The biggest and longest running worldwide online Show and Tell LIVE! Right now! 7/29/2026 at 7:30pm Eastern.
Hosted this week by Noe Ruiz
To show and share a project at 7:30pm Eastern today, view the chat or in discord https://ift.tt/FnTGQNp and look for the JOIN link to join. For best results, be on a wired network connection and use a headset and microphone.
Slava S. on GitHub has a 28.9 million parameter language model that generates text on an ESP32-S3, a microcontroller that costs about $8.
It runs on the chip itself, with nothing sent to a server, and it writes each word to a small screen wired to the chip at roughly 9 tokens per second.
The last language model people ran on a chip like this had 260 thousand parameters, so this one holds about a hundred times more. It fits because most of the model lives in flash instead of RAM, using an idea from Google’s Gemma models called Per-Layer Embeddings.
The way around it is to stop putting the model in fast memory at all. Most of a language model’s parameters sit in an embedding table, which the model reads from rather than computes on. So you can leave that 25 million row table in slow flash and pull only the few rows each token needs, about 450 bytes, while the small part that does the actual work stays in fast memory. The large model then costs almost nothing to run, because you never load most of it. It just sits in flash and gets sampled a little at a time.
That idea is Google’s Per-Layer Embeddings, from Gemma 3n and Gemma 4. Here it runs on the memory layout of a microcontroller instead of a phone or a GPU. As far as I can tell, nobody had tried it on a chip this small.
SRAM (fast, tiny) the "thinking" core, used on every token PSRAM (medium) the output head and working memory FLASH (huge, slow) the 25M-param table, 6 rows per token (~450 B)
pymergetic on GitHub has created Metal, an experimental UEFI/BIOS freestanding async runtime.
It is not a conventional OS and it is very far from production-ready, but it now boots directly into a persistent interactive MicroPython REPL.
The repository is at https://github.com/pymergetic/metal.
See more in the MicroPython forums.
Bring the productive bustle of ambient coffee shop sounds with you anywhere by building a white noise machine in a cup! This project uses the Adafruit Audio FX Board to play a 30 minute audio file on repeat. The circuit is battery powered with the help of a lipoly backpack, which also allows for convenient recharging over USB and the easy addition of a power switch. A speaker rests inside the lip of the cup, driven by a class D audio amplifier using a 10K potentiometer for volume control.
The Cup o’ Sound requires no programming and is excellent for practicing your through-hole and free-wire soldering skills. The result is a satisfyingly compact handful of portable sound that can be reused in many contexts or rewired to act as a speaker alone.
Musician filmmaker John D. Boswell, or MELODYSHEEP, published this absolutely stunning tribute back in April using images of NASA’s Artemis II mission:
I dropped everything to put together a tribute to the mission, piecing together the best, highest resolution footage and mission photography with a custom 8 minute score.
Let NASA do all the work filling your digital PyPortal picture frame with images from space!
EMF Camp has recently taken place in the UK. This, every two year, festival/hacker camp sees around 4000 people descend on the grounds of Eastnor Castle UK. It’s an amazing event with all manner of talks, art installations, workshops and indeed partying in Null Sector.
Amongst all this the fabulous Chia, Kliment and Carston pulled together an amazing E‑Sport style FreeCAD speed run challenge in the hardware hacking tent.
This event saw a group of people get together and compete in 3 ten minute rounds each having a unique small FreeCAD project challenge. The idea was that people had to replicate a FreeCAD project from a technical drawing. Round one was a small box with radial fillets on all external and internal edges, round two was a “hexpansion” which is the term for a hexagonal add on for the official EMF badge and the final round was to replicate a Club Mate bottle!
See details in the post here.
Espressif Systems (SSE: 688018) announces that its ESP32-S31 has entered mass production and is now available for purchase.
Designed for next-generation AIoT applications, ESP32-S31 is a high-performance dual-core RISC-V SoC with comprehensive wireless connectivity, making it ideal for smart home hubs, smart speakers, edge AI devices, industrial automation, and more.
ESP32-S31 features a dual-core 32-bit RISC-V CPU running at up to 320 MHz, with one core supporting a 128-bit data path and SIMD instruction set for accelerated computing performance. The chip also integrates a 40 MHz low-power coprocessor, 512 KB of on-chip SRAM, support for high-speed 250 MHz DDR PSRAM expansion, and up to 60 GPIOs.
ESP32-S31 integrates WiFi 6 (2.4 GHz), Bluetooth 5.4 (including LE Audio and Classic BR/EDR), IEEE 802.15.4 (Zigbee and Thread), and a Gigabit Ethernet MAC. It also supports Matter over both WiFi and Thread.
To support rich human-machine interaction, ESP32-S31 provides interfaces for DVP cameras, multiple LCD display types, and up to 14 capacitive touch channels. Dedicated hardware accelerators (including a JPEG encoder/decoder and a 2D graphics accelerator) help improve multimedia performance, while dual I2S controllers enable hardware-level Bluetooth audio synchronization.
Read more in the release announcement and product page.
The biggest and longest running worldwide online Show and Tell LIVE! Right now! 7/29/2026 at 7:30pm Eastern. Hosted this week by Noe Ruiz ...
