My Favorites
A book I wrote about decision making. Use at your own peril.
On the domestication of AIs and resulting moral implications.
Is the world getting better? That depends on our extinction risk.
Design for uncorrelated failure modes, to best support your friends.
All the rest of 'em
- Let's Talk Tasks (2 min)
The guide is...
2024-11-13
ai-safety - Outside Auditors Need More Access (4 min)
Companies that issue debt do not also rate the riskiness of their debt. Letting companies do this would result in the same outcome as when...
2024-11-8
ai-safety - Paper Review: Concrete Problems in AI Safety (4 min)
2024-11-1
ai-safety - Paper Review: GSM-Symbolic (5 min)
Paper here. Ok, I am actually going to write about this above paper. I think it's more interesting than GSM Symbolic, specifically for task...
2024-10-23
ai-safety - On Linear Metrics for Task Scoring (8 min)
To argue this, I considered the `sha256^n` family of tasks: For humans, performance on this task scales linearly with time given to them -...
2024-10-23
ai-safety - On Task Based Scaling Laws (4 min)
One useful thing to do is to predict your LLM's performance based on how much of a certain resource you give it. For example, this scaling...
2024-10-10
ai-safety - Thinking Out Loud: Task-based Scaling Laws (9 min)
Consider two different problems: On the surface, both of these problems seem similar: ten relatively simple steps to accomplish, applied in...
2024-10-9
ai-safetycreative-writing - Paper Review: Intriguing properties of neural networks (3 min)
Paper here. Overall, this paper was a great read. I love old deep learning papers, as there's so little to keep track of and think about b/c...
2024-9-16
ai-safetycreative-writing - Mechanistic Interpretability: Baby's First Month (14 min)
It's been about a month since I've begun my journey into mechanistic interpretability - I'm the baby mentioned in the title! This post is an...
2024-9-11
ai-safety - Vintage Models: An AI R&D Eval Proposal (5 min)
Model intelligence increases as a function of: When it comes to training data, the stock of human-created training data will likely increase...
2024-7-17
ai-safety - AI Lab builds a LangChain competitor part 3 (5 min)
After defining the types of LLM integration libraries and [clarifying the wishes of library users], we're ready to ask: **what would it look...
2024-7-16
product - AI Lab builds a LangChain competitor part 2 (8 min)
In part 1, we explored LLM integration libraries: level 0 is raw API requests, level 1 is an application-agnostic API wrapper, and level 2...
2024-7-16
product - AI Lab builds a LangChain competitor part 1 (5 min)
If you’re here, you’ve probably heard of LangChain. And if you’ve heard of LangChain, you probably read a roast of LangChain. LangChain is...
2024-7-15
product - Ethical AI Vegans (2 min)
There is much disagreement on: You can find well-reasoned, well-supported perspectives on both sides - from philosphers like Peter Singer...
2024-7-12
ai-safetycreative-writing - Today's Shit, Tomorrow's Treasure (2 min)
In November of 1802, Alexander von Humbolt first encountered bird shit in enormous quanties. As the joke goes, at second-glance, it turns...
2024-7-11
creative-writingworld-model - We Must Dream of Limitations (2 min)
Most every fantasy book brings a different magic system. Lord of Rings, your magic power is a function of your birth - wizards and elves and...
2024-7-10
creative-writing - Paper Review: Attention is All You Need (7 min)
Paper here. Had a lot of useful dialogue with LLMs themselves to help understand a lot of this. Here’s how this all works: This is the...
2024-7-10
paper-review - Paper Review: React: Synergizing Reasoning and Acting in Language Models (3 min)
Paper here. Reading papers following the method...
2024-7-9
paper-review - Domesticating AI (3 min)
Domestication is, excuse my french, super. An animal is more domesticatable if it's easier / more plesant to put in a pen and breed. More...
2024-7-6
ai-safetycreative-writing - Among the Clouds (2 min)
The best lofts for living are full ones - with enough length to go for a jog, and thickness so you can hang your clothes out without them...
2024-2-10
creative-writing - Enlightenment Maybe (5 min)
Here’s our simple model of the world. Steven Pinker argues that human reason leads to technological advances which has led to the world...
2023-11-15
world-modelai-safety - Health and Avoiding Modeling Error (15 min)
At every moment in time throughout history, the state of the art treatments that the medical profession collectively holds is touted as the...
2023-7-12
world-model - Morning Walk (2 min)
Up Stanyon street, past the cyclery that promises decades of service, and the old firehouse, well-worn wooden steps rise from the sidewalk....
2023-4-1
creative-writing - What Becomes the Narrative (2 min)
In my work on blockchain, we always heard two main criticisms: there was the lack of security that some protocols had, and the environment...
2023-2-4
world-model - The easiest path it takes, the easiest path it makes (2 min)
In the stream down on the golf course, there is a rock that splits the flow of the water in half. This rock is a new element of the stream -...
2023-2-1
world-model - Unintelligent Design (2 hr 36 min)
I started writing the below when I saw a squirrel run up a tree, and was at the time thinking about Mito's goal of 5% growth weekly. I ended...
2022-8-15
world-model - Becoming a squirrel (13 min)
My previous post describes where a lack of long-term thinking can still lead to long-term planning that leads to "evolutionary success." To...
2022-4-1
world-model - Squirrel Strategies (9 min)
Squirrels bury nuts for the winter, across a variety of hidey-holes. During the winter, they only dig up about 78% of these nuts. At first...
2022-3-28
world-model - Replicators v1 (3 min)
Replicators are objects that replicate, and they are worthy of study for their applicability to questions of evolution. According to...
2022-1-14
ai-safety - AI Safety and Memes, Initial Research Ideas (20 min)
Any entity that forms lineages of descendants created by some “reasonably pure function of the parent” with occasional random mutations....
2022-1-14
ai-safety - Perfectly priced derivatives (6 min)
Terra is a protocol that, at a high-level does the following: In practice, this leads to TerraUSD being a stable coin. In theory,...
2022-1-10
inventionblockchain - Goodbye, Charlie (4 min)
Charlie died last night. It was probably around 8PM. I was walking out from behind the kitchen bar, to do something or other, and he was...
2021-11-22
reflection - What I think science is part 2 (8 min)
Some things I note: Now, let us say we want to make propositions about the state of the universe. Let us say the fundamental question here...
2021-11-15
world-model - What I think science is part 1 (2 min)
Science is a process that is designed to drive human knowledge towards objective truth. It therefore assumes there is such thing as an...
2021-10-15
world-model - What can I do about climate change (5 min)
I've been reading this and [that] about climate change, trying to get a handle on what climate change really is and what, if anything, I can...
2021-9-15
reflectionworld-model - What I should work on (2 min)
I think that there's a worthy consideration of "what is worth it" in the long term from AI...
2021-8-16
reflectionai-safety - In Motion (5 min)
I spent two years trying mindfullness meditation, and I failed to stick to the habit for more than a week straight, ever. I've successfully...
2021-8-1
reflection - Anthropogenic Effects and Reproduction Success (2 min)
Or: careful what your sampling from, if you’re even sample at all. There is an chain of reproduction, stretching from a first-living...
2021-7-22
world-model - Clarity, Please (4 min)
Clarity was missing from most of the classes that I took in school. I think that this is one of the major problems that I had with my...
2021-4-5
reflectionworld-model - Everyone's Dog is Dead (3 min)
Well, not everyone. But a lot of my friends had dogs that died this year. Practically this manifested as everyone being sad about their dog...
2021-3-21
system-designcreative - On Heuristics (4 min)
A heuristic is any algorithm that solves a problem approximately. In some cases, heuristics are used because "correct" algorithms are too...
2020-8-10
world-model - Spreadsheets are the Ultimate Programming Paradigm (3 min)
Functional programming? Eh. Objects? Maybe sometimes. But spreadsheets... spreadsheets are the ultimate programming paradigm. Naturally,...
2020-6-3
productspreadsheets - Why Version Control is Not Version Control (3 min)
This post disagrees with a post by Abstract, a company one would expect Saga to be aligned with given our "async collaboration is better...
2020-5-19
productspreadsheets - Fuck It, We'll do it Live (5 min)
You're designing a new application, and you want it to be collaborative. Don't worry, you don't don't have to do much thinking. You just...
2020-5-18
productspreadsheets - On Generalized Platforms (5 min)
We launched saga with the intention of building "generalized version control." The thesis: git-style collaboration works really well for...
2020-4-8
blockchainproduct - Taste Reset 2 (2 min)
Yeah, ok, the Taste Reset is not gonna happen. There's just no way I could really solicit good input from my friends with music, on an open...
2020-4-2
reflection - Product vs. Protocol (3 min)
I'm not very good at building products. I don't have a great product sense, I kinda suck at user interviews, and I struggle to understand...
2020-4-2
blockchainproduct - Taste Reset (2 min)
Here's a terrible idea: I'd like to reset my preferences. Less dramatically - more specifically- I'm going to create a new Spotify account....
2020-3-30
reflection - Quarantine 2 (2 min)
Quarantine update! I'm excited to write this post for no reason, which just goes to show you how much quarantine can addle your brain. I'm...
2020-3-29
reflection - Options (6 min)
I know very little about derivatives. I started learning about them yesterday, so take everything I say with a grain of salt the size of one...
2020-3-24
world-model - Induced Demand (4 min)
City streets and computer hardware and blockchain throughput. We dream that "make bigger, make faster" will alleviate congestion in all of...
2020-3-20
blockchainworld-model - Self-Hosting (2 min)
This blog is self-hosted now. I stole the theme from Vitalik. It's simple but pretty and also exactly what I'm looking for. I might put it...
2020-3-12
blog - Quarantine 1 (2 min)
It seems like we're going to enter into harsh "social distancing" mode as a US-of-A pretty soon. Which is pretty scary, but mostly really...
2020-3-12
reflection - Against Index Variables (2 min)
I don’t think that the economists of the future will use index variables in the same way that economists use them today. Simply put, and...
2019-11-1
world-model - Review: NileRed (2 min)
This is the start of a new series on this blog. The goal: review things I find on the internet that I find particularly interesting or cool...
2019-7-17
media - Sorry, Mr. Wallace (2 min)
David Foster Wallace’s This is Water is a senior-year-of-highschool/college classic. Wallace introduces the new graduates to some pervasive...
2019-7-8
media - Data-Driven Happiness (2 min)
What makes me happy: a sunny day in a green park; a good run that makes you swear; a funny friend and a silly joke; a productive day and...
2019-7-8
reflection - Depression from Existential Dread (2 min)
There’s been a recent trend on Twitter and [Reddit] to attribute mental illness trends to “how fucked” the world is. The claim: depression...
2019-6-10
world-model - Trust the Process (2 min)
The New York Times recently created a series called Diary of a Song. Each video takes some famous song and overviews its creation: initial...
2019-6-10
media - Eighth-Grade Optimism (2 min)
Eighth Grade is a painful movie. I fast-forwarded bits of conversation to avoid the second-hand embarrassment. Throughout the movie, I...
2019-3-11
media - Stacking as Human (2 min)
All the greatest things are just stacks of other less great things. Sandwiches? Just a big stack of bread, meat, cheese, pickles....
2019-2-15
invention - Light-speed, Nerf, and Abstraction (2 min)
I used to love Nerf guns. Sometime in middle school, I started trying to modify the few Nerf guns I had to shoot faster and farther. I’m...
2019-2-15
inventions - Robots will Take Our Jobs, But Not Ours (3 min)
Just based on the amount of new automation in manufacturing and supply chains and various other simple-services, it’s easy to say “wow,...
2019-2-7
world-modelai-safety - More Radical than You Think (2 min)
People making predictions about technology is always a terrible idea. But even articles that acknowledge that trying to make these...
2019-2-7
world-model - Management As an Engineering Discipline (2 min)
Bridges conform to rigorous standards, else they fall. And so usually they confirm to the standards. Teams don’t really conform to any...
2019-2-7
world-model - Throwing Better (2 min)
I’ve thrown a single bowl in my life. It was hard as hell, terrific fun, and looked terrible at the end. Is there some improvement we can...
2019-1-21
inventions - VM Versioning with Persistent Smart Contracts (2 min)
The most recent Ethereum Hardfork didn’t happen. Shortly before the hard fork block, a bug was found; pretty much, old code relied on...
2019-1-17
blockchain - Sick, Yo (2 min)
When I was in the 11th grade, at the height of my soccer career, I got hit in the head with a soccer ball. And got a very mild concussion....
2019-1-17
reflection - Quantum Consensus Algorithms (2 min)
DISCLAIMER: I have no idea what I’m talking about. Although I know as little as one could know about it, quantum computing seems very...
2019-1-17
blockchain - Moar Media (2 min)
The new God of War game looks awesome. I haven’t played it, but I watched pretty much the entire game on YouTube. It’s heavily story driven...
2019-1-17
media - Community as a Protagonist (2 min)
Main characters dying in books or movies or video games is usually shocking. Some authors kill characters more frequently than most - and...
2019-1-17
media - Types of Understanding (2 min)
I’m a total type theory noob. When I say I “like Type Theory,” I’m mostly just saying I like type systems that stop me from creating stupid...
2018-12-31
reflection - Recursion in Evolution (2 min)
Disclaimer: my biology knowledge ends at 9th grade bio - so everything in this blog post is pretty much fake. Why are so many things in...
2018-12-31
evolutionworld-model - Formal Methods for Scam Detection (2 min)
I read a paper recently where the papers authors analyzed how often Ethereum ICOs delivered on the what they promised in their white paper....
2018-12-31
blockchainsecurity - Correct by Construction Correct by Construction (2 min)
The Correct by Construction design process is a very interesting. The general idea can be summarized as: The benefits of the CBC process...
2018-12-31
math - Client-side Code and User Guarantees (2 min)
At the time of writing Etherscan.io is currently the most popular Ethereum block explorer. If you check out a smart contract on Etherscan,...
2018-12-31
blockchainsecurity - Corporations as People (3 min)
How much of a company’s advantage in a market comes from the private information it has? How much value does Urban Outfitters get from not...
2018-12-24
world-model - Learning From the Old (2 min)
My dad was big into computer graphics in the late 80s and early 90s. He worked at an IBM research lab - drawing cubes and spheres and...
2018-12-23
reflection - Accountability not Transparency (2 min)
Blockchain is often touted as a corruption-stopper. The idea: government payments can be put on the blockchain, and because these...
2018-12-23
blockchain - Blockchain Governance as an Information System (2 min)
I just finished reading the essay “Common-Knowledge Attacks on Democracy.“ To summarize the essay’s interesting points: democratic states...
2018-12-21
blockchain - A New Idea (2 min)
Welcome to my new new blog. I’m creating this for the standard reasons: a selfish wish to share my personal thoughts, and deep existential...
2018-12-20
blog