Projects Tracker — everything currently open

This folder is the map of what I am working on. Five tracks, one page each, and this page is the way in. Nothing here is a finished document — every page is meant to be edited in place while the work moves. The tick boxes are live: click one in the browser and the .md file on disk is updated.

In one line. One page per track, and each track page owns its own next steps — so there is exactly one place to look, and exactly one place to update. Open the track you are working on.

Full text of every note in this folder and its subfolders — try render, Tailscale, profile, narration. Press / anywhere on the page to jump here, Enter to open the top hit, Esc to clear. Each result links to the section it was found in.

The five tracks

1 · AI / ML learning

Andrew Ng's two specializations, written up as they go. What exists, what is only watched, the size of every note, and the exact video to resume from. The ML Specialization is done; Deep Learning Course 2 is the live one.

2 · ReMapping

The film-rendering system: productions are folders, films are Python, and nothing ships until it verifies against an output profile. Three implementation requests are open.

3 · Smart Devices

The home automation running on the Pi — what is live, on which port, under which service — plus the protocol stack still to study: Matter, Thread, Zigbee, MQTT, BLE.

4 · 5G Systems

One track, four layers: the NR physical layer, the RF front end underneath it, the Open RAN split above it, and where machine learning actually enters. Writing has started.

5 · Pi 5 home cloud + 4 TB NVMe

The always-on home server: parts, NVMe over PCIe, network limits, power, cost, security, and the final build. Parts not bought yet.

The board

#	Track	State	Where it stands
1	AI / ML learning	in progress	ML Specialization written up in full. Deep Learning Course 2 watched, not written — see its next step.
2	ReMapping	active build	System runs; three implementation requests open.
3	Smart Devices	in progress	One device live on the Pi. Protocol stack still to study.
4	5G Systems	writing	Layer 1 begun — the frame structure is written. Next: SRS.
5	Pi 5 home cloud	planned	Waiting on parts — see the budget.

Table 1: Every open track, its state, and where the work sits. The next steps themselves live on the track pages as live tick boxes - this table points, it does not repeat.

Moving now

The goal — phase 0, to 6 Sep 2026
AI / ML — write DL Course 2 Week 1
ReMapping — three requests open
Smart Devices — protocol study

Waiting on something

Pi 5 home cloud — parts not bought
AI-RAN — reading not started
ReMapping manual — EN + FA rewrite

How the tracks connect

The tracks are not independent. The Pi 5 home cloud is the machine the smart-device services and the note engine already run on, and it is where ReMapping renders and 5G experiments would eventually be stored. The AI / ML learning track feeds the ML notes, which is what the top layer of 5G Systems draws on.

Figure 1: How the five open threads relate. The home server is shared infrastructure, not a separate project — everything above it eventually lands on it.

The notes these tracks produce

The tracks are the plan; the notes are the output, and they live in their own folders beside this one. Those folders report their own state back into the track pages (§4 explains the same mechanism for the archive) — but they are also just notes, and these links open them:

Machine Learning Notes

13 notes across four tracks — linear regression through to deep Q-learning, plus the network reference. The written half of track 1.

5G Systems Notes

The NR physical layer, the RF front end, Open RAN and AI-RAN, written from the 3GPP and O-RAN specifications. The written half of track 4.

Smart Devices

One folder per device documenting what actually runs on the Pi — the data path, the API, the traps. The written half of track 3.

ReMapping manuals

The rendering guide, the open implementation requests and the production log. Inside this folder, so it opens however the tracker is served.

These links work however the tracker is served. They point one level up, outside this folder — so if only the tracker is being served, the engine starts a server for the target folder on mdopen's port and redirects you there. Click, and it opens. Serving md-engine/ itself skips the redirect and adds one search box across every folder.

The course archive

md-engine/Courses (RAW)/ is shared raw material, the same way the Pi is shared hardware: lecture videos, labs, slide decks and hand-written LaTeX notes, feeding more than one track. It reports its own contents — the table is walked from disk each time this page loads, so it cannot fall out of date, and a folder nobody has declared shows up in red.

It is inert to the engine — no .md, nothing served, never searched. But it is the reason never to mdpush or mdexport the md-engine root: push and export the note folders individually, as the commands below do.

Working conventions

One page per track. Every track owns a page in this folder, and its next steps live there and nowhere else — as live tick boxes. This page points at them; it never repeats them. A track that grows a body of work gets a sub-folder (as ReMapping has) rather than a second page here.

A goal is not a track. A track is ongoing work on a subject with no end date. If a dated target ever appears — a job, a deadline, an exam — it gets a page of its own and borrows from tracks; it never owns one, and it is removed when it is met or abandoned. The subject survives the goal.

Tick boxes are live. Any - [ ] is a real checkbox in the browser, and ticking it rewrites the source file. Use them for anything with a definite done state; use prose for anything that needs a judgement.

Badges mark state, not priority. done active in progress blocked parked planned

Dates are absolute. Write 16 Aug 2026, never "last Tuesday" — these notes outlive the week they were written in.

Some tables are live, not typed. Content folders elsewhere in the engine report themselves here through widgets, so their numbers cannot go stale — the archive table above, the device table on Smart Devices, the note table on AI / ML learning. Do not hand-edit what a widget prints; fix the folder's status.json instead.

If a track goes quiet for a month, mark it parked rather than leaving it looking active.

Running this folder

Either of these works:

# the whole engine - every folder in one server, one search box across all of them
mddash ~/Documents/md-engine

# or just the tracker - links to the notes folders still work, via a redirect
mddash "~/Documents/md-engine/0_My Projects Tracker"

Serving the root opens the tracker at /md/0_My Projects Tracker/index.md and needs no redirects; it is also how the Pi serves everything, one library root with many folders inside. Serving this folder alone is lighter, and the links in §3 start the other folder's server on demand.

mdpush ~/Documents/md-engine    # copy it all to the Pi over Tailscale

Edits show up on the next reload — the page hot-reloads while a server is running.