Structured Communication is the New Code

this was a talk given by Sean Grove from Openai at the AI Engineers World summit earlier this year.

it won't come as a surprise to anyone that the nature of programming is going to change.

but I think this is one of the most simple and succinct framings of that change.

in the talk, Sean encourages encourages a reframing of the idea that 'code' is the most important artifact.

"code is 20%. The rest is in structured communication - talk to users to understand, distill, ideate, plan ways to achieve those goals."

structured communication.

what does this mean?

computers are getting very good at writing code.

there is obviously debate surrounding how much of the current human developer pipeline can and will be replaced by ai. but the 'if it will replace' will likely become less interesting than the 'how will it change'.

in the talk, Sean said this:

"in the near future, the person who communicates most effectively is the most valuable programmer .. literally, if you can communicate effectively, you can program."

this explains why 'vibe coding' is so interesting.

"vibe coding is communication first, code second".

in a future of increasingly abundant intelligence, computers are going to get better at writing code. but the problem with computers writing code and 'vibe coding' in general, is that it produces 'sloppy' and potentially harmful code.

so the real lever becomes "structured communication", or - "context engineering".

ie,

1 how well can you communicate to the computer what you want, and 2 how efficiently can you validate the outputs to reduce slop

i would argue that (1) will become increasingly important, and (2), less-so. but we'll see.

dex (founder of Human Layer - who build 'coworker agents')

in the old world - developers spend most time writing and reading code; reviews are line-by-line and the codebase is the single source of truth

new paradigm - create compact, structured context (plans, specs, summaries) that agents and people use. Reviews focus on understanding intent and alignment.

and this isn't just theoretical. in the talk, he gives several examples implementing this new approach to become far more efficient while maintaining quality (anti-slop).

this is basically Sean's 80/20 structured communication versus raw-code future of programming in action.

i think this will extend beyond software development.

take a writer, for example.

if 80% of your work, the artifact, is the published writing, you're ngmi. it's the thinking, the research, the conversations, the connections and the experiences behind the writing.

just as computers are getting better at generating code, they'll get better at generating all types of .. everything - so, learning to focus more on the process behind the artifact, and package this in an 'ai-interpreable' way seems to be the path forward.