Comprehensive Thoughts on AI

This document aims to: Provide a high-level understanding of Artificial Intelligence/Machine Learning and how it works. Go deeper on LLMs (Large Language Models) and why they’re important. Help you create a prediction framework for the intelligence level of AI over time. Highlight the non-linear impact of AI on business, government and politics. The second part of the document is more technical, covering a broad range of engineering topics: Explore what AI scaling will look like over time. The technical soup to nuts: from hardware to model. The key takeaways of the document:

Read more →

August 5, 2024

Machine Learning Notes for Developers

Machine Learning systems notes. Some ‘ML’ math at the start, hardware stuff in the middle, training/inference and optimizations at the back. LLMs TBD. Linear Models Simplest linear model: $$ y = mx + b $$ $$ y = 2x + 3 $$ Where ‘x’ is the input, and ’m’ is the parameter in the context of machine learning. A multivariable linear function generates a plane. In this instance, a plane in three dimensional space:

Read more →

March 1, 2024

Managing Managers Notes

[This is a bunch of bullet points on managing managers and teams, and that ended up being the foundation of a large set of ‘mental models’ I wrote. More on how to do that later…] Preface: Congratulations! Our desire for organizational stability is our worst instinct. Seeing an org as a fixed cost is precisely the wrong mindset, you should be creative with your organization and engage with it like a ‘work product’. Right now there is someone on your team who is doing great. They could do more if you let them, but they aren’t because you’re happy with the work they’re doing. That’s the wrong mindset for growth. Are you pushing themselves to the point where they’re starting to fail? If not, they won’t understand their capacity, and you won’t realize the extra ‘yield’ you’ve already got. As a rule, you should double the capacity of your leadership every year. That’ll stay ahead of growth (linear) and cross-functional complexity (quadratic). Let people impress you. Stretch them. When things start to fail, that’s signal on where to hire. “Congratulations on your transition. We call it a transition for PR reasons, but it’s actually a brand new job, and your new API is people”.

Read more →

January 1, 2024

Hiring Mental Model

This is my mental model for hiring people into an established organization. TLDR: it’s about the courtship, how well the individuals career story fits with the opportunity, and overcoming uncertainty] Philosophy and Principles The goal is multivariate optimization: the multi-year experience of the person you’re trying to hire; the teams future experience with the addition of this person; and the company. Each has different goals and agendas, but all must be net-positive to proceed. To hire or not is a decision weighed by the inputs above. Hiring doesn’t stop at the signature. It’s your job to ensure the smoothest possible transition for the individual into the collective team. This is a many month process. The act of hiring is about courtship. Invest about the same amount of energy into it, as you did when you courted a partner. Assessment of Requirement Before hiring anyone, you’ll need a view on:

Read more →

April 27, 2023

Communication Techniques for Mutual Understanding

I was watching “The half of it” on Netflix over the weekend (a lighthearted comedy-drama written by an ex-Microsoft software engineer-turned-writer, Alice Wu) and a scene in the movie just happened to succinctly summarize a common problem I see almost daily at work. The smart loner Ellie, is trying to coach popular-jock Paul, on the art of conversation so that Paul can charm his high school love interest: In a subsequent ping-pong scene, Ellie says to Paul “match energy, match strokes, and just say one thing”, which is an apt sporting metaphor for the basic skill of communicating with another.

Read more →

November 27, 2022

Write down your how

Regularly writing down both what I’m thinking and how I think about things has been the highest impact career tool I’ve learned to date (it’s up there with learning Vim). It helps me get ‘unstuck’ when I’m uncertain, understand my reactions to events, and teach and grow my colleagues. The what and how parts are distinctive, with how being the most important. I’ll explain this distinction using a common example of where I, and the folks I work with, get stuck: making a decision about something important. I’ll then explain why writing down how you think, and then deconstructing and improving that, is the key to rapid personal development.

Read more →

October 27, 2022