If you read only one thing about AI/LLMs this week, make it this:

AI is Writing Code Now. For Companies, That is Good and Bad

June 9, 2023

If there's only one thing you read about #AI this week, make it this: AI is Writing Code Now. For Companies, That is Good and Bad by Isabelle Bousquette at the WSJ.

TL;DR:

- Think of LLMs as a credit card for code - buy now, pay later! Teams can write and deliver more code with AI assistance than ever before.
- But accelerating delivery also accelerates the accumulation of technical debt, and makes governance exponentially harder.
- CIOs in particular will struggle to maintain quality and security as they absorb vastly higher volumes of output.

Dan Mason's hot take:

- Don’t confuse abundance with productivity. This is a classic outcomes over outputs situation — why build 10 things which don’t work when you could build 1 that does? LLMs make it vastly easier to create new things, but you need to improve how you choose what to build, how it integrates into your products and systems, and how it contributes to the long term health of your business.

- Dev teams need to better understand how to use LLMs to help increase capacity across the board, and not just for their “coders”. This will require rethinking roles and responsibilities, the path to production, and the process for separating what is worth deploying from what will just cause headaches down the road. The technical challenge will pale in comparison to the transformation in people and process. Buckle up!

Link to original post by Dan Mason on LinkedIn