Your job isn't repeatable, but that doesn't mean all parts of your job aren't. The scaffolding around everything you do almost always is surrounded by tasks that are. Things like organizing, summarizing, formatting, etc. That's what most people miss.

Maybe they played with ChatGPT during the early rush and sent a few prompts and got some good results, some bad results, and some fun anime versions of their dog. But if you stopped there you're missing how the tools have evolved.

You need to be thinking about how to apply AI tools and processes as systems.

Developers have (mostly) always done this. Write code, test, iterate, refactor, debug, drink coffee, write more code, curse, write code, ship, debug, ship again, drink beer. The system is built into the job. A lot of other work doesn't have that infrastructure baked in, which is exactly why engineering was one of the first fields to be disrupted. (Of course it doesn't hurt they were working on the technology themselves too.)

The people getting the most out of AI right now are the ones bringing systems thinking to work that wasn't traditionally systematized.

Vibe coding was level one. Throwing prompts and full codebases of content at the model and hoping. Agentic engineering is what comes next. It's all about wrapping processes around the work to leverage the things AI is good at and build code around the things it isn't. With a knowledgeable human orchestrator in the middle. Guiding the bots and applying taste and judgement.

That shift isn't staying with developers. It's spreading into writing, design, product management, and anywhere people are willing to think about the systems that make up their work.

The links this week are a signal in that direction.

📺 Watch
Andrej Karpathy: From Vibe Coding to Agentic Engineering
YouTube

The man who named vibe coding is now explaining what comes after it. If you've been treating AI like a faster way to write code, you're still on level-one. The next step is building systems and putting yourself in the middle as conductors guiding and crafting the outputs that way.

📖 Read
The Wiki That Writes Itself
Extended Brain

Agentic engineering for knowledge. I've been thinking a lot about the self-authoring wiki approach and the various use cases for it. I've been experimenting with using it as a way for organizing the topics I want to learn about and see opportunities to expand it to onboard team members to work.

🎨 Design
Introducing Claude Design by Anthropic Labs
Anthropic

For systems thinking bleeding outside of engineering look no further than Anthropic's foray into disrupting design with the aptly named Claude Design. It is early, but the more use they get the better they'll be able to tune the tool. Those thinking about how to put tools like this into their workflows early will gain a distinct edge.

📺 Watch
Collaborative AI Engineering: One Dev, Two Dozen Agents, Zero Alignment
Maggie Appleton, GitHub

Once you have a handle on agentic engineering the next step is going to be how to do it in a way that works for teams. We're still going to get the best results by working with other people and bringing in a wide array of views and ideas. GitHub is exploring what that looks like in a land where code is fast and feedback can come from many people.

🔧 Tool
Cursor Agent Kanban
Cursor

Cursor has a new approach to guiding your agents. Give them a kanban board directly in Cursor where you can assign tasks by dragging around cards and let the agent take over from there. See it in action.

Thanks for reading,

Jason