I didn't understand the OS I built with AI until the MCP gave it analytics For my birthday this year I got a Pimoroni Presto — a little developer gadget with a 240×240 touchscreen and not much else. I also got an environment sensor which can track CO2, temperature, and humidity. I wanted to build something for it. The thing is, I'm not an engineer. I'm a marketer. Until I joined PostHog I couldn't code at all . I've made effort to learn and slowly got myself to the point where I could ship product features , but it wasn't until DeskHog came along that I saw how AI could supercharge my progress. My goal was a desk dashboard that worked at any level of attention. Something I could glance at between meetings and read idly, but that also had deeper things to poke at while I waited for someone to join a call or had lunch. What I built I call it Joe-OS, because naming things is the one part of this I'm actually qualified for. The Presto runs MicroPython. The top two-thirds of the screen is a carousel of panels; the bottom third permanently shows temperature and CO2 from the sensor, with little sparklines of the last half hour. You move between panels by tapping the corners of the screen. The panels are deliberately a mix of useful and useless: Three news readers that stream the latest RSS headlines from The Guardian, BBC News, and Rock, Paper, Shotgun. Oblique Strategies — Brian Eno's deck of creative prompts, plus a few dozen I wrote myself to build it out more. A rotating daily joke, a random fractal, and a Sky panel showing the current moon phase, sunrise, and sunset. The most ambitious part of Joe-OS was Vigil It's a tiny medieval roguelike text adventure about a knight keeping watch over a dying realm. Every turn something happens — a ghoul rises, a house burns, a bandit approaches — and you respond by choosing an Honorable or a Craven option. Sometimes you get a third choice based on your inventory or companions, each of which has their own simple personality. It runs across five acts and 400+ hand-written events, with a two-and-a-half min...
Saved Posts
Keyboard shortcuts:
j next, k previous, o open post, e expand/collapse, s save, f follow-up, Enter mark read & advance, ? help
Summary:
• Brief mention of Kona feature overview from 15Five company
• Very limited information provided about what the feature actually does
• Appears to be a product update or documentation reference
Summary:
• Company built a new system called AdvancedIF to test how well AI follows instructions
• Goes beyond simple tests to more complex real-world scenarios
• Could help make AI systems better at understanding and following human requests
Summary:
• New AI programs are showing promise for finding treatments for thousands of rare diseases
• Could make it easier and cheaper for people to pursue treatments for uncommon illnesses
• Researchers are cautiously optimistic but still testing these approaches