Why Does DARPA Work?

I've had this article in my reader since it came out in 2020. It's in my "constant re-read rotation" primarily because it has rich detail on a few practices that run against my grain. It's also one of the most thorough reviews of an organization's structure and ways of working that I've ever seen.

Have a look!

Why does DARPA work?

🤖🖥🛰🔬💡🤔

Ben Reinhardt

Opacity enables outlier successes

My "normal" stance on transparency vs. opacity is something like, "most organizations have less internal transparency than they need, leading to mistrust, CYA-ing, and hours spent 'finding the truth' instead of 'getting important things done.'"

The opposite at DARPA:

Opacity removes incentives to go for easy wins or to avoid being criticized by external forces. Reporting requirements also add friction around everything from hiring to changing direction to trying crazy things to moving quickly and more.

Opacity is a prized possession, but it's usually got a different name: "Air cover." It does the things that it says on the tin, above ⬆︎. But opacity is almost always abused. It usually isn't in DARPA for a few reasons:

1. High social pressure ("you’re surrounded by people who are complete ballers working on amazing things")
2. Short and regular tenure of only 4 to 5 years, so no incentive to fuck around/do projects to get promoted, etc.

This probably can't be applied to most corporate context without applying all the other things that make DARPA work. But there are other ways to achieve these outcomes (removing incentives for easy wins; preventing criticism; removing friction to change) with other supportive structures.

Almost everything is outsourced

My "normal" stance on insourcing vs. outsourcing is something like, "most companies don't have enough of the critical creative capabilities they need to win, and should look to reconfigure around differentiators vs. requirements."

The opposite at DARPA:

As of April 2020, there are 124 staff and three layers. That number is right around Dunbar’s Number –just small enough for one person to know everyone in the organization.

Actual cutting edge research may require rare equipment or knowledge. There are many pieces of equipment or tacit knowledge that only exist in one or two places in the world and it’s easier to access them through finite projects than purchasing or hiring them.

[Outsourcing] enables strong accountability because for any program there is exactly one responsible person.

An organization of only highly skilled, highly technical managers who are expected to only outsource means you've got ~100 very interesting bets going on at once. Cool. Can you imagine this in a marketing department? Or as the overall corporate structure for, say, a consumer products company? I sure can!

Highly technical advisory sessions

My "normal" stance on the role of leaders and steercos is something like, "they should define the boundaries for a team or initiative, and then get out of the way."

The opposite at DARPA:

The way the tech council meeting was described to me is that it’s roughly like a university seminar in a room full of people who you cannot bullshit and who have enough technical experience to dig into anything about the program. It’s important that the meeting be egoless and clearly focused on making the program as good as possible because this sort of thing can easily go down a rabbit hole if people feel the need to show how smart they are or just destroy the presenter.

The tech council doesn’t have any power besides advising the director on the program’s technical soundness.

(The Director is the one with the power kill programs.)

So there are clear structural safety mechanisms surrounding these sessions. There's no career juice to squeeze. The people providing the advice to the director might someday rely on each other – and on the person who is presenting their program – to help fund each others' programs.