The Day at a Glance

• Seven projects got real attention today, and context switching was the strategy, not the enemy
• Moved automation between machines and spent the afternoon debugging paths that worked perfectly five minutes ago
• The WSJ says AI makes work more intense, not lighter. My Saturday is Exhibit A, and I don’t think that’s bad
• A security researcher breached a major consulting firm’s AI platform through SQL injection from 1998
• Codie Sanchez spoke to 2,000 people recently. Ten were using agentic AI

The Real Bottleneck Was Never the Work

Seven different projects got meaningful attention today. Client work, personal brand, content systems, automation infrastructure, media asset pipelines. A year ago, touching three projects in a day meant doing all of them poorly. The tax wasn’t the work itself. It was the mental overhead of switching between contexts. Remembering where you left off, what mattered, what was blocked.

That tax is basically gone now.

Not because the projects got simpler. Because AI carries the context for me. Every time I pick up a thread, the full state of that project is already loaded. The decisions I made last Tuesday, the files I changed, the problems I was circling. So “switching” from a client deliverable to setting up a content publishing system to debugging path references in an automation script doesn’t feel like switching anymore. It feels like one continuous day of work that happens to span different domains.

The WSJ ran a piece recently arguing that AI isn’t lightening workloads, it’s making them more intense. After a day like today, I think they’re describing the symptom but missing the diagnosis. The intensity isn’t coming from AI demanding more of you. It’s coming from AI removing the friction that used to force you to slow down. When the barriers between projects dissolve, you naturally fill the space. Whether that’s a problem or a superpower depends entirely on whether you’re building something you care about.

Systems That Outlast the Session

The most important work today wasn’t any single deliverable. It was the shift from doing tasks to building systems that handle tasks on their own. Content that publishes itself. Logs that sync across machines automatically. Approval workflows that wait for a thumbs-up before going live.

Nobody warns you about the next part, though. Moving automation between environments is where the real debugging lives. Code that works perfectly on one machine breaks in subtle ways on another. A path reference that assumed a directory structure, an environment variable that existed in one place but not the other. These aren’t hard problems individually. But they’re the kind of thing that makes you realize the gap between “I built a demo” and “I built something that runs reliably without me” is wider than it looks.

That gap is where the craft actually lives now. Anyone can get AI to generate a script. Getting that script to run unattended, across machines, handling edge cases gracefully, recovering from failures. That’s the part that compounds.

From the Vault

A stat caught my eye this week: 6.8 billion people still haven’t used AI. The startup world is fighting over the 16% who are already onboard, building increasingly sophisticated tools for people who already know what a prompt is. Meanwhile, 84% of the market wants specific painful tasks to disappear. No interface. No learning curve. Just the problem going away.

The content systems I built today are a small version of that. Once the pipeline runs, the output just shows up. The interesting question is what that looks like at scale, for teachers, for small businesses, for the local services that make a city work.

Separately, a security researcher breached a major consulting firm’s internal AI platform through basic SQL injection. Two hours, millions of messages exposed. An AI-powered consulting firm getting hacked through a vulnerability from 1998. It’s a reminder that the unsexy fundamentals don’t stop mattering just because the technology is exciting.

One more: Codie Sanchez spoke to 2,000 people recently. Ten were using agentic AI. About 20% had heard of Claude. We’re so deep inside this world that it’s easy to forget how early all of this still is.

The day ended with more running systems than it started with. Not what I shipped today, but what will still be running tomorrow without me touching it.