Beyond Vibe Coding: Agent Onboarding and Orchestration

2 minutes

Like many people in tech, I’ve enjoyed the vibe coding era over the last year. I built dozens of applications, everything from Chrome extensions to full-blown web apps. It’s been a lot of fun working with programming languages and stacks I’d never touched before, and getting to sample tools like Lovable, Bolt, Replit, and Cursor.

Now with Claude Code, it’s remarkable how easy it is to one-shot entire applications. If anything, some of the magic feels gone because there’s no real prompting craft left to it. Even for more complex agentic coding workflows, getting a single agent to onboard to a codebase, write incremental code and tests, and open a PR feels like a pretty trivial skill at this point.

I think the next evolution in AI-driven development is going to be around agent onboarding and orchestration.

Agent onboarding as I imagine it, is the art of quickly getting an AI employee up to speed on what you want it to accomplish. It means giving it full context, objectives, and goals, and creating the preconditions for one-shotting complex tasks. For coding agents, this is already fairly straightforward. But for other functional areas like marketing, data analysis, and legal, the onboarding piece can be time-consuming and complicated. Being an expert at onboarding means your agents produce better quality output, faster.

Orchestration is the next frontier beyond that, where you move from having a single agent working for you to dozens of agents working in parallel. The mental models and harnesses for managing orchestration at that scale are still pretty immature today, but I think they’ll develop significantly over the rest of this year. The end-state skills that I believe will matter most for orchestration are, interestingly, the same ones that matter for managing people in an organization:

  • Being a clear communicator
  • Being able to delegate tasks effectively
  • Setting organizational structures that are well-suited to getting goals and objectives accomplished

By the end of the year, someone who’s great at onboarding and orchestration could build a massive company of one person with thousands of AI employees. This is probably already happening to some degree, but will become more common. I’m not sure exactly how it all takes shape, but it’s helpful to imagine some ways it could manifest itself.

  • AI employees communicating across thousands of email accounts, creating a record of work streams just like in a real company.
  • Slack-like chat environments where agents collaborate asynchronously.
  • Detailed onboarding protocols optimized to give each new agent exactly the right context and depth of training for the complexity of the tasks they’re being given.

The amount of leverage that’s going to be possible is truly mind-boggling.