Insights for AI builders
Tutorials, product updates, and ideas to help you build and ship AI applications faster.
Subscribe via RSS
How to Use AutoResearch to Optimize Any Business Metric Autonomously
AutoResearch runs experiments in a loop to improve any measurable metric—cold email reply rates, landing page conversions, ad copy—with zero human involvement.
What Is the Averaging Cost Problem in AI Teams? Why More Stakeholders Produce Worse Outputs
The averaging cost problem explains why group decisions in AI-assisted work produce mediocre results. Here's how to structure teams to avoid it.
What Is Claude's Agentic Operating System? How Skills Chain Into Business Workflows
Claude Code skills become most powerful when connected into systems. Learn how shared brand context, memory, and chained skills create an agentic OS.
How to Use GitHub Actions to Run AutoResearch Experiments on a Schedule
Deploy an AutoResearch loop to GitHub Actions to run A/B experiments on cold email, landing pages, or AI skills automatically every hour without a server.
What Is the Judgment Density Framework? How to Identify AI-Ready Talent on Your Team
Judgment density, conviction velocity, and execution bandwidth are the three qualities that predict who will thrive with AI agents. Here's how to spot them.
What Is the Learnings Loop? How Claude Code Skills Improve From Your Feedback
The learnings loop lets Claude Code skills update their own instructions based on your feedback. Here's how it works and why it matters for AI workflows.
How to Build a Self-Improving AI Skill System for Marketing and Content Creation
Chain Claude Code skills with shared brand context, a learnings loop, and eval scoring to build a marketing system that improves automatically over time.
How to Build a Self-Maintaining AI System with Heartbeat and Wrap-Up Skills
Learn how to build an AI system that syncs itself automatically using heartbeat scans and wrap-up skills inspired by OpenClaw's memory architecture.
Shared Brand Context vs Context Folder: The Two Memory Layers Every AI System Needs
Understand the difference between static brand context and dynamic context folders in agentic AI systems, and why both are essential for reliable outputs.
What Is Speed of Control? The Attention Management Skill That Unlocks AI Agent Performance
Speed of control—making high-quality decisions quickly across multiple agents—is more important than span of control. Here's how to develop it.
What Is Taste vs Conviction in AI-Assisted Work? The Skill Gap Nobody Talks About
Taste helps you evaluate AI outputs. Conviction is what makes you ship. Learn why conviction is the missing skill for getting real value from AI tools.
Gemini Embedding 2 and the End of Stitched-Together Embeddings
Why Gemini Embedding 2 matters: a primer on embeddings and how a unified vector space replaces the brittle stitching of separate text, image, and audio models.
What Is Nvidia Nemotron 3 Super? The 120B Open-Weight Model You Can Fine-Tune
Nvidia's Nemotron 3 Super is a 120B parameter open-weight model available on Perplexity, Open Router, and Hugging Face. Here's what makes it worth knowing.
What Is Perplexity Computer? The AI Agent That Runs Tasks on Mac Mini Hardware
Perplexity Computer is an AI agent that runs on dedicated Mac Mini hardware to handle tasks like recruiting, slide decks, and marketing analysis 24/7.
What Is an AI Coding Agent Harness? How Stripe, Shopify, and Airbnb Build Reliable AI Workflows
Enterprise teams at Stripe, Shopify, and Airbnb are building structured AI workflow engines. Here's what they are and how to apply the pattern yourself.
How to Use the AutoResearch Loop for Cold Email Optimization with GitHub Actions
Connect your cold email platform API, define a reply rate metric, and run an autonomous challenger-baseline loop on a schedule using GitHub Actions.
How to Use AutoResearch to Optimize Landing Pages and Ad Copy Autonomously
Apply Karpathy's AutoResearch loop to marketing: set a conversion rate metric, connect your platform API, and let agents improve your copy overnight.
ChatGPT for Excel: How to Use AI to Build and Update Spreadsheet Models
ChatGPT now integrates directly into Excel as a sidebar. Learn how to use it for data analysis, budgeting, and live spreadsheet model updates.
What Is Claude's On-Demand Generative UI? How It Differs from Canvas and Artifacts
Claude can now build interactive applications inside your conversation on the fly. Learn how generative UI differs from canvas features and image generation.
What Is ComfyUI App Mode? How to Turn Complex Node Workflows Into Simple Interfaces
ComfyUI's new App Mode converts spaghetti node graphs into clean, user-friendly interfaces. Learn how to build and share apps without touching the nodes.
What Is Domain Expert Building? How Non-Coders Are Becoming Builders with AI
Doctors, teachers, and logistics managers are now building custom software with AI. Learn how the translation layer between expertise and code is disappearing.
Gemini in Google Docs, Sheets, and Slides: What You Can Actually Do
Google's Gemini is now embedded in Docs, Sheets, and Slides for paid users. Here's what it can do and how to use it to speed up your work.
GPT-5.4 vs Claude Opus 4.6 vs Gemini 3.1 Pro: Real Benchmark Results Compared
Side-by-side benchmark results for GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro across coding, creative writing, research, and SVG generation tasks.
What Is Jevons Paradox in AI? Why Cheaper Intelligence Creates More Demand for Human Work
Jevons Paradox explains why AI efficiency gains expand demand rather than shrink it. Here's what this means for your career and business strategy.