Blog
Thoughts on software, AI, and building things that matter.
The ROI of AI Agents: How to Measure What Your Automation Is Actually Worth
The number one thing missing from most AI implementations is a measurement plan. Teams build something, it sort of works, and then six months later no one can answer whether it was worth the investment. That’s not a technology problem....
What to Expect From an AI Consultant (And What You Shouldn't Have to Tolerate)
Most businesses hiring an AI consultant for the first time have no idea what they’re buying. That’s not a knock — it’s a structural problem. AI consulting is new enough that there’s no shared vocabulary for what “good” looks like,...
SMBs Are Canceling Salesforce. Here's What's Actually Happening.
The Information reported this week that five small companies, each with 20 to 70 employees, canceled their Salesforce and HubSpot contracts over the past six months. They replaced them with applications they built using Claude, Lovable, and Replit. Software costs...
What Anthropic's Claude for Small Business Gets Right (And What It Misses)
On May 13, Anthropic launched Claude for Small Business — prebuilt agentic workflows wired into QuickBooks, PayPal, HubSpot, Canva, DocuSign, and Google Workspace. The pitch is that 33 million small businesses can now access the kind of AI automation that...
Karpathy Bet on Anthropic. That Tells You Where the Operational Edge Is Going.
Quick context if the name is new. Andrej Karpathy was on the founding team at OpenAI, ran AI at Tesla, and taught the free course “Neural Networks: Zero to Hero” that trained a large chunk of the people now building...
AI Agent Teams Are Having a Moment. We've Been Running One for Months.
A tool called Paperclip launched in early March and hit 30,000 GitHub stars in under three weeks. The pitch: give your AI agents an org chart, goals, and budgets, and watch them run your company with zero human intervention. The...
What Agent Orchestration Tools Get Wrong
There’s no shortage of agent orchestration frameworks right now. CrewAI, AutoGen, LangGraph, dozens of others. They all promise the same thing: give your AI agents roles, let them collaborate, watch the magic happen. I’ve been running a multi-agent AI team...