Rebooting Page Street Labs
If you are a new subscriber, this might be an odd post to start, but welcome!
Some time ago, I started a newsletter experiment called Page Street Labs (PSL). With AI becoming so lucrative, we started seeing sites and newsletters pop up everywhere. Often these newsletters simply aggregate content from elsewhere with very little original content, and many times, caught in copy-pasta, they propagate errors. In response to that, I wanted to create something fresh -- an "un-newsletter" called PSL. Here's what I wrote in the "About" section.
Page Street Labs is an organization dedicated to bringing clarity and new possibilities in artificial intelligence and related technologies. Every week our inboxes are flooded with AI newsletters, so why do we need another one? First, this is an un-newsletter. Think of it as a magazine with sporadic issues. Each issue focuses less on news and happenings in AI and more on developing a deeper understanding of something of broader interest. Every issue has a singular focal point, though we may have occasional experimental sections. We will dive deep into one topic and bring out novel findings, new ways of looking at existing things, providing a synthesis of subject matter typically limited to a few experts, but in an accessible language. Each issue aspires to bring a level of creativity/expression and scholarship that lies at the intersection of a typical literary journal and an academic journal, to create more informed conversations around the topic. The standard length of an exposition will be somewhere between 2,000-10,000 words. All writings come from wetware than GPUs; if we use generative AI technologies for writing, we will explicitly call those sections out.
Sadly, after two years and significant effort, the idea wasn't sustainable in my context (which doesn't mean it won't work for you). If you are bootstrapping a PSL-like venture, like me, and also have a family to support, the goal of pumping out 2000-10000 words or well-researched original content that's not related to your day job becomes untenable. Even for a modest 3000-word piece, I probably wrote 2-3x that length as "prep work", followed by meticulous editing and fact-checking. If you add to that time investment, the resource costs, it quickly becomes untenable. This is not just my experience, but Chris Olah said something similar about distill.pub.
Instead of trying to get too creative about how building support for a different entity, I am consolidating all writings in PSL under my personal site, where I continue to write for fun.
What does this mean to you?
Not much. You will still get that quality of content you signed up for in your inboxes via Substack; just the name will be different, and it will be thematic and interesting. For me, it's less overhead and the joy of having a single outlet for all my writing (except Twitter).
What to expect in the new few issues?
A lot of Large Language Model stuff :-)