RSSAmplifier

Blog

iter.ca

Recent blog posts on iter.ca

iter.caRSS feed ↗10 posts

Latest posts

Malleating Git commit signatures

Recently there was an article published about how you can malleate Git commit signatures ( Git Hash Chain Malleability ) : given a signed commit A, anyone can create a new signed commit A’ that is identical in all respects except that it has a different (still valid) signature and therefore also a different commit hash. Despite what the article says, I don’t think this is actually a…

Inkhaven 2 retrospective

I just finished doing the second iteration of Inkhaven, the 30-day writing residency where you have to write a blog post everyday or you have to leave. I enjoyed it a lot; I had a lot of fun both from blogging and from being at Lighthaven and doing the various activities available to me at Lighthaven. Initially I thought Inkhaven would be like a Lighthaven conference, but for a whole month! It…

I went to Stripe Sessions

For reasons that are really unclear to me, Stripe decided to give me a free ticket to Stripe Sessions, their two-day annual conference where they talk about payments and other stuff Stripe does. I don’t know why they gave me a free ticket – they sent an email with an offer for free tickets to my old Stripe account that I haven’t used for processing payments in ~6 years (and I only processed ~$25…

Loans are bets on doom

Some people want to bet on the end of the world. Or that we will live in a post-scarcity utopia where money doesn’t matter (and so nobody cares about bets being repaid anymore); economically these two scenarios are pretty similar (in both cases we live in a world where money isn’t useful anymore) so I’ll refer to both of these as “doom” in this post. The obvious way to bet on doom would be “in 10…

What is Value Aligned Research Advisors?

Value Aligned Research Advisors (or VARA) is a shadowy private investment fund. Despite managing ~$9.94 billion in assets, there’s very little coverage of the fund (which doesn’t have a website, or any other public-facing information about it), and nobody has tried to understand what their investment strategy is. They have only a few people involved with running the fund, and the head office they…

I used to make a lot of bad decisions

sorry if this is a jumble of thoughts, i had to write something for inkhaven quickly Yesterday I wrote a post that involved describing things I’ve done in the past. To write it I had to recall many things I had done over the past couple years, and try to remember why I made various decisions. It was often hard for me to figure out why I did certain things, and throughout writing it I got pretty…

Morphing meetups and me

In May 2024, I was looking through the attendee directory for Manifest (a prediction market conference I was going to next month) to find interesting people who also lived in Toronto, and read about the Toronto AI safety meetup from someone’s blog 1 . TASM had a pretty big influence on my life, and helped me understand a lot about the field of AI safety. TASM and Manifest made me realize I was…

LLMs are pretty good at making slideshows now

I’ve used various LLM agents to make slideshows for me in the last few months, and they’ve gotten very good at it. What works Here are three things I’ve found work well for improving slideshow creation with LLM agents to allow them to edit the slides better: Give the agent a way to see what the slides look like. Otherwise it ends up making slides with overflowing text or misaligned elements that…

Nine cool programs I use

Here are nine cool pieces of software I like to use! Soulver Soulver is a combination of a calculator and a notebook. It’s like Jupyter Notebook but the cells are math expressions instead of Python and are run automatically on changes (along with dependencies). It’s like a spreadsheet but with a bunch of lines in sequence, and with comments. Logseq I use Logseq to manage my personal memex where I…

How I spent my time in 2021

Yesterday I read fellow Inkhavener Nicky’s post about what she did on one day of her childhood . I also tracked how I spent my time as a child, but instead of for one day, tracked what I was doing every ~45 minutes (sampled at random seconds, where each second had a 1 in 2700 chance of being sampled) from July 2020 to May 2023 (ages 15-18), with 33,235 data points in total. Excited to outdo…