Eight years after World66 went dark, we are bringing it back as World66.ai : a free travel guide where AI is used to maintain the commons, not strip-mine it. The code is open source , the text is free to copy, there are no ads or paid links, and anyone can contribute by putting an agent to work and opening a pull request. The bet is simple: AI does not get to bury the web under infinite travel…
Goodhart's law is the benchmarker's curse: when a measure becomes a target, it stops being a good measure. Coding-agent benchmarks are almost designed to trigger it. The tasks are public, the result is one number and the leaderboard inevitably fills up with harnesses that are, often without meaning to be, overfit to the benchmark. That does not make the go-to standard Terminal-bench useless, but…
There's a World Cup every four years, and every four years I build a model to predict the winner with whatever tool I happen to be using at the time. Here's this years . Eight years ago it was Python. Machine learning was riding high, I was writing a book about it, and a small regression model said Brazil was going to win. The model was simple enough to explain: learn team strengths from past…
I wrote torreable , a tool that lets AI agents publish websites to the dark web. The first reaction is usually that this sounds like the start of a dystopian short story: robots quietly spinning up onion sites by the thousand, the worst of both worlds in one Python package. It's a reasonable concern. But Tor's reputation has very little to do with what Tor was actually built for. The protocol is…
More than 20 years ago I was working on a travel startup, world66 , and in the process I got a list of all countries and their coordinates. Geo on the internet was young, so this was non-trivial. So I wondered what else you could do with this data. Back then you could scrape Google News if your Python script claimed to be Internet Explorer 5.5, so that was easy enough to code up. I didn't think…
Around 2018 I took an UberPool across Manhattan for $1.78. Cheaper than the subway. Today the same ride is about $20 and nobody finds that surprising. The rides were never $1.78. The difference came out of somebody else's wallet, and eventually that somebody wanted it back. Tokens. While supplies last. The same thing is coming for tokens. Last month ccusage estimated my Claude usage at $5,098.85…
Democracy might be the worst form of government, except for all those other forms that have been tried from time to time, but within that worst form of government there are a number of choices. The Dutch system is purely proportional, so growing up I naturally assumed that this was the right way; if a party has 3% of the vote, they should have 3% of the seats. But there's something about district…
The BBC never published a digital edition of the Battle of Waterloo, but in the latent space of a Large Language Model (LLM), that page does exist. LLMs are famously trained to predict the next word in a sentence. Somehow, as a side effect, they also learn facts, concepts, and relationships. Among those relationships is the link between a URL and its content. This results in something like a…
When I was a kid I wanted to become an astronaut. But then I found out you have to be a pilot first and that seemed boring. So I decided to become an astronomer. But that involves working at night. But then I read in a science magazine aimed at children about how you could simulate planets coming into being using a computer. So I decided to do that. I fired up BASIC, implemented my understanding…
(this first appeared on Block's blog ) I hold the generally unpopular opinion that we write too many tests. Most tests written have negative value and should not only not have been written, we're better off actively deleting them. Most tests don't ever catch bugs. They fail when the underlying code changes as intended, not when it breaks. The result is a growing tax on improvement disguised as…
I learned something about myself and the world. I discussed it over a beer with a good friend of mine and he thought it was interesting enough for a write up-so here we go. You're an idiot Everybody is an idiot It's fine Ideally you realize these truths in sequence. When you're young, you are unaware of all of this. But then as you grow older, you start to notice how much you don't know and how…
There are many election models making the round, from the good old 538 (now under new management) to the economist and Nate Silver . The all agree it is basically a toss-up with Kamala having a slight advantage. There are 7 swing states that most likely will decide the election. What if we treat each state as an actual coin toss? That gives you 2⁷ = 128 possible outcomes. Turns out Harris wins in…
Another one of Paul Graham's essays is making the rounds: founder mode . Paul's writing is generally thoughtful and to the point, but I am not sure I am getting it this time around. The idea seems to be that as startups scale up, their founders are told to become managers and that makes them feel unhappy. That I get. Your nimble startup has become the slow moving corporation to quit. It's…
As Franklin would have it, nothing in this life is certain but death and taxes . And when it comes to US states it is generally thought there's a trade-off between the two. Live in a high tax, high quality of life state or cut down on your tax bill and your life expectancy. But is it true? I did some data digging to find out. Let's start with the answer because it makes for a pretty graph: I would…
I was looking around for a spreadsheet with all Y-Companies, but couldn't find one, or at least nothing recent. The top Google Result is deprecated and leads to a website that also seems to have only old data. So I wrote a little scraper that collected the data and put it in a Google Sheet. See the bottom of this post for technical details. Or if you are here for just the sheet, click here . The…
Every year during the gathering of the world leaders in Davos, Oxfam comes out with a report about how the rich are getting richer and the poor are getting poorer. It's good timing and press repeats the conclusion in capitals. This year the claim was: World's five richest men double their money as poorest get poorer . I decided to read it. The sources it is based on, the Forbes Billionaires lis t…
Two weeks ago Neptyne, my startup, announced that you can now access the OpenAI APIs like ChatGPT for free when building apps inside Neptyne. We're happy to now announce that we've extended this capability to four more APIs that would normally require registration and payment. How does this work? It's easy. You just call the API as you always would, but there is no need to provide a token in the…
While ChatGPT definitely has grabbed the headlines when it comes to the AI revolution, using LLMs to automate all kinds of tasks has yielded some interesting results too. A number of frameworks have sprung up like LangChain, AutoGPT and BabyAGI. They allow users to go beyond the simple chat interface and connect all kinds of components to GPT3.5/4, like web search, memory stores or even code…
The first viral thing that I ever built was a Visited Countries web app. You know the idea, you have a list of countries, you click the ones you've been to and it maintains a map suitable for bragging purposes. This was a long time ago when people still just had their own blogs and generally were able to insert html straight into what they were doing. The app generated a little bit of html with a…
The 2020 presidential election was marked by a number of significant trends and developments, including record voter turnout and a deeply divided electorate. As the votes continue to be counted and analyzed, one notable pattern that has emerged is the relationship between life expectancy and political party affiliation. According to data from the Wikipedia, states with higher life expectancy rates…
As the United States continues to grow and develop, understanding the population changes of its individual states can provide valuable insights into the country's overall progress and demographics. In this post, we will show we can use Neptyne, the programmable spreadsheet to import the population data, refine that data and visualize the growth in an appealing way in just a few lines of Python…
It seems like Machine learning and Deep Learning are everywhere these days. AI can do anything! From writing essays to cheat tests to generating to having an infinite image generating machine running on your local machine , the possibilities are endless. So it should come as no surprise that you can use deep learning to predict the outcome of major sports tournaments. Robot predicting the future…
So yeah, it's been kinda quiet around here - famous last words on any blog, but I thought I should at least post something about what I've been up to here. Programmable Spreadsheets. Neptyne ! We tried this whole stealth mode thing. I'm not sure, maybe just developing out in the open is better. But here we are, we're ready to talk about what we are up to. After a bit over a 1000 pull requests,…
It's hard to imagine that it was only January when OpenAI's DALL-E-2 took twitter by storm by showing a new way to generate images from text. Initially there were only a very limited number invitations were available. Fast forward and now not only is getting access relatively easy (though no longer free), there's an open source alternative: Stable Diffusion . You can run Stable Diffusion on Google…
Ever since DALL-E launched a few months ago , OpenAI's machine learning tool has been blowing the Internet's mind with its uncanny ability to convert any sentence into an image. I've been playing around with DALL-E and one of the things that I stumbled into, is making zoomable movies. A number of people asked me how to do it and even though the tooling isn't perfect, I though I'd give it a…