I was wrong about AI assisted coding...

(well, maybe not entirely)

I thought integrating AI into a dev team workflow was all about building context.

You know the drill - clean docs ADRs that someone actually maintains, PRDs that don’t read like corporate jargon.

Feed the agents good context, get good code out.

And boy... I was wrong. Well – not entirely. Context matters.

Without solid docs, your coding agents will produce mediocre output. That part is true.

But there’s something more important. More elusive.

Strong engineering culture.

I’ve worked with teams with beautiful Notion wikis and zero AI adoption. And scrappy teams with messy repos who shipped 3x.

They were faster because they had something else:

• Team members openly sharing tips and workflows: “hey, I got Cursor to do X with this prompt” lands in Slack and everyone leveled up by lunch

• Ownership over the process, not just the task – engineers who ask why before how, who push back on wrong directions instead of just closing tickets

• Challenging each other’s decisions; in code review, in architecture calls, in “wait, should we even build this?” conversations

  And here’s the beauty about that kind of culture; it doesn’t matter what wave of new tech we’re riding. Cloud and microservices yesterday, AI today, the next sexy thing tomorrow.

Teams that share knowledge, own outcomes, and challenge assumptions don’t get disrupted by new tools. They absorb them.

Context makes your agents smarter.

Culture makes your team unstoppable.