So much AI angst comes from the difference between thinking about an LLM as a compressed representation of the language in the training data vs thinking about ChatGPT as a product experience.

AI models will be like Cloud providers in that redundancy and diversity bring value: if you want to validate you need independence. Guaranteeing or maximising orthogonality will be a hard problem.

There is one tool that is essential for coding that is rarely listed in a dev’s toolbox: pen and paper.

One of the smartest people I ever worked with had a pad of graph paper on his desk, with no cover. By the end of each day it would be entirely covered in ideas workings doodles. One page a day.

I love it when my bank app notifies me of our cloud bill and asks “would you like to add a receipt?” Always makes me chuckle.

I’ve never known a subject that divided experts as much as AI. Half the folks I know think AGI is happening next week, the other half think we are nowhere near. I can’t recall any similar polarisation amongst people genuinely expert on a subject.

I wonder if the major non-random factor in startup success is founder resilience.

Parts of my software stack. Wish I had not had to test so much to get here:

CRM: HubSpot. I’ve tested everything cheaper than Salesforce.

Collab: Slack.

Calls: Zoom.

Knowledge base: Whimsical. It’s great. I have tried everything else. Everything.

IDE: Jetbrains, always.

TBC. Add yours ⤵️