guess instead of execute.
Ações a tomar
🗂️Workspace Setup
Claude becomes far more useful when the project explains itself through structured files and folders.
This week, create CLAUDE.md, roadmap.md, and review.md in your project root and populate each with business context, current goals, and quality standards.
A repo organized into /context, /customers, /spec, /demo, and /routines gives Claude a place to work and a brain to think with.
Create those five folders today and drop at least one real customer note and one demo script into them.
📋Planning and Execution
Plan mode forces Claude to inspect context and propose an approach before touching any files.
On your next task, prompt Claude to use plan mode, list files to change, risks, and verification steps, then wait for your approval before editing.
Vague tickets like 'make the app better' force Claude to guess, which turns you into a cleanup crew instead of a manager.
Rewrite your next three prompts to include one task, one finish line, and one reviewable change.
A transcrição e os insights são gerados por IA e podem conter erros. A precisão depende da qualidade do áudio e da clareza dos falantes — se algo parecer errado, o áudio original é sempre a fonte mais confiável.
Episódios e vídeos prontos para ler
Episódios de podcast
87 | On the joys of being a young scientist
Night Science
13 de jul. de 202638:07EN
What makes a good game? | Eric Zimmerman
TED Talks Daily
22 de set. de 202614:16ENEpisode 24: Why Big Bettors Still Play Offshore
Zero Latency: An Eilers & Krejcik Gaming Podcast
4 de mar. de 202424:09EN
#404 | POR QUE TANTAS EMPRESAS ESTÃO PEDINDO RECUPERAÇÃO JUDICIAL?
Market Makers
25 de ago. de 20261:31:16PT
Folge 1: Koronare Herzkrankheit (KHK) – Medizin auf Deutsch: MedVibe
Medizin auf Deutsch | MedVibe
9 de mar. de 202625:05DE
#6「エヴァってめっちゃ難しくない？」
朝井リョウ・加藤千恵 信頼できない語り手
13 de mar. de 202649:21JA
Vídeos
Making $$$ with Loop Engineering
Greg Isenberg
13 de jul. de 202639:44EN
Sam Altman in conversation with Patrick Collison
Stripe
30 de abr. de 202658:32EN
The Secrets of Ageing Better | Future Lives | u3a UK
u3a
31 de out. de 202258:10EN
COMO SER GENTIL ESTÁ ACABANDO COM A SUA AUTORIDADE? | Fabiana Bertotti #152
Como Você Fez Isso?
30 de jul. de 20261:19:47PT
#39 Die Pest
99 mal Geschichte
8 de jan. de 20261:02:15DE