The 2023 SensibleDefaults Manifesto

Introduction

So the last month or two have been transformative for me - I started toying with GPT-3 because the results were pretty good, then GPT-3.5 (ChatGPT) came out, and changed everything. Results are still so-so, but the bar for just trying things got incredibly low for everyone in software.

The biggest problem with GPT-3 is that it'll bullshit you - and preventing that is a hard job sometimes. We've all seen this if we've tried it for code: It still beats me at pandas, but when it is wrong, it is wrong. There's problems and tricks to working with it, but it's not too tough to pick up when you start using it.

It took me about two months to get started. In that time countless LLM based startups have sprung up - Doing things we'd typically contract out at work. It's coming for us next. My big fear is it takes everyone this long to get started - Not everyone can afford that time, and we just can't afford to let people (especially large companies) build things in private that run us all out of the market.

So I'm Trying Something

I'm putting together a small team. Initially I'll call them CollectiveDefaults, and they'll be writing both here and on our public notes/resources site to come. My brother Mike did the first post of the series Here if you want to see the kind of thing we're aiming for.

The goal is to assemble community notes and guides aimed at everyone - even (way!) outside of tech. My goal is to help lower the bar for the planet and get people started thinking with large language models in their everyday lives.

Getting perspectives from outside of normal tech is a big part of this - we don't know what we don't know, and modern LLMs (and the newness of them!) make the bar for contribution incredibly low.

Why Now?

We went from bagphones to iphones overnight in the LLM world - and now there's public pressure for the big players to release their models. If you're looking to get into tech, or you're already here you have to learn a new way of thinking and using tech tools fast.

My team is doing that in public and noting what they found helpful - It took me too long to really 'get' why LLMs matter, and I want to shorten that pipeline for the next round of learners.

My goal is to help you surf the automation wave and focus on your most important work.

We're going to see more monumental shifts in tech in the next year, I'm hoping we can help workers make it happen.

John Partee

SensibleDefaults.io