53. Why Your Vibe Code Needs Tests, Ditching SaaS & AI Agents That Work While You Sleep

Beskrivelse

James and Kieran get into the practical, unglamorous side of building software with AI that nobody seems to talk about. After a HubSpot integration silently broke his video testimonial app in front of a client, Kieran shares his journey into writing automated tests, and how Claude wrote a full test suite in five minutes that would have taken a developer a week. They also cover replacing paid SaaS tools with custom-built alternatives (blogging platforms, course hosting, even a video-to-scope tool for freelance clients), why Convex is generating buzz as a Supabase alternative, how Apple and Shopify app stores are buckling under the wave of AI-generated submissions, and running OpenClaw agents autonomously in Slack as a marketing team of one. Plus a first look at Anthropic's brand new scheduled tasks feature and what it means for running AI workflows without keeping your laptop open. Topics covered: * Writing tests for vibe-coded apps * Convex vs Supabase * App store review bottlenecks * Replacing SaaS subscriptions with your own builds * OpenClaw as an autonomous marketing agent * Anthropic's scheduled tasks * The rise of solo founders 🎟️ Create With Conference — June 25th, Brighton👉 createwith.com

52. Your AI Co-Founder: OpenClaw, Manus & the Zero-Person Business

🎙️ Special Guest EpisodeJames and Kieran are joined by Matt Roberts, co-founder of Happy Operators, to discuss what it's actually like using OpenClaw day-to-day — beyond the hype. From building an asset management system while walking the dog to shipping automated marketing tools on a $10/month server, Matt shares what's working, what's risky, and what's just noise.We also cover the new Buildprint MCP server for Bubble, a surprising study on Claude.MD files degrading performance, and why the builder mindset matters more than technical skill.⏰ Timestamps[00:00] VibeCoding Olympics Wrap-Up — 1000+ participants, 100 submitted products, and how attitudes are shifting overnight[04:09] OpenClaw Real Talk — Matt's approach: separate accounts, treating it like a new hire, and running it for under $10/month[10:36] Use Cases That Actually Work — Content Bloom (automated Shopify marketing), an asset manager built via Telegram, and multi-project management[18:11] Security Reality Check — The Meta AI incident, sandboxing strategies, and what permissions to actually give it[22:10] The Update Problem — Why "Bob" couldn't restart himself, and the messy reality of self-hosted agent maintenance[27:02] Manus vs OpenClaw — Meta's acquisition, the 300-ad-creative toggle, and why every platform is racing to add autonomous agents[30:30] Bubble's AI Revolution — The Build Prints MCP server, live editing from Claude Code, and how it compares to Bubble's native AI[31:30] Claude MD Files: Help or Harm? — New research suggesting project instructions may actually degrade model performance[33:36] The Builder Mindset — Why persistence beats technical knowledge, and how desktop apps are lowering the barrier💡 Key TakeawaysTreat OpenClaw like a sandboxed employee — give it its own accounts and limit what it can accessThe real value isn't unique capabilities; it's the user experience of building via Telegram while you're out and aboutSecurity through isolation beats trusting the agent with everythingClaude MD files may be doing more harm than good — worth revisiting your setupYou don't need to be technical, but you do need to be persistent and curious🔗 Join the CommunityCreate With Community Kickoff Party📅 Wednesday, March 4th🌐 createwith.com/joinA supportive space for non-technical builders learning AI tools — office hours, challenges, and no stupid questions.#OpenClaw #AIAgents #ClaudeCode #NoCode #VibeCoding #BuilderMindset #CreateWith #HappyOperators

25. feb. 202637 min

53. Why Your Vibe Code Needs Tests, Ditching SaaS & AI Agents That Work While You Sleep

Beskrivelse

Kommentarer

1 måned kun 9 kr.

Alle episoder