UX for AI
Hosted by Behrad Mirafshar, CEO of Bonanza Studios, Germany’s Premier
Product Innovation Studio, UX for AI is the podcast that explores the intersection of cutting-edge artificial intelligence and pioneering user experiences. Each episode features candid conversations with the trailblazers shaping AI’s application layer—professionals building novel interfaces, interactions, and breakthroughs that are transforming our digital world.
We’re here for CEOs and executives seeking to reimagine business models and create breakthrough experiences, product leaders wanting to stay ahead of AI-driven product innovation, and UX designers at the forefront of shaping impactful, human-centered AI solutions. Dive into real-world case studies, uncover design best practices, and learn how to marry innovative engineering with inspired design to make AI truly accessible—and transformative—for everyone. Tune in and join us on the journey to the future of AI-driven experiences!
UX for AI
EP 109. I Built a Blog Automation App in 2 Weeks Using Claude Code (Two-Agent CTO Approach) - Video On Spotify!
Watch the full episode on youtube (or Spotify): https://www.youtube.com/watch?v=ttf2tQy0x_U
We build MVPs in 2-week sprints. This sprint is an internal “Blog Maker” app. I’m wrestling the last mile, always the hardest part of any launch.
👉 What the app does
+ Pull YouTube videos from our channel
+ Analyze transcript + do SERP/competitor checks
+ Use two AI agents — researcher → writer — to draft
+ Push the draft straight to Webflow for editing where it lives
Saves hours per post. The writer agent repurposes for our ICP and SEO rules.
👉 Why we built an app, not an automation
+ Complex automations turn into field soup and never-ending maintenance.
+ I find it easier to change and maintain codebase. The AI agent can do most of heavy lifting rather than me clicking myself to suffocation on n8n.
👉 The two-agent CTO approach
+ Agent 1: “CTO” — ruthless reducer
+ Agent 2: “Visionary” — blue-sky adder
They argue over the same prompt. I cherry-pick, then iterate the prompts every run. I find giving long prompts to one agent is a bad practice. just produces garbage.
👉 Lesson learned:
I burned ~$150–200 on API calls testing YouTube transcript paths so far in less than a week. Should’ve used a reliable paid API for transcripts and moved on.
👉 Status right now:
Content quality is there (2,500–3,000 words), SEO linter checks are running, but markdown + external linking still need love before I auto-publish.
👉 Tool stack:
Stack (today): cloud code (plan-mode), ChatGPT-Codex (debug mode), a lot of terminal, Git. I let cloud code run with full permission (YOLO), then review.
Part 2 next week: either “celebration mode” or “here’s the fix I shipped.” If you want me to share the two-agent prompt skeleton or file structure, comment “AGENTS”.
P.S. At Bonanza Studios | Digital Transformation, we run bi-weekly sprints building MVPs like this for mid-market digital teams. Your idea that's been stuck in meetings for months? We develop it in 2 weeks. DM if you're interested.
#MVP #AIPROTOTYPING #claude #MVP #appdevelopment
Interested in joining the podcast? DM Behrad on LinkedIn:
https://www.linkedin.com/in/behradmirafshar/
This podcast is made by Bonanza Studios, Germany’s Premier Digital Design Studio:
https://www.bonanza-studios.com/