Launching an AI Investment Tool: Lessons from 6 Months of Building KeepRule

Six months ago I started building KeepRule (https://keeprule.com), an AI-powered platform that helps investors apply principles from Buffett, Munger, and other great investors. Here's what I've learned — the good, the bad, and the things I wish someone had told me.

Why This Product

I'm William, and I've been at the intersection of AI and finance for a while. The insight that led to KeepRule was personal: I kept making the same investing mistakes despite "knowing better." I'd read all the books, understood the principles, but in the heat of the moment — when a stock was crashing or a hot tip landed in my inbox — I'd abandon my rules.

I realized this wasn't a knowledge problem. It was an execution problem. And execution problems are exactly what software can solve.

The Tech Decisions That Mattered

Go for the Backend (Go-Zero Framework)

I chose Go over Node.js or Python for the API layer, and I'd make the same choice again. Go-Zero gave me a clean, opinionated structure: define your API in a .api file, auto-generate handlers and types, write your logic. Zero decision fatigue on architecture.

Performance-wise, the Go backend handles concurrent AI requests beautifully. When a user is chatting with the AI while another is loading scenarios, the server doesn't break a sweat.

Vue 3 + Nuxt for SSR

SEO matters enormously for this product. People search for things like "Buffett margin of safety rule" or "value investing checklist" — and I need those pages to be indexable. Nuxt's server-side rendering was the right call.

TypeScript was non-negotiable. The data structures in an investment app are complex (rules, categories, tags, user notes, scenarios), and type safety has caught countless bugs before they shipped.

Claude API for AI Features

I tested several LLMs for the "chat with investing masters" feature. Claude won for two reasons: it follows instructions consistently (critical when you need the AI to stay in character as an investment advisor), and its reasoning about nuanced financial concepts is genuinely good.

The prompt engineering took longer than I expected. Getting the AI to give thoughtful, principle-based responses — without giving actual financial advice or being overly cautious — required weeks of iteration.

User Acquisition: What Worked

Content-Led SEO (Slow but Compounding)

I write detailed content targeting long-tail investment queries. Posts like "How to calculate margin of safety step by step" or "Buffett's rules for when to sell a stock" bring in highly qualified organic traffic.

The key insight: people searching for these terms are actively trying to become better investors. They're the exact audience for KeepRule. Conversion rates from organic search are 3-4x higher than from social media.

GitHub Gists and Dev.to Articles

Surprisingly effective. I published Python scripts for stock valuation, portfolio tracking, and dividend analysis — all with natural mentions of KeepRule. Developer-investors are a real audience, and they share useful code.

What Didn't Work

Product Hunt — I launched too early. The product wasn't polished enough, and I didn't have enough existing users to rally votes. If I did it again, I'd wait until I had a stronger user base and a more compelling demo.

Twitter/X Threads — The investment content space on Twitter is dominated by accounts with huge followings. My threads got minimal traction. I've since focused on platforms where content quality matters more than follower count.

Cold outreach — I tried reaching out to finance bloggers and YouTubers. Response rate was near zero. Warm intros and genuine community participation work much better.

Product Iteration: The Surprises

Users Want Interactive, Not Static

My first version was essentially a well-organized content library — investment principles categorized by topic and master. It was useful but not sticky.

The game-changer was adding interactive elements:

• AI chat: Users discuss their actual investment decisions with a Buffett-trained AI
• Scenarios: Real-world investment dilemmas where you practice applying principles
• Bilingual support: I added Chinese (zh) content and saw a huge spike from Chinese-speaking investors

The Feature I Almost Didn't Build

Scenario-based learning was a "nice to have" on my backlog. I almost didn't build it. Now it's the most engaging feature by session duration. Users spend 3-4x longer on scenarios than on static principle pages.

Lesson: sometimes the features you think are secondary turn out to be the core product.

Pricing Is Hard

I'm still figuring this out. The free tier needs to be valuable enough to attract users but limited enough to motivate upgrades. Currently experimenting with:

• Free: Browse principles, limited scenarios
• Premium: Full AI chat, all scenarios, personal rule-building

What I'd Tell Past Me

1. Ship the interactive features first. Static content is table stakes. The AI chat is what makes people say "wow."

2. SEO from day zero. Every week of delay is a week of compounding you'll never get back. Start publishing content immediately, even before the product is ready.

3. Build in public, but selectively. Share your journey on communities like IndieHackers, but don't spread yourself across 10 platforms. Pick 2-3 and go deep.

4. Your first 10 users will teach you more than 100 hours of research. I spent too long in "research mode" and not enough time talking to real people.

5. Niche down harder than you think. "Investment tool" is too broad. "AI-powered rules-based investing for individual investors who follow Buffett's principles" — that's specific enough to build for.

Current State and What's Next

KeepRule is live at keeprule.com with a growing user base. Revenue is early but the metrics I care about — retention and engagement depth — are heading in the right direction.

Next priorities:

• Personalization: Let users build and track their own investment rules, not just browse curated ones
• Mobile optimization: Most of my traffic is mobile, and the experience needs work
• B2B exploration: Financial advisors have reached out about using KeepRule with clients

If you're building in the AI or fintech space, happy to chat. This community has been invaluable for perspective.

What's the hardest product decision you've faced recently? Would love to hear from others in the trenches.