DEV Community: Isak Kim

Building in Public Month 1 Real Numbers Real Failures Real Lessons

Isak Kim — Sun, 17 May 2026 22:36:34 +0000

Month 1 Summary

69 products. $27 revenue. 0 email subscribers. 180+ hours invested.

What Worked

The 4-hour product rule. Claude Code as documentation partner. Whop affiliate infrastructure.

What Failed

No pre-launch validation. Audience building came after product (wrong order). Zero email subscribers.

The Metric That Matters

Email subscribers. 1,000 social followers = 30 see your post. 1,000 email subscribers = 450 see your email. Currently at 0.

Month 2 Plan

Content: 2 articles/week. Reddit: 3 posts/week. Bluesky: 5 posts/day. Affiliates: 10 targets at 40% commission. KPI Day 60: 100 subs, 5 affiliates, $500 revenue.

The Compound Lesson

This article goes to 40 people. Next goes to 41. Compound interest, week one.

Catalog: whop.com/gaiaratio

Why 90% of AI Prompt Packs Fail (And What McKinsey Got Right)

Isak Kim — Sun, 17 May 2026 22:06:25 +0000

The Prompt Pack Graveyard

Search "prompt pack" on any digital product marketplace. Thousands of results. Most fail.

Not because the prompts are bad. Because they solve the wrong problem.

What Most Prompt Packs Get Wrong

A typical prompt pack: 'Here is a prompt for cold emails. Here is one for Instagram. Here is one for LinkedIn.'

The buyer downloads it, uses each prompt once, never opens it again.

This is the one-size-fits-all failure - the prompt knows nothing about you, your brand voice, your audience. It produces generic output. Generic output doesn't convert.

What McKinsey Got Right

McKinsey's 2025 State of AI identifies the most productive AI organizations as those with context-rich, workflow-integrated AI systems - not loose prompt collections.

Organizations with documented AI workflows outperform ad-hoc AI use by 3.4x on efficiency metrics.

The 3.4x comes from context persistence: your AI knows your business every session, not just this one.

Generic Prompt vs. Context-Rich System

Generic:

Prompt: "Write a cold email for [product] to [audience]" Output: Generic. Could be anyone's.

Context-Rich:

System Prompt: "You are [Brand]'s outreach specialist. Audience: [specific ICP]. Tone: [specific voice]. Product solves: [specific pain point]. Differentiator: [specific claim with data]. Write cold emails that feel like a real person who understands the recipient." Output: On-brand, on-audience, on-message. Every time.

The second takes 30 minutes to build. It replaces every cold email prompt you'll ever need.

Three Things That Make AI Prompts Work

Specificity of Context - Generic context = generic output
Workflow Integration - Prompts in a separate doc get abandoned. Integrated prompts get used.
Output Validation - What does good look like? What are the red flags?

This is what separates systems from suggestions.

What I Built

Each of my 69 products at whop.com/gaiaratio includes:

Context template (your inputs)
Core prompt system (tested, not theoretical)
Example output (what good looks like)
Customization guide (3-5 fields to change)

Systems. Not suggestions.

30-day refund if it doesn't work for you.

GaiaRatio | Day 34 of Building in Public

From Idea to Product in 7 Days A Brutal Sprint Breakdown

Isak Kim — Sun, 17 May 2026 22:05:18 +0000

Day 0: The Constraint

7 days. One product. Price point: \,997.

That's the constraint I set when building the MRR Vault - 69 AI productivity systems sold with Master Resell Rights.

The \,997 price positions against the second option: building the products yourself. 69 products at 4 hours each = 276 hours = 6.9 weeks. At \/hour consulting rate: \,800 in time cost. So \,997 is an 85% discount on the build cost. The math must be visible for the price to make sense.

Day 1-2: Architecture

Before building anything, two days on structure:

Pricing ladder:

Starter: \ (MRR rights to 30 products)
Standard: \,497 (MRR rights to 50 products)
VIP: \,997 (MRR rights to all 69)

The math: anyone who buys Starter and sells 1-2 units has already made their money back. The math creates trust.

Day 3-5: Build Sprint

Claude Code for drafting. Me for testing and documentation.

Insight from day 4: The 4-hour rule broke. I built a multi-step sales funnel builder - it took 9 hours. I almost cut it. I kept it. It's now one of the highest-perceived-value products because the complexity is visible.

Week numbers:

Week 1: 8 products
Week 2: 14 products
Week 3: 24 products
Week 4: 23 products
Total: 69 in 30 days

Day 6: Whop Setup

Listing 69 products took 7 hours.

Key decision: 40% affiliate commission. At \,997 VIP, that's \/sale for affiliates. High enough to motivate serious promotion.

Day 7: Launch

Revenue: .

Not surprising. Zero audience = zero launch day sales. The product launch is not the moment that matters - distribution is.

What \,997 Buys (And What It Doesn't)

You get: 69 AI systems + MRR license + affiliate support + 30-day refund.

You don't get: An audience. Guaranteed sales. A passive income autopilot.

The people who succeed with MRR: they already have an audience and need inventory. Or they commit to building an audience alongside.

7-Day Scorecard

Day	Task	Hours
1-2	Architecture	8
3-5	Build sprint	30
6	Whop setup	7
7	Launch	2
Total		47 hours

47 hours to build a \,997 product. One sale = \/hour on the build. Ten sales = \/hour.

Full catalog: whop.com/gaiaratio

GaiaRatio | Building in Public | Day 34

The Hidden Cost of Not Systemizing Your Business in 2026 (Real Numbers)

Isak Kim — Sun, 17 May 2026 21:59:58 +0000

Most Entrepreneurs Are Using AI Wrong

Open ChatGPT. Type a question. Get an answer. Close the tab. Repeat tomorrow from scratch.

That's not using AI. That's paying /month to have a conversation with yourself.

McKinsey's 2025 State of AI report quantifies what "using AI wrong" costs: SMBs without systematic AI workflows lose 30% operational efficiency compared to peers who've documented their processes.

Let me translate that into your money.

The Math for Your Business

If your business generates ,000/year:

30% efficiency loss = ,000 in labor equivalency Time equivalent = 600+ hours/year Per week = 11.5 hours you're not getting back Per day = ~1.5 hours evaporating

That's not a productivity hack. That's a structural cost that compounds every quarter.

The Difference Between "Using AI" and "Systemizing AI"

Using AI (what 90% of businesses do):

Ask Claude a question ? get answer ? close tab
Re-explain brand voice to every session
Rebuild the same email template four different times
Trust AI output without a validation step

Systemizing AI (what the 10% do):

System prompt library: brand voice captured once, loaded every session
n8n workflow templates: repeatable tasks automated end-to-end
Output validation SOPs: Claude output passes through a human QC step
Documented edge cases: when Claude should escalate vs. handle

What Systemization Actually Looks Like

Here's a concrete example. Email response handling for a solo creator:

The manual version (45 min/day):
Read email ? Think about response ? Open Claude ? Re-explain context ? Get draft ? Edit ? Send

The systemized version (8 min/day):
Email arrives ? n8n categorizes by type ? Loads appropriate system prompt with full context ? Claude drafts response ? You review + send

Time saved: 37 minutes/day = 3 hours/week = 150 hours/year.

At /hour billing rate: ,000/year recovered from one workflow.

The Three Systemization Priorities

Priority 1: Capture your context (Week 1)
Build a master system prompt that includes:

Your brand voice (specific words you use/avoid)
Your business context (what you sell, who to)
Your output requirements (length, format, tone)
Common edge cases you've encountered

Priority 2: Automate the repeatable (Week 2)
Identify your top 5 recurring AI tasks. Build n8n workflows for each. Estimated time: 3-4 hours per workflow. Expected savings: 1-2 hours/week per workflow.

Priority 3: Monetize what you've built (Week 3+)
The workflows you build for your own efficiency are sellable. Digital product distribution platforms like Whop let you sell SOPs and workflow templates with zero listing fees. Your efficiency investment becomes an asset.

The Time Investment vs. Return

Action	Time to Build	Weekly Savings	Break-Even
System prompt library	3 hours	2 hours	1.5 weeks
Email automation	4 hours	3 hours	1.3 weeks
Content workflow	5 hours	4 hours	1.25 weeks
Client onboarding	6 hours	5 hours	1.2 weeks

Every systematized workflow breaks even in under 2 weeks.

Where to Start

If you want to skip the 6 months of figuring out which workflows to build: I packaged 69 of mine at whop.com/gaiaratio. Plug-and-play. 30-day refund.

Or build your own. The math works either way.

Stop using AI like it's Google. Start using it like it's your most productive employee.

GaiaRatio - building in public since March 2026

I Built 69 AI Products in 30 Days Using Claude Code (Here's the Code, Cost, and Confession)

Isak Kim — Sun, 17 May 2026 21:59:22 +0000

The Number That Started Everything

$13,440. That's what I calculated I lost in 2025 by not having systematic AI workflows.

Not money I spent. Money I didn't make - because I was manually rebuilding the same email templates, re-explaining my brand voice to Claude every session, and treating AI like a search engine instead of a system.

McKinsey's 2025 State of AI report calls this "AI productivity debt." SMBs without systematic AI workflows lose 30% operational efficiency every 18 months compared to peers who've built proper systems.

For a solo creator billing /hour, that's ,440/year. Gone.

I decided to stop leaking it.

The 30-Day Sprint

March 15th. One goal: build 69 AI productivity systems in 30 days using Claude Code as the primary development tool.

Why 69?

Because 70 felt like I was lying, and 68 felt incomplete. Also because I needed a number specific enough to be credible and ambitious enough to force real discipline.

The Stack

Claude Code (CLI) ? Core prompting + documentation n8n (self-hosted) ? Workflow automation Playwright ? Browser automation where APIs don't exist Beehiiv ? Email sequences Whop ? Distribution (0% listing fees)

The Build Process

Each product followed the same template:

Problem statement (1 sentence): What specific task does this eliminate?
System prompt (50-200 words): The Claude instruction set
Workflow SOP (numbered steps): How to use it
Example output (real, not fabricated): What good looks like
Customization guide (3-5 fields to change): Making it yours

Build time per product: 2-4 hours once the template was established.

The Numbers by Week

Week	Products	Notes
1	8	Learning the process
2	14	System locked in
3	24	Template reuse accelerating
4	23	Quality control pass
Total	69	30 days

The Honest Results

Revenue in 30 days: **** (one sale, day 24).

The "Inbox Filter & Reply Bot" - built at 3am during my lowest point of the sprint - sold first. I'd been debugging a more complex product for hours and rage-built this one in 45 minutes.

The lesson I'm still processing: authenticity converts. The products built from genuine frustration outperform the ones I agonized over.

What Claude Code Actually Did

I want to be specific here because "I used Claude Code" can mean anything.

What it actually did:

Drafted initial system prompts (I edited heavily)
Generated documentation templates (I filled with real examples)
Wrote the setup guides (mostly kept as-is)
Suggested edge cases I'd missed (invaluable)

What it couldn't do:

Test the prompts (I did every test manually)
Know what would sell (market intuition is still human)
Replace the 3am debugging sessions (unfortunately)

What's Next

30 days of distribution. Content like this article is part of it.

If you want to see what 69 systematized AI workflows look like - or resell them through MRR rights - the catalog is at whop.com/gaiaratio.

30-day refund on everything. No questions.

Build in public. Ship in private. Distribute everywhere.