We just wrapped our yearly all-agency conference. It's a chance to get the full team together and share knowledge across the disciplines that make up a full-service ad agency. This year I ran a workshop where we gave teams a business and asked them to take it from idea to functional prototype in an hour. I also did a talk about finding the repeatable tasks hidden inside larger everyday work that are actually suited for automation or AI assistance.
The reaction that stuck with me was the genuine surprise when people saw what a well-structured system could produce. Being in engineering it's easy to see the ways harnesses and systems improve the output, and easy to forget not everyone is seeing that as clearly in their day-to-day work. People were asking questions and figuring out where it fits in their own work. That kind of interest is worth building on.
Something that stuck out to me is that we have champions across departments who have started using AI tools in their day-to-day. They're testing, iterating, and finding what works. But we don't have a formalized structure for bringing that to the wider teams. Right now it's a few Slack channels and hope.
Solving that has been sitting in the back of my mind ever since. Coming from a WordPress background, the core contributor team model comes to mind. Instead of leaving adoption to whoever's motivated enough to self-direct, you build teams around defining how tools get used and spreading that to the wider group. Building out the agentic systems and getting them into more hands.
Because AI tools are only as good as the systems behind them. And people aren't going to adopt tools that aren't making their work better.
📖 Read
Welcome to the personal software revolution
The Verge
My favorite part of vibe coding is being able to dream up an idea and have a working app in a few compute cycles. It may not be public ready, but sometimes you just need an app for a specific use case. As these things get better and we put systems around them, anyone is going to be able to build their own single-purpose apps.
Turns out managing a team of humans and managing a fleet of agents requires the same core skills. Breaking down work, setting context, and defining what is actually good. Engineering managers are suddenly in demand as ICs because they already know how to delegate tasks. Now they're just doing it to agents.
🔧 Tool
dcramer/dex: Task tracking for Agents
GitHub
As teams adopt agentic engineering practices it's going to be important that many people and agents can work across a codebase without duplicating work. Dex is an open source CLI tool that puts some structure behind task tracking for coding agents. Whether it's a team or your own personal fleet.
📖 Read
AI Handbook — WordPress AI
WordPress
WordPress stood up a formal AI contributor team with a public handbook covering workflows and best practices for AI in core. Read it less for the WordPress specifics and more as a model for how an organization could formalize AI practices across a distributed group with wildly different skill levels.
📖 Read
Google's Guide to Optimizing for Generative AI Features on Google Search
Google Search Central
Google still dominates search. So when they post about how to optimize for their own generative AI features, it's worth taking note and implementing.
Thanks for reading, Jason
