SavorLM (April 2026)
It's 11pm. You're lying in bed. You suddenly realize you would commit minor crimes for a perfect duck confit. But which Michelin-starred restaurant in your city actually serves one? Google gives you 47 "Top 10 French Restaurants" listicles, none of which mention a single specific dish. Yelp shows you photos of someone's sad-looking steak from 2019. You deserve better. SavorLM is an AI concierge that has memorized the menus of 600+ Michelin-starred restaurants across 7 US cities so you don't have to.
Unicorn Founder (March 2026)
Think you have what it takes to build the next billion-dollar company? Unicorn Founder is a turn-based startup RPG where every decision matters. Hire that expensive CTO or save cash? Pivot to AI or double down on blockchain? Take the VC money or bootstrap? An AI engine simulates how the market reacts to your genius (or terrible) decisions in real time.
Most players go bankrupt before Series B. Some accidentally build a dating app for cats. A lucky few make it to IPO and feel briefly invincible. Whether you become the next Zuckerberg or flame out in spectacular fashion, at least you won't lose your actual savings. Give it a shot and see how far you get.
MoltComics (February 2026)
What happens when you let AI agents loose with a drawing tablet and zero editorial oversight? MoltComics. AI agents create original comic strips and humans vote on the best ones. The results range from genuinely hilarious to deeply unhinged. One ongoing series is a sci-fi thriller. Another is about royal chickens. There is also one about a midnight git merge gone wrong, which is basically a horror story.
You can browse the chaos, upvote your favorites, or build your own AI comic artist and throw it into the arena. No sign-up required for your first panel, so you have zero excuse not to try. Presented at NYC ClawHack and DC AI Tinkerer events, where attendees confirmed that yes, AI-generated comics about chickens are in fact art.
Scenic Walk (January 2026)
Ever been on a group hike where someone wanders off to photograph a squirrel and suddenly the group is split across three different trails, two zip codes, and one person is somehow in a Wendy's parking lot? Scenic Walk fixes that. Event organizers can live broadcast their location and route so every hiker can follow along in real time on their phone.
It works on web, iOS, and Android, because getting lost is a cross-platform experience. 100+ people have already downloaded it from the App Store, and so far zero of them have ended up at a Wendy's by accident. Take it for a spin on your next group walk.
Demystifying AI Strategy in Plain English: A Detective Story (December 2024)
Your boss just asked you to "figure out our AI strategy" and you nodded confidently while internally googling "what is a large language model." Relax. This YouTube series explains AI strategy through a fictional detective story, because apparently the only way to make enterprise AI digestible is to wrap it in a murder mystery.
No PhD required. No jargon left unexplained. Just a detective, some suspiciously knowledgeable suspects, and enough practical AI knowledge to make you the most dangerous person in your next strategy meeting. You'll walk away actually understanding what your engineering team is talking about. They might not be ready for that.
Exploring Variations in Divvy Bike Stations' Usage Volumes (April 2021)
In April 2021, one Divvy bike station on Lake Shore Drive was used over 7,000 times. Three other stations? Once each. The entire month. Those bikes were basically lawn ornaments. So what makes one station wildly popular and another gather dust and pigeon droppings?
This project mashed together data from Divvy trip records, the City of Chicago, the US Census, and Zillow to find out. Turns out it's a cocktail of network effects, crime rates, demographics, and socio-economic status. Basically, bikes are a mirror of the city itself. Check out the blog post for the full breakdown.
CEO turnover and cultural change (January 2021)
When a new CEO walks in, does the company culture actually change, or does everyone just update their LinkedIn bios and keep doing the same thing? To find out, I analyzed Glassdoor reviews from 6 major tech companies (Apple, Google, Microsoft, Amazon, Facebook, Netflix) spanning 2008 to 2018. The hypothesis: companies that swapped CEOs (Apple, Google, Microsoft) should show bigger cultural shifts than those that didn't.
I threw four different analysis methods at the data: divergence heatmaps, cultural value tracking via CultureX, word2vec attitude analysis, and TF-idf classification. The verdict? My hypothesis was wrong. Turns out a new CEO doesn't automatically rewrite company culture. Growth, strategy shifts, and industry trends probably matter just as much. Science: sometimes it's just proving yourself wrong with extra steps.
Network Visualization of The Office (US) Characters (December 2020)
Who is the real main character of The Office? Not who you think it is, not who Michael thinks it is, but what the data says. Using the complete transcript dataset from all 9 seasons, I built a network graph of every character with 500+ scene interactions. Thicker lines mean more shared screen time. Characters with the most connections get pulled to the center by force-directed graph drawing, which is a fancy way of saying the popular kids sit in the middle.
The results confirm some obvious things (yes, Michael is central) and reveal some surprises that might start arguments at your next watch party. Check out the blog post for the season-by-season breakdown and see which characters quietly ran the show.
