The tools accelerate you past the busywork to the crux of the problem. And at the crux, you need other people. The tools do things for us, but somehow they’re making us do things together more than we did without them.

We called it a pitch pause. Two weeks where product and engineering at Validic got dedicated time to stop, learn, and build with agentic AI tools. 43 repos. A happy hour. And a conviction that creative exploration needs to become part of how we work.

We spent seventy years making code more readable. Machine code to assembly to C to Python to English. The ladder doesn’t go higher. What happens below it is the interesting question.

Teams are using Claude Code to build apps, then punting the hard parts to LLM API calls instead of engineering solutions. The tool that helps you build faster can also help you skip the thinking faster.

Everyone building with AI independently gravitates to the same projects: meal planners, knowledge bases, Slack plugins, email triage. The pattern is funny. What it says about build vs. buy is not.

Workflows imply linearity. Working with agentic AI doesn’t feel like that anymore. It’s loops - individual and concurrent. The trap is mistaking motion for thought. The real work is getting both you and the AI into a flow state before the loops start.

A step-by-step walkthrough of prompt injection using a real example: an AI agent that reads Jira tickets and writes Confluence docs. What could go wrong, what it looks like, and what your team should do about it.

I blocked off an entire day for open office hours on agentic AI. Sales, product, engineering, support all showed up. What I watched wasn't people learning a tool. It was people dusting off ideas they'd been carrying around for months.

Context engineers. Agent orchestration architects. Codebase ergonomics specialists. The unit of expertise is moving up a level of abstraction, and most of the roles that will matter don't have names yet.

I spent 3 months using Claude Code on 40+ projects. Then I asked Claude to go through all the conversation history and write blog posts about what happened. 17 posts covering the journey from typing "check out this git repo" three times to orchestrating eight research agents in parallel without writing a line of code.

Written by Claude. Reviewed by the human. About the collaboration between them.