Discussion about this post

User's avatar
Peter W.'s avatar

While I agree that your approach would indeed make chip design a lot more approachable (digestible?) for AI, it also faces the same obstacle that is IMHO a key reason why documentation is currently mediocre at best:

people don't like to obsolete themselves. Poor documentation is like a virtual moat that helps keep the barbarians out.

It would be interesting to know how, for example, Nvidia has dealt with this. I picked Nvidia as they started using AI tools to assist their chip design earlier than most.

Expand full comment
CitizenSleeper's avatar

>"The real challenge is not technical, but cultural."

This is a very important observation. Many a times, engineers who do not document are the not so good ones or mediocre engineers at best.

Added to this managers and management who don't care about documentation are the biggest liabilities for the company. Forget about AI for a second, it becomes very difficult to onboard a new hire when the documentation in the org is bad as it takes a long time for someone to get up to speed and be productive. Due to this the company is forced to keep these mediocre engineers thereby spoiling the culture of excellence. I have seen plenty of lifers doing the same part of the chip design for the entirety of their career.

Some of the best engineers I know wrote meticulous notes for themselves. It was a pleasure to read through it and it felt like a textbook. These type of engineers were also ideal troubleshooters in other domain as they were systematic and they could always come back to what they were doing as they had logged their work.

Some of the inspirational engineers for me are Bob Pease and Jim Williams who wrote some of the best documentation that I keep referring back to. Also TI and ADI has some legendary documentation which is like a textbook for people like me.

Expand full comment
1 more comment...

No posts