Normal People

The world is waking up.

Part 1

Introduction

The metamorphosis of prime intellect

What was soon going to happen was kept among several hundred people, most of them in San Francisco.

Two years after, the world has realised what's going on, roughly, but it's still within a small subset of people. Not hundreds, but tens of thousands of engineers and people who work closely with the frontiers and help enterprises transform their businesses.

As much as I'd like to quickly write down some simple ideas, I want to continue on this essay with facts that I can verify and test agasint. The outline below moves from energy and land, to model training, compute economics, government capacity, robotics, hardware factories and post-AGI society. I want to answer a few questions:

  1. Landscape tracker pipeline to map energy, land use, and critical material production against what is claimed in the news - most importantly because of urban planning has its own guidelines aim to protect civilisation and land usage, is it ok to build that much power plants and whether they are abusing this planet? More so how much energy is enough?
  2. Observability tools for LLM training and serving, CPU, HBM, and data-processing bottlenecks directly
  3. Semiconductor simulation and discoveries around H100, Blackwell, Cerebras, and relavant systems to produce intelligence
  • 2 and 3 are adjacent to the questions I want to know. We all know there are a lot of optimisation can be done during computation, but I am afraid we are reaching the limit of it - What else can we done to optimise for computation? Are there new technology can help us beyond photolithography and current SOTA of packaging? How much energy and planet resources can we save from doing so? Would we increase production of raw source of computation power and able to build self-replicatin factories by doing so?

4.RL and eval environments to test the real cost of self-learning loops in production

  • Rich Sutton mentioned that stochastically people will be benefited from general use models, yet there are still numbers of companies being found based on RL and custom models (Prime Intellect, Hud, to name a few). What will be the biggest impact from an RL environment that's widely adopted? How'd humanity be more efficiently benefit from it? If people are not widely adopting the mechanism, what type of human and machine interaction would be driving it forward?
  1. The path to automated AI research, automated AI alignment

... and interviewing a lot of founders, engineers, operators across US, UK and EU.

Two years after

It's been two years since the infamous situational awareness. I completely agree with what Leopold said, "You can see the future first in San Francisco,". I wanted to find out where humanity is going in the next 20 years and I can't think of any other places to look for it other than San Francisco. Important enough that I paused whatever idea I was working on to stop the humming noise in my ear and flew in to San Francisco from London.

The first ultraintelligent machine is arriving.

  • 2024: we would complain that it (coding agents) couldn't handle low-level networking tasks.
  • 2026: it can competently do the majority of the infrastructure work as long as the specs are clear, and handle small parts of eigendecomposition, convex optimisation, change detection and more in pure math.

If you are in San Francisco you can probably feel it, even engineers who used to work on low-level engineering tasks no longer need to write code.

We forgot the order of magnitudes (here and after OOM). We only saw what's now but forgot to relate now to the past and the speed it has been, the order of magnitude. What was soon going to happen was kept among several hundred people, most of them in San Francisco. Two years after, the world has realised what's going on, roughly, but it's still within a small subset of people. Not hundreds, but tens of thousands of engineers and people who work closely with the frontiers and help enterprises transform their businesses. The large majority is enjoying, or suffering from (expand more in later section), the "AI boom", but this is not the end of the story.

"I don't know how to get people AI pilled anywhere outside of SF."

"The substrate of what's coming is scary."

People from SF are struggling to communicate with people outside of tech industry, but actuall I found interesting angles to communicate without trying to get people AI pilled. Even people from labs are terrified of what's going to happen. Looking at every industrial revolutions, the time of transition always seem scary and when it happened every time, everything comes after was always very fast. We will talk more about the history and how we may peaked into history and Sci-Fi to unpack the global future landscape.

I witnessed a lot of my friends treating every new generation of large release like a pet - "Why is it not responding to me?", "How can I make it work smarter?", "They cancelled the release? I am down today. I don't want to code." An invisible emotional bond with a junior employee who you have trained for the past couple of years (how I see the world post-agi). Once a good model gets retracted, they comment, "There was nothing like it. I don't want to use anything else now." The [prediction] was true. Automated AI research is happening. Self-replicating factories, data centers, robots are still in research phase, but it's going to happen.

If the doomsday machine is meant to be made, we will need every power we have on this planet to survive our own curiosity. Humanity has this unwavering desire to explore and understand the unknown and we withstand all odds to do so. We ought to keep driving this machine forward to know the answer to"The world and economy will be much more efficient based off it". Witnessing the fear, I'd much rather we don't get driven by fear, but by the child-like wonder to explore. If it doesn't work out, perhaps Elon's SpaceX can help us sustain our life elsewhere.

I will reference a lot from Leopold's situational awareness (appreciation and gratitude to him who wrote it), combined with what I have heard and read, to fact-check what has changed. This is an article I read again and again, wish I have read, and anyone of my friends in London have read in 2024. Witnessed China's hypergrowth first hand, worked in San Francisco, I want to offer United Kingdom a way forward.

Let's see where we are at after two years.

Lois, June 2026

Outline

About me

I am Lois. I am a bundler engineer primarily work on infrastructure projects, orchestration and deployment. There are not many of us on this planet (probably 300-400 of us?). Previously I worked in San Francisco, was founding engineer in Toma (backed by YC and a16z) on voice AI orchestration infra, runtime over the air update for mobile, tooling support for open source projects like Bun, Rspack (by ByteDance web infra) and the Gotenberg Project. Funny enough I was also Knight Frank's No.6 employee for their Capital Market Team (commercial transaction Toshiba, DJI, Huawei and more) in South China in my past life. Bootstrapped a bunch of business from China, Macau and the UK including smart city consultancy (800k+ ARR before I left China), internet routers and more.

Grew up in China, studied hospitality, started my career as a waitress and housekeeper in Sheraton, I have came a long way.

I live between London and San Francisco.

I like it hard and deep.

Talks (Globally)

  • "How to be extremely lucky", TEDx Youth, BrookGreen, London, United Kingdom, 2022
  • "Database: MySQL and Postgres' indexing engine, query planner, metadata", McKinsey London HQ, United Kingdom, 2023
  • The superhuman at Supabase - Conversation with Supabase CEO, Paul Copplestone, JustEat HQ, 2024
  • Unfuck your bundlers Computer Science Museum, Mountain View, California, United States - Monorepo World 2024
  • "Fixing your bundlers" speech in City JS Athens, Greece 2024
  • "Building runtime over-the-air update for React Native" speech in RenderATL, Atlanta, USA, 2025
  • "Detangle the mess of NHS" talking about document processing pipeline in Zoopla HQ, London 2026
  • Finalist speaker - React Summit, NYC, USA, 2025
  • diagnostic_channel deserves better on infrastructure monitoring and observability - NodeEU, Bologna, Italy, 2026