I was personally on call for some gpt-oss training runs, so you know those models were looked after well.

Codex took maybe fifteen minutes to write it, run the tests, and fix the bugs. My day off beat a pre-AI workday.

I've written on why you should never trust a senior engineer who doesn't code, and on the agile waterfall.

Drawing on TPUs at Google and GPUs at Nvidia, I wrote it up as a free document that grew from 80 to 151 pages.

I worked on Facebook's internal cloud, Confluent's data in motion, and Amazon's People Engine for 2M colleagues.

I also worked on TPUs there. Seven years after leaving Google, I still typed 'blaze build'.

After too many years consulting at a large bank, I wrote a satire on designing unmanageable, insecure systems.

I was writing BASIC and assembler then. In 1990 I started a Comal interpreter in C, now open source.

I was personally on call for some gpt-oss training runs, so you know those models were looked after well.

Codex took maybe fifteen minutes to write it, run the tests, and fix the bugs. My day off beat a pre-AI workday.

I've written on why you should never trust a senior engineer who doesn't code, and on the agile waterfall.

Drawing on TPUs at Google and GPUs at Nvidia, I wrote it up as a free document that grew from 80 to 151 pages.

I worked on Facebook's internal cloud, Confluent's data in motion, and Amazon's People Engine for 2M colleagues.

I also worked on TPUs there. Seven years after leaving Google, I still typed 'blaze build'.

After too many years consulting at a large bank, I wrote a satire on designing unmanageable, insecure systems.

I was writing BASIC and assembler then. In 1990 I started a Comal interpreter in C, now open source.