The Day at a Glance

• Eight projects in one day, and the thing that broke wasn’t the AI or the code
• Scattered .env files revealed the real cost of moving fast across two machines
• Stopped shipping to spend the day on infrastructure nobody will ever see
• A WSJ piece nailed it: AI isn’t lightening workloads, it’s making them more intense
• The question I can’t shake about what happens when the management layer becomes its own job

The Bottleneck You Don’t See Coming

Eight projects today. Client work spanning healthcare, financial services, SaaS, and enterprise. Personal brand work. An AI assistant that needed debugging. A media asset repository that needed to exist on two machines. A brand new site spun up from a template in minutes.

None of that is the interesting part.

The interesting part is what broke. Not the AI. Not the code. The organization layer. Somewhere between projects three and five, my environment configuration was scattered across multiple locations. Different machines, different files, different assumptions about where things lived. Everything worked, technically. But the seams were showing.

So I stopped building and started consolidating. One source of truth for environment variables. One deployment package that works regardless of which machine runs it. The unsexy work that nobody writes blog posts about, except here I am, because this is the part that actually matters.

Infrastructure Days

There’s a temptation, when AI lets you move this fast, to just keep shipping. New site? Twenty minutes. New automation? An hour. Content across five platforms? Afternoon project. The speed is real. But speed without structure is just organized chaos with good PR.

Today was an infrastructure day. Cross-machine sync. Deployment packages audited and fixed. Path references that assumed one machine corrected to work on another. A daily automated content pipeline wired up end to end. None of it produced a visible output for anyone except future me.

A Wall Street Journal piece I bookmarked recently nailed this tension: AI isn’t lightening workloads, it’s making them more intense. That tracks. The volume of what’s possible in a day has exploded, but the organizational overhead to sustain that volume hasn’t automated itself. You still need to build the scaffolding.

The Adoption Gap Is a Mirror

While consolidating today, I kept thinking about a stat I saved recently: only about 16% of people have meaningfully adopted AI tools. The other 84% aren’t waiting for better models. They’re waiting for someone to make specific painful tasks disappear. No prompts, no learning curve, just results.

That gap felt personal today. Even on the builder side, the constraint isn’t capability anymore. My AI can write, research, debug, deploy, and manage context across wildly different domains in a single session. The constraint is the human systems around it. The file structures, the deployment configs, the environment variables that assume you’ll always be on the same laptop. The boring stuff.

From the Vault

A few ideas that crossed my desk and stuck around:

“Don’t mistake AI visibility for actual control.” Someone made the point that AI dashboards create a false sense of oversight. Five-person teams now outproduce thirty-person teams, but only if those five people have real systems underneath them, not just monitoring screens. Felt that one today.

“Be careful what you practice.” Jordan Peterson’s idea that your brain constantly offers easier alternatives that feel productive but dodge what actually matters. Infrastructure days are the antidote. They’re never the thing you want to do. Always the thing that makes everything else work.

“Software engineers going extinct by 2026.” A bold claim from someone who should know. Whether or not the timeline is right, the direction is clear.

The question I keep circling back to: at what point does the system you’ve built to manage AI output become its own full-time job? And when it does, what does AI do about that?