Code Was a Big Part of the Bottleneck

Writing code may not have always been the bottleneck, but it's horseshit to say that coding agents when properly directed don't meaningfully change the overall calculus in building software.

I think when most people think of "code," they think of the full process to create something new that ends with software being released. It does have many more steps that are more important.

But let's be realistic: the biggest chunk of calendar time in that process used to be implementation, and I can tell you from direct experience that efforts that would have taken months end to end from a full team can now release safely in weeks, maybe even days if the conditions + foundation are right. With much fewer people involved per effort. And yes, on systems with paying customers, not just completely speculative 0 to 1 things.

Complicated migrations from one database to another. Building completely new end to end testing suites to enable upgrading years old frameworks by multiple versions. Those now took weeks instead of the months it would have taken before even with a full team - all done by myself or with some help from another part time senior dev.

I recently shipped a decent sized customer feature request from 0 to production in one day. With 11 other improvements from the previous couple of days.

You can make bigger bets and find out whether they're going to pay off sooner. Cost of delay can go down on everything.

Can you make a mess faster? Hell yes if you're not careful.

Can you get feedback on something real from the market faster if you do it right? Also hell yes.

Build your agent management skills or find somebody who complements you who has them.

Bad foundation? An agent army with the right leader can help fix it. It will make a difference. You'll have more at bats to figure out the right things to build with the same budget.

What are you doing now that didn't make sense before?

Join the conversation on LinkedIn