Skip to content
The Observatory

Building & Agents · 2 min read

Hire the Risk Away Before the Agent Runs

The best teams are not résumé collections — they are purpose-built instruments for eliminating the specific risks that will kill a venture before it ships. Vinod Khosla's "Gene Pool Engineering" makes this concrete: enumerate your biggest risks first, then hire the exact people who have retired each one before.

Featuring · studied & reframed for AI-native builders“Gene Pool Engineering” — Vinod Khosla ↗

Map Your Risks Before You Map Your Org

Most founding teams assemble around enthusiasm or availability, not around a hard-nosed list of what will actually fail. The Khosla method flips that. Before you post a single role, write down your five likeliest causes of death — technical, regulatory, distribution, trust, whatever fits your terrain. Each gap on that list is a hiring target, not a job description waiting to be dressed up.

In Agentic Systems, the Risks Are Specific

When agents do the work, your failure modes are not generic software failure modes. They are: agents taking wrong branches because no evaluation harness caught the edge cases, tools scoped too loosely so agents wreak havoc with excess permissions, and context handed to a model so carelessly that it halts or hallucinates. The gene pool you need has retired those risks specifically — people who have built evals before shipping, who have designed tight tool surfaces, who know that a workflow is not an agent and that conflating the two is how production systems fail quietly.

Judgment Is a Gene You Cannot Fake

Even a well-engineered agentic stack requires humans who can set goals, make tradeoffs, and own accountability when the system surprises you. That capacity is not a soft add-on; it is a primary risk in any autonomous company. When you engineer your gene pool, explicitly hire for demonstrated judgment under uncertainty — someone who has held a hard line when an autonomous system pushed past its boundaries. That track record retires a risk that no clever prompt or eval loop can retire on its own.

The List Is Your North Star, Not the Org Chart

An org chart describes the company you already have. A risk list describes the company you need to survive to build. Write the list first, treat each item as a hiring constraint, and resist the temptation to fill a role with someone impressive but mismatched to your actual threats. An AI-native company designed around agents and evals from day one does not have the slack to carry talent that addresses risks you do not have.

Try the Desktop OS

Explore Atlanta’s startup scene, live.

Everything you just read lives inside our Desktop OS — an interactive workspace that opens right in your browser. Its Atlanta desk maps the city’s startup scene: who’s building, what’s happening, and where. Open it and take a look.

Open the Desktop OS →