TCL's NXTPAPER 14 tablet is a 14.3-inch eye-care-focused device built with a nano-material display layer that reduces glare, blue light, and finger marks while supporting both full-color and ink-only reading modes. Priced at $379.99 with the stylus and flip case included, it runs Android 14 on a MediaTek octa-core processor with 8GB of RAM and 256GB of storage. A dedicated NXTPAPER key toggles between color and paper modes, and the tablet also features quad speakers, a 10,000mAh battery, and an Extend Mode that turns it into a secondary PC monitor.
- The 3:2 aspect ratio (2400x1600) is well-suited for document and note viewing but lower than many competing tablets
- Wi-Fi only; no cellular connectivity
- 8MP rear camera with AI scene detection, 13MP front camera
- 33W fast charging; full charge in about 2 hours 55 minutes
- No fingerprint sensor; facial recognition only
- Headset jack is via USB-C, not a 3.5mm port
Shekhar Vaidya writes about utilizing a $15 ESP32 and a small OLED screen to display live home server stats (CPU, RAM, disk, and load) without requiring an open dashboard or phone app. He selected the Komodo monitoring service for its 5-second refresh rate over Beszel's richer but slower 60-second data, and the project took several days of troubleshooting after a wiring error fried the original display.
- The home server is a repurposed Dell Latitude laptop running Debian with 20+ self-hosted services
- A superadmin API key and secret are stored in plain text on the ESP32
- The display is not a replacement for monitoring tools; it serves as a glanceable trigger to investigate further
Colonel Panic sells open-source, ESP32-S3 based counter-surveillance hardware that passively detects drones, body cams, license-plate readers, and other monitoring devices. The flagship OUI SPY board ($85) runs "Unified Blue" firmware with six modes, including Sky Spy for FAA drone Remote ID, a multi-target BLE scanner that flags axe-on body cams and Taser 7/10, a headless AXEOFF beeper, a Foxhunter proximity tracker, Flock-You for Flock camera detection, and raw Wi-Fi/BLE capture for Wireshark. The mesh-detect v2 ($110) pairs a Seeed XIAO ESP32-S3 with a Heltec LoRa 32 V3 for 5–15 km distributed drone detection and real-time web mapping over Meshtastic. The project has drawn significant press attention—from EFF, Forbes, the Wall Street Journal, and Hackaday—after being credited with spurring local council action against Flock Safety cameras and being featured on the DEF CON 34 main stage.
- A Nerd Nite talk in Asheville is credited by the WSJ with directly prompting the city council to vote against Flock Safety.
- AXEOFF detects axe-on body cams, Taser 7/10, and Signal devices without a UI—just proximity beeps.
- A transparent flexible PCB patch version of the OUI SPY ($75) is sewable onto clothing as a punk art statement.
- A "Flock-You" mode exists but the site admits Flock rapidly changes firmware, so detection coverage is limited.
- Skull earring PCBs ($25) are sold as artistic DIY kits compatible with XIAO ESP32-S3.
Simon Batt writes about unidex, an open-source operating system for the 1.54" Waveshare ESP32-S3 ePaper display that turns the tiny board into a pocket productivity device with a Tamagotchi-like virtual pet. It displays the next class with a live countdown, converts voice notes to Markdown for Obsidian, and allows the pet to interact with other users' pets.
- Installable directly from a browser
- Includes a WiFi collection game and four one-button games
- Free to fork and customize for non-commercial purposes
This documentation introduces how to fine-tune large language models into decision models using Unsloth. These models are designed to make decisions with calibrated probabilities rather than simply generating text. Users can train models like Qwen or Gemma to read input, score provided options, and return a specific choice. The process is supported through the Unsloth Desktop app, free notebooks, or custom Python code, and the resulting model can be integrated directly into the Decision API for practical use.
Tyler August writes about a retrocomputing enthusiast who acquired a 1994 DEC AlphaServer 2100 at Vintage Computer Festival Midwest and recorded a tour of the machine. The server is a single-CPU model with a 64-bit Alpha RISC processor, capable of hosting up to 2GB of RAM, though this example only has 256 MB. The owner has gotten it powered on but is experiencing what appear to be power-related issues preventing it from booting an operating system like OpenVMS.
- Dated to the height of the RISC wars when multiple vendors were competing with their own high-end RISC architectures
- Alpha machines boot differently from typical x86 systems
- Comments note the machine could potentially run Windows NT 4.0, DEC OSF/1, or OpenVMS
- One commenter shared a story about running Exchange 5.5 on a similar Alpha machine under NT 4.0
REA is an open-source Model Context Protocol server that connects coding agents to local reverse-engineering tools for inspecting native binaries, JavaScript/Electron apps, .NET assemblies, Android APKs, firmware, websites, and network captures. It integrates with Hopper, Ghidra, and IDA for deep native analysis, and provides static JavaScript and .NET inspection without requiring a native analysis engine. Results include evidence and limitations behind each conclusion, and the project is available under the MIT license.
- 96.4k stars and 21.5k forks; active development with 35 releases and 100 contributors
- Showcases include reconstructing a DX-Ball sound-pan calculation that passes 3,205 original test cases, and recovering a 16-bit DOS bullet-ring calculation from the game TH04
- Supports Claude Code, Codex, Cursor, Gemini CLI, and Grok Build out of the box
- Also analyzes EVM bytecode, WASM artifacts, recorded Linux crashes, Android devices via ADB, and JEB projects
mikedeblin writes about a portable MP3 player built on the ESP32-S3 that plays music from a microSD card or streams it from a DLNA/UPnP media server (tested with Synology Audio Station). The device features a 1.8" colour TFT with four display modes, an OLED status display, a rotary encoder with a 16-LED ring, and an output compressor and limiter that protect the DAC from being overdriven.
- Uses a PCM5102A DAC in hardware I²S mode with a ReaComp-style compressor and brick-wall lookahead limiter
- Network streaming uses a 256 KB PSRAM ring buffer to absorb WiFi drop-outs
- Supports ID3v2.3/2.4 tags in UTF-16, UTF-8, Latin-1, and Windows-1251 (Cyrillic)
- LED ring is driven through an MCP23017 I²C GPIO expander with 125 Hz multiplexing
- MP3 only for now; AAC and FLAC decoders exist in AudioTools but are not yet wired in
Silvio Werner writes that Banana Pi has announced the BPI-CM7, a compact compute module intended as an alternative to Raspberry Pi compute modules. It is not a standalone board and requires a carrier board, which makes it well suited for ultra-compact custom devices that only need a few breakout interfaces. It uses an Allwinner H618 with four Cortex-A53 cores and an ARM Mali-G31 GPU, and pricing and availability have not been announced.
- 55 x 40 mm form factor with two 100-pin high-density connectors
- Confirmed variant has 2 GB LPDDR4 and 32 GB eMMC
- Integrated Wi-Fi 5 and Bluetooth 5.0
- Supports 100 Mbps Fast Ethernet and USB 2.0
- Exposes GPIO, UART, SPI and draws power from 5 V
Abhishek Kumar Mishra writes that he created a background bash script to automate service monitoring and self-healing on his Raspberry Pi Zero 2 W. The script is scheduled via systemd to check the status of Docker, Pi-Hole, Unbound, and Tailscale at regular intervals, restarting failed services without requiring manual SSH intervention. It acts as an additional protection layer beyond Docker's built-in restart policies and only triggers a system reboot when multiple checks fail.
- The script was generated using ChatGPT and selected over a Gemini output because it included appropriate delays between checks.
- Lightweight approach is preferred on the Zero 2 W because heavier tools like Uptime Kuma or Grafana consume too many resources.
- The script saves logs to a file, providing a baseline for further troubleshooting when manual intervention is still required.