The Limiting Factor in Modern Coding

By Admin20 May 2026

By Raju Vegesna, Chief Evangelist, Zoho 

When you drive on a one-lane road, the speed of traffic is set by the vehicle in front. That first vehicle's speed becomes the 'limiting factor'. To let more traffic through, you add more lanes to remove the limiting factor.

Something similar has been true for writing code. Code takes time to write. That time was the limiting factor. To go faster, we put teams together so multiple people could write code in parallel. It worked, up to a point. After a certain team size, the cost of collaboration starts to outweigh the gains from parallel work. That has been a known trade-off for a long time (the mythical man-month is a well studied phenomenon).

With AI, writing code is no longer the limiting factor. Collaboration is. It is taking more time to align with another person than it takes to write the code itself. The gap between the cup and the lip is shrinking. The limiting factor is moving from code to collaboration, decisions, QA, review, security, etc. This has implications for software development lifecycle.

One direction is to break software into smaller atomic parts that each person (or a very small team) can build independently, with a system in place to fit them together. From what I observe about Anthropic, they are organized into very small teams (sometimes a team of one person) working very closely reducing this limiting factor in collaboration. I do not have a clean answer for the shape of teams in this new environment. My guess is we will figure it out by experimenting. I am curious if anyone else is seeing this play out on your teams? If so, please share your observations.

One thing I find interesting: At the chip level, the direction is the reverse. CPUs and GPUs used to be independent atomic units connected by a bus. AI required them to 'collaborate' better, which led to shared memory and higher bandwidth communication between these historically atomic units.

Increasing collaboration at the chip layer may be leading to more atomization at the human layer, the opposite of what we have seen in the past.