Korva is a scroll of the internet’s best ideas.

Why people are needed

A concentration of people in San Francisco really know how to deploy agents, eval agents, and be AI-first in engineering and other areas. That knowledge isn't out there yet, and the talent isn't available for every enterprise to have its own expertise at this point in time.

The long-term view

Eventually it will be, and teams to set up AI agents for every enterprise may stop being needed, with the work becoming a job function similar to software engineering. That is maybe a decade-long change.

One view holds that services are just an excuse for a weak product. The counterpoint is that it is a knowledge dissemination problem, and services bridge it while the knowledge spreads.

Who does the work

Forward-deployed engineers suit a particular profile: great communicators who cut through to the source of a problem and simplify. Engineers like that also make great founders, and Mercor has many alumni who went off to start companies.

Where did forward-deployed engineers come from?

Palantir, founded in 2003, popularized the role, placing engineers inside customer organizations to adapt its software to their data. The company called them forward deployed engineers. The title borrows from the military use of forward deployment, where forces are positioned near the area of operation. Other enterprise software companies have since adopted similar embedded roles.

Why do new technologies need installers at first?

Economist Paul David's 1990 paper The Dynamo and the Computer pointed out that factories saw little productivity gain from electric motors for decades after the first power stations opened in the 1880s. Gains came once factories were redesigned around the motor, which required people who understood both the technology and the factory. Enterprise software went through a similar phase, when consultants installed and customized large systems for each client.

How did software engineering become a common job?

The term software engineering was popularized at a 1968 NATO conference in Garmisch, Germany, convened over concern that large software projects ran late and over budget. Over the following decades, universities added degree programs and companies adopted shared practices such as version control and code review.

More from 20VC with Harry Stebbings

Tech & AIToken spend depends on the use caseCoding-agent token spend can compound into gains, while a customer service agent whose tokens cost more than its revenue is in a bad position.
Enjoyed it?