Skip to content

Recent articles

Insights on modern web development, performance, and building better digital products.

7 min read

Why Focused Landing Pages Convert Better

Most landing pages lose conversions by adding more. Nine cognitive principles and Princeton's processing fluency research explain what to remove instead.

6 min read

Context Substrates for Long-Running Agents

Production agents often work better with reliable context retrieval across sessions than with persistent memory. Here is the two-camp taxonomy that separates the two markets, with the public signals most buyers are missing.

9 min read

Hermes Agent in Production: the Day-30 Problem

Most Hermes Agent 4-profile teams work on day one and show signs of voice convergence by day 30 in production deployments. The operator layer that prevents it: handoff contracts, memory-KPI audits, and policy gates per role.

9 min read

Why Most AI Agent Setups Run Far Slower Than They Could

The architecture around the model explains the gap between modest and substantial AI productivity outcomes. Five builders published the same thesis in one week.

8 min read

The AI Knowledge Layer: 127 Pages, Zero Vector Databases, and What I Got Wrong

Karpathy's LLM wiki gist hit 99,000 bookmarks in a week. It resonated because it named what every AI user feels: your agents have no memory. I run a knowledge layer in production. Here is what works, what does not, and how to build one in 20 minutes.

7 min read

Every SaaS Becomes an Agent Harness

A sales team replaced an entire sales SaaS with one Claude skill and three MCP servers. This pattern will define which software categories survive 2026.

7 min read

From Rules to Results: What 22K Stars on a Single CLAUDE.md Reveal About AI-Assisted Development

The karpathy-skills repo proves that AI coding bottlenecks aren't about model capability. They're about the behavioral contract between human and LLM.

7 min read

Why AI-Generated Software Still Needs Engineering Review

Andrej Karpathy coined "vibe coding" in February 2025. Since then, a wave of AI-generated apps has shipped that work in demos and break in production. The problem is using AI tools without engineering discipline.

6 min read

Anti-Slop Content Strategy: Why LLMs Won't Cite What They Can Already Generate

If ChatGPT can write your article from the title alone, it won't cite it either. Here's the content framework that optimizes for LLM citations instead of backlinks.