Field notes

Codepet journey

Field notes from building AI products in the open: what real users teach us, and how to grow a second brain that thinks with you.

Building AI ProductsFeatured

Streaming AI Responses: Why Fast Feels Smarter

Streaming makes your AI app feel dramatically faster without changing the model. Here's how to add it, what breaks along the way, and when not to bother.

July 6, 20266 min readRead more

Insights in your inbox

Free emails about the latest Codepet posts on building products with AI. Unsubscribe anytime.

The Aha Moment: When AI-Assisted Coding Finally Clicks

Most new developers hit a wall before AI coding feels natural. Here's what we've observed about the moment it all clicks — and how to get there faster.

July 5, 20265 min read

Building AI Products

How to Scope an AI Feature Before You Build It

Most AI features fail before they're built — because the scope was wrong. A practical framework for defining what your AI should do, how good is good enough, and when to stop.

July 4, 20265 min read

Codepet journey

Streaming AI Responses: Why Fast Feels Smarter

Insights in your inbox

The Aha Moment: When AI-Assisted Coding Finally Clicks

How to Scope an AI Feature Before You Build It

The Decision Log: Stop Relitigating Choices You've Already Made

How to Give Your AI App Memory

Spaced Repetition and AI: Make What You Learn Actually Stick

How to Write a Product Spec with AI Before You Write Any Code

How to Use AI to Turn User Feedback Into Product Decisions

What Happens When You Actually Ship: The First-Time Builder Experience

When to Use AI Agents — and When a Simple LLM Call Is Enough

Structured Outputs: The Reliability Layer Your AI Product Is Missing

How to Design AI Features That Users Actually Trust

How to Run a Weekly AI Reset (and Stop Your Project Drifting)

Comprehension Debt: The Hidden Cost of Coding With AI

LLM Tool Calling: A Practical Guide for AI Product Builders

Context Window Management for AI Product Builders

How to Give AI Better Context and Get Answers Worth Acting On

How to Debug Code with AI: What Works, What Doesn't, and Why

Prompt Versioning: Treat Your Prompts Like Production Code

How to Build an AI Learning Loop That Actually Sticks

AI as a Crutch vs. AI as a Coach: The Habit That Builds Real Skills

How to Write System Prompts That Actually Work in Production

How to Build a Personal RAG System for Your Notes

Why New Builders Hit a Wall — and How the Best Ones Push Through

How to Write LLM Evals for Your AI Product

What Is Vibe Coding, and How to Do It Without Shipping a Mess

Ask AI to Review Your Code, Not Write It

How to Use AI as a Thought Partner, Not a Search Engine

Measure Before You Optimize: The Codepet Bug That Proved Us Wrong

Building an AI Product in Public: What Codepet's First Year Taught Us

Beyond Autocomplete: Building a Second Brain That Thinks With You

What 500 Beginners Taught Us About Learning to Code with AI