New Maker Tech: Luckfox Builds a Linux PLC, ESPHome Gets a $40 Starter Kit, and Gemma Runs Offline on a Pi 5

Luckfox has taken its small Linux platform and turned it into a real programmable logic controller. ESPHome now has an official $40 starter kit designed to remove much of the friction from building your first smart-home device. Google Chrome has quietly filled one of the biggest remaining gaps in the ARM Linux desktop experience. And Google engineers have built a completely offline Gemma-powered voice translator around a Raspberry Pi 5.

None of these are simply faster versions of hardware we already had. Each one takes a familiar platform and makes it useful in a slightly different way.

Luckfox has expanded its Lyra family with the Luckfox Lyra PLC, an industrial controller built around Rockchip's RK3506B.

Rather than simply putting the Lyra inside an enclosure and calling it industrial, Luckfox built a complete control platform around it. The PLC runs Debian 12 with Linux 6.1 on a triple-core Arm Cortex-A7 processor clocked at up to 1.2GHz, paired with 512MB of DDR3L memory and 8GB of onboard eMMC storage.

There's no VPU or NPU onboard, and the GPU is limited to 2D acceleration. This chip was never meant for heavy computing workloads, but that's fine. Raw performance isn't really the point here.

The interfaces are what matter.

Two 11-pole terminal blocks expose an RS-232 port, a shared RS-422/RS-485 interface, CAN bus, and two relays, one normally open and one normally closed, rated at up to 10A. There's also a 4-pin I2C header, a 4-pin GPIO connector, dual 10/100 Ethernet ports, and 2.4GHz Wi-Fi 6 with Bluetooth 5.2 LE.

A small 1.9-inch, 320×170 touchscreen is built directly into the controller, along with an RTC and temperature/humidity sensor.

Power can come from USB-C at 5V, a 7–24V DC terminal input, or an optional PoE module supporting IEEE 802.3af. The entire system fits inside an 87.8 × 71.84 × 60.5mm DIN-rail enclosure.

What makes the Lyra PLC more interesting than an SBC with a few industrial connectors bolted on is the software.

Luckfox provides a browser-based dashboard for network configuration, file management, logs, process management, system settings, and even a browser terminal, so there's no need to keep a monitor and keyboard attached.

The PLC section is where it gets genuinely useful.

It can configure Modbus RTU, TCP, and UDP devices, control the onboard relays, collect and report data, and publish information over MQTT, TCP, UDP, or HTTP. It can also act as a protocol bridge, with CAN-to-serial, CAN-to-Ethernet, serial-to-Ethernet, and Modbus RTU-to-TCP conversion supported out of the box.

That's much closer to a finished industrial controller than a development board with a Modbus library installed.

Luckfox lists the Lyra PLC starting at $83.99 through Waveshare, with a PoE version at $89.99. It also appears on AliExpress for a little more.

That price range puts it in an interesting position for small automation systems, test benches, home labs, and DIY industrial projects.

It's not about to replace an established PLC platform on a factory floor, but that was never really the pitch. A small Linux computer with Modbus, CAN, relays, networking, and a ready-made WebUI opens industrial-style automation to projects where a conventional PLC might be overkill or simply too expensive.

One caveat worth noting for readers outside the US: Rockchip restricts sales and support of RK3506B-based products to certain sanctioned countries and regions, so availability may vary depending on where you're ordering from.

Speak your PR description, bug reproduction, or Cursor prompt. Wispr Flow auto-tags file names, preserves variable names, and formats everything for immediate paste into GitHub, Jira, or your editor.

No re-typing. No context gaps. No mangled syntax. Works natively inside Cursor, Warp, and every IDE at the system level.

4x faster than typing. 89% of messages sent with zero edits. Used by engineering teams at OpenAI, Vercel, and Clay.

Try Wispr Flow free

ESPHome has become one of the easiest ways to build custom smart-home hardware around inexpensive microcontrollers.

But "easy" is relative.

Someone building their first ESPHome device still needs to choose a microcontroller, find compatible sensors, wire everything correctly, and get comfortable enough with the configuration to make it all work.

The new official ESPHome Starter Kit, built by Apollo Automation, is designed to remove most of those barriers.

It's the first officially branded ESPHome starter kit, developed in partnership with the Open Home Foundation, and it sells for $40.

The kit is built around the ESPHome C6 board, which uses an ESP32-C6-MINI-1 module with 2.4GHz Wi-Fi 6, Bluetooth LE, and Thread and Zigbee radios on the same chip.

Instead of giving beginners a breadboard and a bag of jumper wires, the modules connect through two 14-pin FPC ribbon cables. No soldering is required.

Four modules are included: a PIR motion sensor, an AHT20F temperature and humidity sensor rated to ±0.3°C and ±2% RH, a button module, and an ESPHome Casita notification board with ten RGB LEDs and a piezo buzzer.

The main board has two FPC connectors, so two modules can operate at the same time. Enough to build something like a combined motion and environmental sensor right out of the box.

The bigger change may actually be on the software side.

ESPHome's new Device Builder desktop app for Windows, macOS, and Linux provides a visual interface for configuring supported hardware, meaning a first-time user doesn't need to begin by writing YAML.

Plug in the PIR module, configure it visually, and you've got a motion sensor. Swap in the AHT20F and it becomes an environmental monitor. Add the Casita board and the same platform can flash lights or sound an alert.

From there, the device can operate on its own or integrate directly with Home Assistant. Everything runs locally, with no required cloud account or subscription.

Apollo Automation is the Open Home Foundation's second commercial hardware partner, after joining in December 2025, and the company says the majority of the profit from each kit goes back to the Foundation, which supports ESPHome, Home Assistant, and more than 250 other open-source projects.

Arduino and Raspberry Pi starter kits helped a lot of new makers get through that awkward first stage where even wiring a button correctly can turn into its own mini-project.

ESPHome now has something similar built specifically for home automation.

For anyone who's looked at ESPHome before but never wanted to deal with breadboards, pin assignments, or configuration files, this may be one of the easiest entry points yet.

Arm-based Linux computers have become much more capable over the last few years, but desktop software support can still reveal just how dominant x86 remains.

One surprisingly persistent example has been Google Chrome.

Chromium has run on Arm Linux for years, but Google never shipped the same official Chrome build it offers for x86 Linux even as Arm Macs and Arm Windows PCs received native versions years ago.

Google said back in March that native ARM64 Linux builds were coming in Q2 2026, promising the same account sync, Chrome Web Store access, and Google ecosystem integration available on its other supported platforms.

That target slipped by a few weeks.

Rather than arriving with a major launch announcement, ARM64 .deb and .rpm packages quietly appeared on Chrome's official download page in early August. Some users even discovered the packages by manually changing the amd64 download URL to arm64 before Google's own support documentation caught up.

It's live now.

Chrome's installation page currently lists a 64-bit ARM .deb for Debian and Ubuntu alongside the traditional x86-64 package, as well as an .rpm for Fedora and openSUSE.

Testing on a Raspberry Pi 5 shows the browser installs cleanly through dpkg, includes working Widevine DRM support, and performs similarly to Chromium in areas such as WebGL and video acceleration.

The biggest practical difference is account sync.

For makers, a browser update isn't as exciting as a new SBC, but it matters in a quieter way.

Faster Arm processors don't do much good as desktop replacements if users still need workarounds for mainstream software. Google also highlighted its NVIDIA DGX Spark partnership when discussing the ARM64 push, alongside the broader effort to support Arm Linux distributions.

Chrome is a significant one to finally cross off that list.

The Raspberry Pi has also become the target of an interesting experiment from a small team of Google engineers: a completely offline voice translator.

The open-source Gemma Translator runs on a Raspberry Pi 5 with 8GB of RAM and uses Google's LiteRT-LM runtime to run the gemma4-e2b model entirely on-device.

The model has roughly 2.3 billion effective parameters, or about 5.1 billion total when its per-layer embeddings are included.

Once the system is set up and the model has been downloaded, no internet connection is required for translation.

The hardware requirements are straightforward: a Raspberry Pi 5, microphone or USB audio interface, speaker, and display.

Google's interface was designed around small screens, with a 480×320 panel used in the demonstration unit alongside a custom control board with push-to-talk buttons and a rotary encoder for language selection. STL files for a 3D-printed enclosure are included with the open-source release.

Speech-to-text is handled by Moonshine, an open-source low-latency ASR engine, while Kokoro handles text-to-speech for the translated output. Both run locally alongside the Gemma model.

In testing, translation runs at roughly six tokens per second while using less than 1.5GB of RAM on the Pi 5.

Two conversation modes are supported: one person can control both sides of the translation, or two people can each use their own microphone and language selection for face-to-face conversations.

Google demonstrated the system translating live between Spanish and Mandarin speakers.

The complete project including source code, setup scripts, deployment tools, and enclosure files are available as open source. Google is also clear that this is a prototype rather than an officially supported consumer product.

That "one clearly defined job" is what makes Gemma Translator more interesting than another Pi-runs-a-chatbot demo.

Running locally provides an actual benefit.

The device can work somewhere without reliable internet access, and conversations don't need to be sent to a remote server just to be translated.

That's a much stronger use case for edge AI than simply proving another language model can run on a Raspberry Pi.

These four developments don't have much in common on a spec sheet.

One is an industrial controller. Another is a beginner smart-home kit. Chrome is a browser update, and Gemma Translator is an open-source AI project.

But they're all moving in roughly the same direction.

Maker and embedded platforms are becoming easier to turn into devices that actually do something useful.

Luckfox is packaging Linux, industrial interfaces, and automation software into something that can be deployed like a real PLC. ESPHome is removing the breadboard-and-YAML barrier for new users. Chrome closes another gap in the Arm Linux desktop experience. And Gemma Translator shows how a small local model can become a dedicated appliance instead of just another benchmark run.

The hardware itself still matters.

But increasingly, the projects that stand out aren't simply the ones with the fastest processors. They are the ones that make all of that hardware easier to actually put to work.