TechSambad · Friday Edition · June 25, 2026 I Stopped Talking to My AI and Started Designing Its Problems— Or: How Kunia Made Me a Loop Designer Instead of a Prompter By Subhankar Pattanayak
🔴 The ConfessionFor two years, I thought Vibe Coding was me talking nice to Claude. Be polite, be specific, break it down step by step, and the AI would build me an app. I was a prompter — and I was proud of it. Then, on a random Tuesday morning in June 2026, two things landed in my feed within minutes of each other:
- A senior Anthropic engineer dropped an 11-page PDF on "Loop Engineering" — Schedule → Discover → Build → Verify → Repeat
- The creator of Claude Code said he doesn't prompt it anymore. Loops prompt it. His job is to design the loops.
The universe was screaming at me. And the funny thing? I'd already built a loop without realizing it. 🎠The Prompter's DelusionWe all thought we were conductors. We were just very polite telegraph operators. "Please generate...", "Could you kindly...", "Let's think step by step..." — we were trying to befriend our microwave. The best prompters in the world were simply the best typists. We'd built a whole economy around talking to AI politely. The uncomfortable truth hit me: the person who writes the best prompt is still just a prompter. The person who designs the system that generates the prompts — that's the loop designer. And that's where 2026 is headed. 🤖 What My AI Assistant Taught Me (Without Either of Us Noticing)I set up Kunia — my personal AI — to curate, analyze, queue, and publish. The workflow goes:
Link arrives → Fetch content → Analyze & summarize → Queue in Findings → Queue in Featured → Publish in TechSambad → Clear for next day
Notice something? I wasn't talking to Kunia. I was designing the system that processes information without me touching it. I wasn't writing prompts — I was building a loop. Schedule → Discover → Curate → Verify → Publish → Repeat. Sound familiar? Meta punchline: I learned about loop design by accidentally building one to run my newsletter. Kunia didn't teach me to be a better prompter. It taught me to fire myself from that job. 🔗 The Two Links That Landed the Same MorningLink 1 — The Anthropic Engineer's Loop Engineering PDF
The five-move loop for agentic systems:
1️⃣ Discovery — The agent finds its own work: failing CI, open issues, recent commits. You don't hand it a list.
2️⃣ Handoff — Each task gets an isolated git worktree. Parallel agents don't collide.
3️⃣ Verification — A second agent, told to assume the code is broken, reviews the first. The "thing that can say no."
4️⃣ Persistence — Results go to disk, never left in a context window that gets flushed.
5️⃣ Scheduling — An automation wakes it on a timer. That's what makes it a loop.
Link 2 — 5 Levels of Loop Design (by 0xMorty)
The Claude Code creator distilled it in one sentence: "I don't prompt it anymore. Loops prompt it. My job is to design the loops."
Level 1: Type prompts manually → "I vibe code"
Level 2: Save prompts as templates → "I have a knowledge base"
Level 3: Write a script that runs prompts → "I'm automating workflows"
Level 4: Build a loop that discovers its own work → "I do loop engineering"
Level 5: The loop builds better loops than you could → "I'm not sure what I do anymore but my agent is happy"
⚡ What This Means for Vibe Coding's Next ChapterVibe Coding 1.0: You + LLM = app.
Vibe Coding 2.0: You design the loop. The loop builds the app. Other loops review it. The question is no longer "what should I ask the AI?" — it's "what should the AI ask itself, when, and who checks its homework?" The key insight that's going to save thousands of developers from embarrassment: an agent grading its own work always praises itself. You need a second agent — one that's cynical, skeptical, and told to find the bugs. This is the "thing that can say no." It's the single most important component in any agentic loop.
🎯 The TakeawayI started 2026 trying to write better prompts. I'm ending it designing loops that don't need me. If you're still typing every prompt by hand, you're doing 2024 work in 2026. The shift from Prompter to Loop Designer is the most important skill you can develop right now. Not because AI is taking your job — but because the people who design the loops will outperform the people who just talk to the AI. Final thought: Kunia didn't teach me to be a better prompter. It taught me to fire myself from that job. And that's the best thing that's happened to my productivity since I discovered that LLMs exist.
Links: • Anthropic Engineer's Loop Engineering PDF • The 5 Levels of Loop Design by 0xMorty — Curated & written with Kunia (my AI assistant) —
TechSambad · Daily AI & Vibe Coding insights |
