---- 1-7-2026 - New Year, Old Plans (9) ---- 2026! I'm entering this year with a burst of energy and productivity. My recent post about being more public with my FRST project was more aspirational, since I haven't actually followed through on any real promotion of the project... I've just felt like there were a few important long-running-threads I needed to close out before any kind of serious launch. For one, while I have the PHP e-commerce system mostly functional, I'd like to tweak a few things on both the web and gopher sides. Another stalled thread is the schematic capture / PCB work for the Data Crystals. It's not going to be too bad, though -- I've nailed down the main parts selections, at least. JLC does such quick work, I should be able to get boards back by VCF East, which is my hopeful launch target. The BBS software that drives the crystals is fairly complete, but I've got lots of real-world testing that needs to take place. I'm planning on operating at least one TCP node on a public IP for a bit, at least. I'd like to start dabbling in FTN networking, and this seems most suitable to that end. But the main update I wanted to offer this evening is the one that is most current. One long-running task/thread that has remained in 'suspend' for far too long involves my earliest cyberdeck concepts. The so-called 'Model 1 Cyberterminal' was my first attempt at a truly ultra-personal, minimalist computing device. I drew inspiration from many sources and the result was a fairly elegant mix of hardware -- a 4.3" B/W ePaper display, driven by an ESP32S3 with 8MB PSRAM, an SX1262 radio module, RGB LED front-light, and a Corne split keyboard with ZMK nice:nano!s. I originally picked the ESP32S3 for its promised USB Host support -- I wanted to use my standard USB HID QMK/ZMK keyboards. Alas, the library support was lacking -- I don't honestly know if ESP-IDF has ever really filled that in fully -- as I quickly moved on to Zephyr, for its far better driver support. The ePaper+framebuffer+cfb stuff alone was worth it, but Zephyr also provides a great LoRa driver. I waited on good USB support for a while before deciding to go BLE. But getting the ESP's BT to play nicely with the nRF52-based ZMK fw proved elusive. I never found the right alignment of BLE settings. In the end, I built two such cyberterminals -- useful for proving out most of the driver stack, but never coming together into the fully-operational terminal. The breakthrough to this problem arrived only recently -- germinating since November and culminating over the holiday break. I realized this would be the perfect test for my 9P4Z library. I'd proved out 9P over L2CAP already, so I decided to rip out all the HOGP/GATT stuff in ZMK, creating a new 'keyboard server', serving up a Plan 9-inspired 9P filesystem. This server just provides file i/o for the existing HID reports coming from the primary ZMK functions. 9P4Z's portability allowed me to truly realize the potential of the Plan 9 concept of 'grid computing' when applied to embedded systems. ....a grid, made up in this case of two different microcontroller platforms from separate vendors, each responsible for very different hardware, sharing resources over a simple network file sharing protocol to form a complete personal computing system. The keyboard and display components become peers while Bluetooth fulfills its one-time conceit of offering users a so-called "Personal Area Network." ## TLDR: 9P allowed me to build the perfect pair of ePaper cyberterminal devices -- now made whole with a unique computing environment built on Zephyr -- completed by an 'rc' workalike shell, with builtin commands for 9P-related and namespace management things, along with general 9P-over-L2CAP connectivity -- including Aether Data Crystals -- and interactive chat over LoRa -- completing my initial goal/vision for these devices! Yeah, I'm stoked. --- A Note regarding "AI": --- | | I want to acknowledge the role that Anthropic's Claude 4.5 | (Opus) model had in these recent breakthroughs. I continue to | be impressed at the performance of this latest model in most | any programming task I throw its way. I will confess that | I remain conflicted on the use of tools like this current | crop of agentic LLMs. I have grown very appreciative of its | ability to help me more quickly arrive at the product of my | imaginings, cutting through so much of the mundane frustration | of navigating modern tools, endless APIs, and other minutae. | While I can confidently claim authorship of this work, I am | also happy to share authorship with Claude. (The code is | full of bugs and LLM cruft, after all.) Will this Faustian | bargain cost me my soul? Will my brain turn to mush as I | increasingly rely on the LLM? I don't know, but I'm inclined | to continue to use the tool if it continues to offer new | brushes to paint with and unlocks new creative potential. | --- .