Key Takeaways

  • Ryan Carson eliminated resumes, phone screens, and live culture calls to hire Untangle's first engineer, replacing them entirely with unedited desktop screen recordings.
  • The hiring bar focuses on how candidates structure prompts, manage latency, and correct AI coding agents rather than their ability to write manual syntax.
  • Candidates who pass initial review get access to Untangle's live Devin agent environment to build a real feature before receiving an offer.
  • Claire Vo observed that Carson evaluates human hires the same way teams test autonomous agents: through deterministic verification loops instead of conversational rapport.
  • This entire evaluation structure is codified in Carson's Asynchronous AI Operator Hiring Funnel.

The Carson's Asynchronous AI Operator Hiring Funnel

  • Phase 1: Full-Screen Workflow Audition: Candidates record an unedited, full-desktop video demonstrating how they use AI coding agents to build and ship a real feature for an existing application. No resume review, phone screens, or culture-fit calls take place.
  • Phase 2: Agent Management Evaluation: Evaluate candidate submissions specifically on how effectively they prompt, correct, and supervise agents, assessing error handling, tool harness usage, and execution speed.
  • Phase 3: Production Sandbox Trial: Shortlisted candidates receive access to the production agent workspace (e.g., Devin environment) to complete a scoped company ticket while capturing video replay telemetry.
  • Phase 4: Brief Verification Call and Offer: Hold a single, brief introductory meeting to confirm logistics, followed by immediate hiring based on verified agent output.

When This Works (and When It Doesn't)

Carson's hiring funnel is built for teams where engineers spend their days orchestrating 10 to 15 autonomous agents in the cloud rather than typing raw code in an IDE. When software output depends on how fast an operator spots hallucinations, debugs broken test suites, and steers background runs, traditional whiteboard algorithms tell you nothing. As Carson explained: “All I want you to do is record a video full screen, everything, your whole desktop of you building a new feature for an app that already exists. And I just want to see the whole thing. And I don't want to talk to you. I don't want to have a meeting with you. I don't want to get to know if I like you. I just want to see how good of an agent manager you are.”

This method breaks down if you are hiring for roles that require deep low-level systems programming, custom hardware integration, or complex organizational consensus. If an engineer must align five cross-functional teams before writing a line of architecture, completely skipping conversational interviews leaves real blind spots. But for early-stage solo founders building AI-native products, conversational interviews are often an expensive distraction from actual execution skill. “Think about how much better this is in this insane way we used to hire,” Carson pointed out. “You like I like this person... and then you're like we're going to kind of ask you to do some weird technical project.”

What to Do With This

If you are hiring your next technical operator this week, delete your standard 30-minute introductory Zoom link and update your job post with a single audition prompt. Ask candidates to submit a Loom or unedited screen capture showing them adding a concrete feature to an open-source codebase using their agent setup of choice.

Review the first ten minutes of the video on 1.5x speed. Ignore whether their voice sounds polished. Look at their screen layout: Do they let the agent run blindly down rabbit holes, or do they catch drift early? Do they craft clear system instructions, or do they fight basic error loops? Invite the top two performers into a shared workspace with temporary access to Devin or Cursor, assign a real backlog ticket, and extend an offer to whoever ships clean code to production first.