Revisit the educational journey of packing 24 motors, 96 switches, 12 RFID readers, 12 motor controllers, six ESP32s, one Raspberry Pi, a touchscreen, and 24 CDs into less than 500mm for the (not so) simple joy of automatically ejecting CDs. If you’ve ever run out of IO pins in a project, come see how off the shelf electronics can be combined with 3D printing to create compact, modular components that form an expandable I2C-based distributed system.

Minus is an open-source HDMI man-in-the-middle device that detects and blocks streaming advertisements in real-time using on-device machine learning. It sits between your streaming stick and your TV, captures frames from the HDMI passthrough, and runs a tiered inference pipeline entirely on a Rockchip RK3588 with a dedicated neural processing unit. When it spots an ad, it composites over the region and mutes the audio, all while maintaining sub-millisecond passthrough latency. No cloud, no subscriptions, no browser. Just bare-metal ML inference on every frame that crosses the wire using a $100 single-board computer.

Modern displays make time look continuous by presenting a rapid sequence of discrete updates. Usually, that granularity goes unnoticed. This talk revisits one experiment in which it changes what people report seeing.

The audience will experience the experiment on an LED panel while the conference projector shows timing, instrumentation, and the instructive failure.

My six-year-old has two obsessions: speedometers and seven-segment displays. So I built him one he can take with him anywhere: a kid-sized arm band with GPS, a seven-segment readout, and rainbow-comet LEDs, so he can see exactly how fast he’s running, or how fast the car is going on the highway.
This talk walks through the adventures and misadventures of the build, design trade-offs, the enclosure I CAD’d mostly by vibes, and the three parts that were the toughest lifts — the enclosure, the firmware, and everything GPS.

Sit down with ArtCenter faculty members Estefannie and Justin Andarza for a fireside chat exploring the world of alternative controllers and experimental game design. In this conversation, they’ll break down why the artistic intention and social awareness behind a physical object is just as crucial as the code and electronics inside it.

They will trace the journey of taking experiments to three-time alt.ctrl.GDC award winners, and ultimately into Warped Worlds—a collective dedicated to reimagining play. They’ll discuss the power of physical-digital interactivity, the joy of making “weird” tactile art, and how tangible, custom hardware is shaping the future of truly immersive player experiences.

Running DOOM on an FPGA is much more interesting than just a software port. We’ll start with a completely blank fabric and then instantiate the processor to run the code.

I’ll be harnessing Luke Wren’s Hazard3 RISC-V core. Yes, this is the same core used on the Raspberry Pi RP2350 and Pico 2. Of course we’ll need more than just a CPU: External SDRAM, HDMI, JTAG debugging, timing closure, and more. So what does it actually take?

We’ll walk through major steps and unexpected challenges such as having the design fail to meet timing requirements and how sweeping seeds for placement and routing can help. The first pass had a surprisingly slow display frame rate, so we’ll also talk about some unplanned DOOM optimizations.

This is the story of how we compressed all 6,570 frames of the Bad Apple music video and all firmware into a single 65kB binary that runs on a 13-cent RISC-V microcontroller. With development spanning a decade, this project has been an incredible experience and we wanted to share the coolest things we learned along the way, with all the twists and turns that gave us deeper insights into the nature of compression.

This talk will consist of various strategies of compression, show trade-offs between different ways of storing, compressing and decompressing data, going into comparing different lossless, and lossy compression mechanisms like LZSS, exponential-Golomb, Huffman trees, range coding, and glyphs, as well as the code needed for performing decompression, with a focus on embedded environments.

Sweden’s videotex era ran from 1982 into the late 1990s – Teledata, Datavision, Videotex, then TeleGuide, Televerket’s 2-billion-kronor answer to Minitel. When TeleGuide collapsed in 1993, the terminals went to landfill – except a few, which can now relive what they once dreamt of. The network itself didn’t even die: it retreated into government service as Posten’s PostNet, quietly serving vehicle-registry lookups until the web finished it off around 1998. I have an original PostNet smart card. It still logs in – just not where you’d expect.

Thirty years later I tracked one down, refused to modify it, and rebuilt the world around it instead. The terminal now dials a modern AI agent as if it were a period videotex service. The bridge board will be on stage, live. You’ll leave knowing how to put any loop-powered telecom device back on a line you control, for about $40 in parts.

Integrated photonics is a relatively nascent field of engineering and physics. In the last several years, it has grown rapidly, mostly due to the utility of photonics within the telecom sector. However, practical aspects and tacit knowledge of the development process are still obscure and often intentionally kept secret so that labs and foundries can better preserve their competitive edge.

In this talk, I will describe some of the less well-known applications of integrated photonics, give an overview of the foundry landscape for photonics, walk through the basic development cycle, and show some of the results from chips that we fabricated.

Through this talk, I aim to spread awareness of what is now possible with photonic circuits, and provide the audience with some ideas as to how their optical systems might be able to scale by integrating photonics as well as the basic outline of how to begin that development process.

Why don’t MCUs run Linux? The classic answer is that Linux requires the CPU to have a Memory Management Unit, or MMU, which gives every process its own virtual address space. With an MMU, two programs can both refer to the same memory address and never collide, because the MMU translates each process to different parts of physical memory.

So, you can’t run Linux on an MCU… except there’s a “NOMMU” kernel config option?

In this talk, I’ll describe what I learned in my attempt to port Linux to the Teensy 4.1 microcontroller platform, and how it differed from my past experiences with RTOSes. This project was the first time I’ve worked with many of these technologies, and I’ll do my best to explain each component in laypersons’ terms.

This talk explores how we can mimic principles of nature to come up with solutions to our problems and how engineering isn’t all math and science, but is full of interdisciplinary ideas. It will use everyday examples of nature in our lives to show how innovation is only hidden in plain sight.

This talk details how I designed and fabricated a series of open-source custom ASIC GPUs. It will cover the entire workflow: setting the specs, digital design in Verilog, FPGA testing, and fabrication via TinyTapeout. I will also explain the architectural techniques used to squeeze a complete GPU pipeline into a severely resource-constrained chip space.