Peter Azuolas

I listen to a lot of audiobooks, and there's no good way to know when an author you like has something new coming out. RecentAudiobook.com is the sister site to RecentMusic.com, built to scratch the same itch.

Follow the authors and narrators you listen to and you get an email when a new book comes out, another when a release date gets announced, and a weekly digest. No ads, no data selling, no algorithm deciding what you hear about.

The catalog is over 776,000 audiobooks across 140,000+ authors, 103,000+ narrators and 76,000+ series. Keeping that current takes a fleet of about 29 worker machines scanning around the clock... they peak at around 2,300 books an hour, so a full rescan takes about two weeks.

The public side is AstroJS again, just bigger. Every author, narrator, book, series, publisher and category gets its own page, about 1.2 million pages in all, rebuilt every 4 hours and swapped in atomically so nobody ever lands on a half-deployed site. The member side is a React app that picks up where the public site leaves off ( click Follow on any public page and it's completed after you sign up ). Members can bring their library in from Goodreads, a CSV or Excel file, or a pasted list instead of following authors one at a time.

It's also where I've been doing most of my AI work. Book categories are assigned by Jev from TypeSafe, a model that returns a probability for each question instead of generated text, so it can't invent a category that doesn't exist. Re-categorizing 4,030 books took 48 minutes and cost $0.90. The fallback is a self-hosted LLM ( Qwen3 on a Mac Mini ) with its output locked to a schema of valid categories. Every model choice was made from measured runs on real catalog data, not gut feel.

I love music, and have an addiction to hearing the latest releases from artists I listen to. Unfortunately, Spotify has limited tools for the discovery of new releases, so RecentMusic.com was developed as a way to feed my addiction.

I used it as an excuse to live at the bleeding edge of static site generation technology, and developed the site using AstroJS.

With a truly MASSIVE amount of data, a build of RecentMusic.com contains over 360,000 files and directories, exclusive of the visual assets. So large, and with such frequent updates, that it "breaks" github, I've developed custom build and deployment systems that allow the system to operate in a fully hands off manner.

Includes a fully automated print-on-demand store at shop.recentmusic.com, people can purchase one of over 5 thousand custom designs.

Leveraging multiple new technologies, including AI image generation, AI consensus research, deep API integrations and complex data warehousing, it's been a challenge to create a site with such depth and breadth while still keeping it cheap to operate.

Still under development, the public side has been released but the member features are still in closed alpha. No marketing or traffic growth has yet been attempted, and yet despite this, RecentMusic.com is receiving 10 thousand+ visitors a month purely from search engines and word of mouth and the wait list for membership is growing.

I've been watching the evolution of AI with excitement, especially fascinating is the quickly evolving space in image generation.

Between closed (OpenAI) to "open" (Flux, StableDiffusion, etc) the possibilities are pretty endless as to the type of images you can create if you are creative and patient.

SampleImages.com grew out of my experiments and the 10s of thousands of images I have created over time. Every 10 minutes, a new image from my archives is passed into a Vision AI for description and tagging and an Astro SSG site rebuilds.

While I find it entertaining to just browse the results, the site is intended to help people see some of the possibilities for AI generated imagery.

While I was involved in the domain name industry in the past, it has been a number of years since I had more than a surface level interaction with the industry. I touched on all aspects from sales, portfolio management, parking, acquisition, MVP development, and drop catching.

These days, my interaction with the modern domain industry is mainly as a buyer and acquisition consultant. I sometimes help friends try to get the domains they want for their projects.

I came across the term "blood mouth" (a mild insult used by some vegans to describe those that eat meat) and saw the name was going to drop soon so I dipped my toes back into the drop catching industry. The landscape has changed with most of the registrars themselves preventing the names from relisting and instead auctioning or direct selling them.

Once I had the name, I needed something to do with it, so a tongue-in-cheek tshirt store seemed like a good idea. I hired a couple of artists for the art work, and used it as an opportunity to familiarize myself with a few new-to-me industries.