2x Engineer = ?

Latest data shows that there is now twice the number of merged PRs for companies that actively use AI than those that don’t. How do you stop that from being twice the amount of crappy code?

6/19/20261 min read

Right now, most companies are still in the first stage of agentic maturity, where developers are supervising agents from within the loop. The same industry data shows that the median number of agents that developers are managing is 4. This is the maximum number of ways that the developer can split their attention. This constraint comes about because the engineer is still embedded in the loop.

A consequence of the developer being embedded in the loop is that the quality of the output of the loop reflects the quality of the engineer. An average engineer will give average quality prompts to the agent, will spot an average amount of weaknesses, may not spot the hallucinations, so will produce average code. A great engineer will give better prompts, spot all the weaknesses, so craft great code.

Agents help all developers write more code. It’s unlikely that they’ll help developers write better code. So you get more of the same. If you’re stuck in this loop, how do you break out?

I recommend a couple of patterns. The first to try is to create agents that will prioritise your learning. There are plenty of posts out there that have great ideas; here are few to get you started. In planning mode ask agents to search the web before planning out their tasks. They should give options of coding patterns, technology libraries, and explain the strengths and weaknesses for you to choose. They should give provenance e.g. library versions so that you can check that the brand new code doesn’t already need a refactor. This learning agent will cost more in token burn, but your next prompts will be better.

The second pattern is much more powerful: shift your perspective and put the human above the loop. Stop coordinating every implementation step and instead govern the system that performs the work. The human’s role shifts from operator to steward: clarifying intent, shaping constraints, reviewing outcomes, making great decisions and retaining accountability for the consequences. Elevating the constraint is the way to go from 2x Engineer to 10x Engineer

Avi Sinharay

CEng, MIET, MEng, MA (Cantab.)

Fractional CTO, Director/VP of Technology

Core Expertise

Technology Leadership, AI Native Dev, Operating Model Design, Engineering Culture

Domains

Health Tech, Media Tech