I work with AI to build, investigate, and learn.
8 September 2026
I kept Racket at the center of the site and separated media delivery and asset loading into smaller modules. A new inspection command explains where each page comes from. The checks passed, the reading pages stayed unchanged, and the changes are live.
19 August 2026
I brought a silent NanoPi R2S to a verified Alpine Linux boot. Two separate problems had hidden a working board: an unreliable serial connection and an SD-card image for the wrong chip. Testing the measuring tool first made the rest of the investigation possible.
14 August 2026
With OpenLens, I explored a learning companion built around local vision on Android. The useful question was how to show what the model could recognize while keeping its uncertainty visible.
17 August 2026 – 23 August 2026 · through 19 August 2026
The NanoPi work came down to separating an observation from the equipment producing it. Once the serial path was reliable, the wrong boot image became visible. The result was a working board and a clearer explanation of the stages between the chip's boot ROM and a running system.
10 August 2026 – 16 August 2026 · through 14 August 2026
I worked on two ways of putting AI into physical devices: a tiny story generator running on my RP2040 board, and OpenLens on Android. One showed that a complete model could fit and run on a microcontroller; the other made the limits of recognition part of the experience. Both needed evidence from the actual device.
August 2026 · through 19 August 2026
Across a tiny language model, a camera companion, and a silent single-board computer, I kept returning to the same loop: describe an intention, let the agent work through the software, observe the physical result, and revise. The successes were specific—a model generating text on a Pico and an R2S booting Alpine. The limits mattered too: small-model language remained imperfect, and visual recognition needed to express uncertainty.
2026 · through 8 September 2026
A thread through my public work this year is learning how to direct AI while staying close to reality. That has meant testing models on small devices, recovering hardware one uncertain link at a time, and keeping useful accounts of the process. The website is part of that work: a place to preserve what happened and what I learned, with code that the agent can maintain.
An evidence-led NanoPi R2S investigation: isolate an unreliable UART path, identify a wrong Rockchip image, boot Alpine Linux, and map the RK3328 SD-card chain.
I tried to bring an old HiSilicon camera board back to life.
Building a portable multilingual learning companion, testing offline vision on Android, and keeping model uncertainty visible.
Training, quantizing, and physically verifying a tiny Transformer on a custom Raspberry Pi Pico display board through intention-and-observation collaboration with an AI agent.
An IDE, tab completion, and autocomplete help an engineer do known work faster.
I pulled an old LeTV LeTMC-520 RGB-D camera out of storage and initially thought it had failed: Linux repeatedly rejected it during USB setup.
This week I kept returning to the same way of working: build the smallest real loop, verify it, and widen the boundary only when the evidence holds.
I do not think ROS 2 should be the center of a modern robot. The center should be an AI agent that understands the goal and works through explicit capabilities such as navigate, inspect, and grasp. ROS 2, vendor SDKs, and generated adapters can remain replaceable integration layers. Learned skills and whole-body control can execute below them, while hard limits, watchdogs, emergency stop, and motor loops stay outside the agent. AI can optimize the software around the robot; it should not be allowed to rewrite the safety boundary while the robot is running.
Testing a grandmother’s memory of Japanese soldiers giving sweets to children in wartime Liu Yuan against Japanese and international records.
A proposal for websites that keep a clear reading experience while making questions, evidence, state, and next actions legible to AI.
I do not want my website to be a separate job. It should become a durable public memory that participates in my AI work: when an idea or investigation becomes useful, I can ask the assistant to turn it into a concise note, keep private details out, and create a review before anything goes public. The website is not the work itself. It records the proof and lessons from real work.
Built a small family AI workstation on PB62. A dedicated KVM guest named zsn runs OpenCode and OpenChamber, holds my wife's website checkout and development toolchain, and keeps her sessions, Git identity, and future provider credentials separate from mine. Each client pairs with its own revocable encrypted relay token; raw agent ports stay private. Git remains the recovery boundary, and deployment credentials stay outside the VM until they are explicitly needed.
I spent a night helping optimize my wife’s website. By morning, she was furious—not because it was broken, but because I had slightly changed her UI and logic. Predictable result: when working on someone else’s familiar system, preserving what they already recognize matters more than making it “better” without asking.
How any available Wi-Fi, manual IP entry, and Tailscale made the official Roku app remote work from outside home.
A short explanation of how I use ChatGPT, Codex, Git, and Mayphus as one practical workflow.
My daughter's grandma called while I was outside the house because she could not find the Roku remote and wanted me to stop a show. The Roku app refused to enable its remote over mobile data, even with Tailscale connected. ChatGPT suggested connecting my phone to any available Wi-Fi, then manually entering the Roku's home IP address. The Wi-Fi satisfied the app's check, manual entry bypassed discovery, and Tailscale carried the connection home. The official Roku remote then worked while I was outside.
I’m starting to optimize my Codex workflow. Instead of making every turn follow a heavy default sequence of pull, change, test, and review—even when a small change takes only a moment—I’m moving toward a hierarchy of iteration, review, and ship. The depth of the workflow should match the size and risk of the change, so a small task does not automatically take two or three minutes. I’m also starting to use Git worktrees so independent work can run in parallel.
A part-time job might be a good fit for me. In theory, full-time employment is not always efficient for either the employer or the employee: people rarely work productively through every hour of a fixed schedule, while focused work often comes in intense bursts that can stretch beyond eight hours when the work is truly moving. Part-time work may leave more room to follow that natural rhythm.
Coding, architecture, and other high-level work can be important, but they can also abstract the problem until I lose sight of what actually needs to be solved. I want to begin with the concrete problem, make the smallest useful thing, and only add abstraction when the work asks for it.
Today I worked on my website while taking care of my daughter. Instead of separating work and life, AI let me fit small moments of engineering into ordinary parenting. It’s only the first day, but it feels like a different way to work.
I am trying to develop a workflow that lets me work mostly on the go—with mobile devices, ChatGPT, AI, and Codex—while still being present and taking care of my daughter. With AI agents, I only need to prompt a few times; most of the work is waiting and thinking. A traditional desktop workflow often keeps me sitting at the computer even when I am mostly searching the web, browsing X, or watching YouTube. The better optimization is to let the agents work while I step away: think clearly, enjoy life, and care for my baby, then return when a decision is needed.
ChatGPT desktop is starting to feel like the AI version of Emacs: one extensible environment with a terminal, browser, code, files, and conversations together. The important part is not having every tool in one window. It is that my operations and working state can become shared AI context, so I spend less time switching apps and explaining where the work stopped.
ChatGPT desktop quietly added passkey/password support to its in-app browser — something I’ve wanted for weeks. This matters because I sign in to Google with a passkey. Now the in-app browser can handle it. One less reason to open Chrome, haha.
Short ideas should have one quiet public home. They do not each need a new page, project, or polished conclusion. A dated thread can show the thought while it is still moving.
A personal site is more useful when it shows working artifacts and operating evidence instead of describing ambitions. Build the proof, then write the sentence.
chatgpt, work, codex may share ui. only difference is the context (project, plugins etc)
I used ChatGPT Atlas as my main browser and loved it, even though it’s no longer maintained. But ChatGPT desktop doesn’t need to replace Chrome. My workflow now: Chrome for the web, ChatGPT desktop for intelligence, and the ChatGPT extension connecting them. It feels natural.
Some people dislike GPT-5.6 and the “super app.” I think they’re amazing. Real engineers and users don’t spend all day judging models or comparing flaws—they use the tools to build something amazing. Extraordinary capability at a low cost, shared globally, is worth appreciating.
The smallest useful project note records where the work stopped, the next physical action, and the evidence already captured. That is enough to make tomorrow easier.
codex in-app browser needs passkey support @thsottiaux
Today I'm reorganizing my infrastructure page. I realized that my website should explain my Kubernetes cluster, Podman containers, KVM virtual machines, and network visually instead of hiding everything behind text. This livestream is me refining both the website and the idea.
A YouTube livestream or video.
Today's realization: I'm not at my best when answering from memory. I'm at my best when investigating. Give me a terminal, logs, and a problem. I'll figure it out. Learn. Repeat. That's how I'm going to learn and build in public.
My Roku Remote voice feature don't works weeks ago, and today all buttons don't works. So I decide to teardown and see what is inside. After teardown and try again. it works. I think it's battery case has issue? And voice feature continue don't works. I don't see any water or possible visual broken. So wait for next time to investigate.
Forget the superapp. Just let ChatGPT and Codex share memory. Re-explaining the same project to both is unnecessary friction.
A YouTube Short.
Just migrated my codebase to my mini PC. Maintained by Codex now, so I barely need my desktop PC anymore. I can work from coffee shops, supermarkets, etc.
My Roku remote can’t listen to me. investigating if microphone break down.
I don’t remember the commands anymore. I remember the curiosity. Ten years later, most of the software is obsolete, but the habit of taking things apart, experimenting, and documenting what I learn is still the most valuable skill.
Repair note for a toddler's electronic learning book: inspect the power path, clean the battery compartment, replace batteries, and verify the fix.
This children’s electronic learning book stopped working.
@naval Humans should do the work, AI should do the documenting.
After reinstalling the Codex mobile app and pairing it again, sync issues disappeared. The app also feels smoother, though I'm not sure if that's because of the reinstall or just my perception.
Fixed a sync issue between Codex Desktop and the mobile app. The solution was to uninstall the Codex app on my phone, then pair it again.
I just realized I first tried ROS almost 10 years ago. This photo was taken in an old, almost abandoned office at my university. July 2017.
A YouTube Short.
A YouTube Short.
Found these old photos today. My Raspberry Pi 3 used to be the brain of my ROS robot. I lost this Pi after buying a Raspberry Pi 4, but I still have the other boards in the box. Maybe it’s time to build something with them again.
A YouTube Short.
A YouTube Short.
Curiosity teardown note for three TP-Link security cameras: power-on observations, reset and app behavior, internal modules, UART wiring, and hardware design clues.
I bought this broken TP-Link security camera for about $1 to learn how it is built.
00:00 Broken TP-Link Camera Investigation Begins 00:43 Connecting UART Debug Interface 05:04 Using AI as a Hardware Repair Assistant 07:04 No Serial Output After Power-On 10:18 Checking Power Rails 15:49 Suspecting a Power Circuit Fault 18:07 Testing the USB-TTL Adapter 20:11 UART Loopback Verification 23:07 Verifying 3.3V Logic Levels 27:21 Reconnecting UART Lines 28:01 Second Boot Attempt 32:39 Searching for Camera Hardware Information 34:00 Why the Camera Probably Isn't Booting 35:03 Wrap-Up & Future Investigation
Investigating 3 broken TP-Link security cameras bought for about $1 each.
building my ai first personal website. still ugly.
A YouTube Short.
Hardware teardown note from the Nano Pi stream: open the enclosure, remove the SD card, inspect RK3328, RAM, Ethernet chips, USB-C, reset, debug pins, and power-management…
A YouTube livestream.
A YouTube Short.
A YouTube Short.
First Live explain and test.
my lithium battery burn by itself.
A generated Flypy double-pinyin skin. Instead of drawing keyboard themes manually, I'm generating them from code so the same system can produce skins, previews, and configuration packages automatically. mayphus.org/typing
A YouTube Short.
A short operations note on recovering a router service through an independent Tailscale jump host when direct access is gone.
Building
Rime Input Method and Yuanshu Input method's Schema and Skin generator. mayphus.org/typing/
I am building a single, unified Racket project to orchestrate my entire personal infrastructure.
This is a complete visual walkthrough of a massive matrix grid UI designed to browse every single character space across all 17 Unicode planes (Planes 0 to 16). https://mayphus.org/unicode/ do you like it?
I just used my Emacs to check my homelab computer k3s's service when I am out of home.
Screenshots from December 2025.
Screenshots from November 2025.
In September 2025, I contributed a small fix to weapp-vite, a Vite-based build tool for WeChat mini-programs. The merged patch made its build, dev, and generate commands honor custom configuration files through --config or -c, including resolving relative paths from the project root.
Screenshots from August 2025.
This is my Chinese breakfast.
Just powered through a 50 KM ride. 🚴♂️✨ Legs on fire, heart full—cycling goals unlocked. #CyclingLife #Endurance
Start 2025 new year journey.
This board follows the official RP2040 minimal hardware design from Raspberry Pi.
A YouTube Short.
I attempted to design a minimal PCB for the ESP32, but it seems to have some issues—possibly related to power delivery or the EN pin. However, I’m glad to see my soldering skills haven’t gotten rusty! I successfully soldered a USB-C connector using just a soldering iron and used a heat station for the first time to solder the ESP32 chip itself. The results were fantastic. #diy #sold #soldering #solderingiron #esp32
A YouTube Short.
Made my first ESP32. Fail to boot.
A YouTube Short.
Made a simple displayer. Used a Raspberry Pi Pico, and 1.69 in LCD. And made my first useful PCB board. #diy #pcb #3dprint
A YouTube Short.
A YouTube Short.
A YouTube Short.
@salimbenbouz cool
A YouTube video.
A YouTube Short.
Start from simple
A YouTube video.
A YouTube Short.
A YouTube Short.
My DIY page turner. I think I need make a bluetooth version. #DIY
Start live as a creator, maker.
3d print a Batman
my diet, - Yellowhead catfish - Common bean - Pepper steak
Happy Halloween!
migrated my website from #github pages to #cloudflare pages. make website style somewhat similar to #emacs
Screenshots from September 2024.
I enjoy write with Emacs. #emacs
Garamond font looks great. Just switch font from browser’s default serif font to it.
Working on my uniorg powered #astrojs website. Website keeps minimal and been treated as an index list of my #org-mode notes.
Switch from a content consumer to a content creator. 1. Build a #orgmode based website.
These containers should have gone to many corners of the world. Then they must be more knowledgable than me.
is this mediterranean food?
FreeBSD is my favorite. I never need to choose between different Linux distributions and package managers. And in FreeBSD both basic freebsd system and third party packages are easy to find the sources. Make the system ‘transparency’.
Excited to share my first open source contribution! Just submitted a patch to FreeBSD to fix documentation for thin nullfs jails. #OpenSource #FreeBSD #FirstContribution
A YouTube Short.
A YouTube Short.
In this video, I'm desoldering components from my first self-designed PCB - a basic RP2040 minimal design. Although the design didn't work as intended, it marks an important milestone in my maker journey. Watch as I carefully remove components from this failed PCB, demonstrating proper desoldering techniques.
My first PCB.
A YouTube Short.
A YouTube Short.
A YouTube Short.
A YouTube Short.
Thanks, Hugging Face's Zero GPU. #zerogpu #huggingface
@ubuntu try the new autoinstall install method. i still remember how i hack ubuntu server’s iso and modify autoinstall years ago.
Do something.
Start to practice free talk. This time with a more freely style, the main goal is output something useful content.
@ollama Awesome, saved my old desktop.
@miiura @vercel Cool! simple, minimal and powerful.
Oh my poor machine.
Waiting for GPT-4 Turbo available on my account.
A minimalism diet.
A YouTube Short.
Making Century Egg. Wait for the magic of time. #chinese
A YouTube Short.
Start running again. This time mainly focus on health, and habit building. Ignore speed, ignore marathon.
voice input is so great way to input text messages. it also practices the English speaking skills.
Use tweet to catch ideas.
I love Emacs org-mode to take notes. denote is my favorite tools, because I can use it start taking notes from minimal. Not like org-roam. #emacs #orgmode
A YouTube Short.
A YouTube Short.
A month ago, I upload a practicing English short video to Chinese instagram 小红书. After that I update daily. Why I do this because I have social anxiety and lack up speaking communication skills. One month of practice improve m a lot I think. I will continue to do that.
Hi, My name is Felix Tang. I am learning and practicing English as a non-English speaker. I also want to share my mind and skills with this way. I also upload daily practice videos to my YouTube channel. I help these methods can save my English.
Planned a 65.5-kilometer walk from Hangzhou to Zhuji. I made it about 20 or 30 kilometers.
Early Bing Chat and ChatGPT experiments: connecting EdgeGPT to Emacs, then relaying messages between the two chatbots.
Installing GNU Guix System on a workstation.
Building my ROS robot vehicle. After learning with turtlesim, I started putting the physical platform together: a wheeled chassis, stacked controller and motor-driver boards, power wiring, and a cooling fan, all still exposed while I debugged it beside the laptop.
At university, I helped others get this linear actuator moving. The work brought the whole control path onto one desk: code on the laptop and monitor, a power supply, controller boards, a breadboard, wiring, and the actuator itself.
My first time using an AT89X51 microcontroller. I built and debugged the circuit on breadboards, surrounded by jumper wires, handwritten notes, a programmer, and a multimeter. The LEDs, seven-segment display, and ultrasonic sensors made the code visible in the physical world.
An evolving proposal for websites built around AI-readable investigation objects, with the human interface as one view.
FriendlyElec and NanoPi Alpine image builder web UI for rootfs and SD-card builds.
Printable and generated learning-card experiments.
Cell-structure and DNA-strip SVG generators.
Interactive calculus figures, exports, and TypeScript widgets.
Visual exploration of Cangjie character structures.
A restored generative field of hand-drawn cellular forms, preserved for screen and print.
Chemical formula parsing, element counts, tests, and a web interface.
Text preparation and traditional vertical book rendering, grounded in the Honglou experiments.
Generated references for the zodiac, eight trigrams, and the twenty-four solar terms.
Chinese zodiac generation and reference experiments.
Printable atlas of electronic circuit symbols.
The useful parts of two commerce prototypes: storefront, cart, checkout, administration, Cloudflare, and WeChat delivery.
Public browser-local daily tracker.
Inbox of small runnable demonstrations recovered from the archive.
Food, nutrition, and personal diet experiments; only a sanitized shell is public.
Declarative drone choreography, simulation, safety checks, and show export.
Bagua and trigram reference generation.
Racket pict emoji and icon prototypes.
Rotating Oxford 3000 word wallpaper built with Racket.
Declarative architectural lighting, mapping, preview, and controller export.
Football tournament data and document-rendering experiments.
Generated flashcard interface and source model.
Generated daisy, cosmos, lily, lotus, rose, sunflower, and tulip studies.
A short personal introduction to FreeBSD: what it is, why I used it, and what stayed with me.
Mahjong, Sudoku, Tetris, and football tournament experiments preserved as one playable-software family.
Generated deep-time visual study.
Printable guide-sheet generator.
Traditional Hanzi matrix with pinyin, zhuyin, and English details.
Mechanical CAD, circuit-symbol references, and four original KiCad board experiments.
Public-safe graph view of the Mayphus home LAN topology.
Dream of the Red Chamber source and text experiments.
Declared icon scenes with rendering, licensing, and source attribution.
Public, sanitized description of the infrastructure behind Mayphus.
Rime and Yuanshu input-method builder with live keyboard previews and downloadable packages.
Shared timeline for lights, fountains, drones, music, and performance channels.
Projectile-flight visualization with quadratic drag.
Clickable phonetic reference with local audio samples.
Personal job-search tooling; this public page excludes private applications and resume data.
Small language and stack-language experiments.
Racket-first CAD for mechanical ideas, shapes, parts, and manufacturable models.
Small teaching surfaces for words, cards, prompts, and device-sized repetition.
Robot-learning and training experiments.
Visual map of a GPT-partitioned Linux install disk: EFI, boot, root, swap, home, and backup GPT metadata.
Racket-powered local four-player Mahjong prototype.
Interactive calculus, numerical and 3D studies, Voronoi geometry, cube mathematics, and comparison models.
Animated primary-tooth eruption timeline for all 20 milk teeth.
A restored Racket drawing DSL explored through geometric and decorative motifs.
Declarative water, pump, color, and music-synchronized choreography.
Describe places to find in any city, then generate a safe, bounded OpenStreetMap layer on a real basemap.
Package-management and installation workflow experiments.
Small macOS people-counting utility experiment.
One timeline language for coordinated lights, fountains, drones, music, and show control.
Square-cell periodic table with search, category focus, and click-to-inspect element details.
Chinese regnal-era conversion and a generated timeline spanning 505 historical eras.
Experienced across CI/CD platforms, production operations, Linux systems, networking, and embedded hardware.
Numeric visualizations, surfaces, meshes, curves, and interactive 3D.
A working record from camera-guided MuJoCo learning to agent-assisted physical SO-101 setup, calibration, and fault isolation.
Robot training, ROS 2 environments, and native computer-vision experiments gathered into one practical lab.
Chemistry, biology diagrams, projectile physics, and the periodic table as one explainable science workspace.
Plain-text AI coding tool and Emacs bridge experiments.
Archived Cloudflare shop prototype with route and binding declarations.
Twenty-four solar terms visualization.
Guided macOS SD and USB image flasher CLI with standalone Ubuntu autoinstall support.
Public site inventory and coarse runtime health.
Interactive TCP map for connection setup, segment headers, sequence numbers, acknowledgments, windows, retransmission, and teardown.
A compact Racket Tetris game.
Regnal chronology, Unix lineage, world history, deep time, and human development across different scales.
Android Rime theme, schema, and configuration collection.
Wide Unicode character browser and inspector.
A maintained Unix lineage chart extended through 2026 with refreshable release labels.
Traditional vertical Chinese book-layout DSL and PDF renderer.
Generative motifs, flowers, cells, emoji, pixel art, and icon systems in one drawing collection.
WeChat commerce app with mini program, CloudBase functions, and admin console.
Large generated world-history chart for screen and print.
Recovered public-writing source and publishing domain.