this Wednesdayoptionsthe week challenge ↗
Summer Teachers Institute · NM Tech · Socorro

Lessons Learned (and created).

Before we build a simtable, feel the power and control of the device already in your hand. Everything below runs locally, and you turn on each sense yourself. Pick anything — they're rungs of one ladder: one device → one algorithm → the whole room.

🔒 nothing leaves your phone unless you send it
the main bead ai-edu · 9a528020-a078-4dc3-a81c-75227494ec91 Every lesson here was built by an agent working in a bead — a GUID-addressed folder holding the source, the prompts, the decisions and the chat log. This one owns the method. Open it, read how these pages were made, and fork the process for your own classroom.

Pick a thing to open

one device · one algorithm

Laser tracking

Track a laser dot off your webcam through three algorithms — brightest pixel, color mask, centroid — where each one's failure chooses the next. It opens on the prompt that built it: how to turn a student's question into a learning path.

Open the laser lab ▸
your device · your data

Phone exploration

A wall of small prototypes: photos, your hand and face as controllers, location, your sensors. Real camera, nothing uploaded — you grant each sense one at a time, and only then do we sense together.

Open the prototypes ▸
many devices · one model

Understanding SimTable

A model everyone can stand around: a live wildfire simulation projected onto real terrain. Scan the code and your phone joins the shared table — the supercomputing session.

Open it & join the table ▸
the code you already scanned

How a QR code works

Type a string and watch it re-encode live: version, masks, and the Reed–Solomon parity that lets you punch a hole in it and still scan it. Find the cliff with your fingers — and learn why scattered damage is far worse than a blot.

Break one ▸
motion · a field your turtles can read

Optical flow

How does a computer know something moved? Frame difference says where but never which way. Chase that flaw down to the aperture problem — then turn the motion into a patch grid of flow-x/flow-y and let your turtles ride it.

Wave at it ▸
light vs ink · Melrose, NM

The colors of Melrose

Mix light (RGB) and mix ink (CMYK) and watch them disagree. Light adds toward white; ink subtracts toward mud — which is exactly why printers buy black. Palette drawn from a village of 622 at 4,406 feet that once held the smallest federal art center in the country.

Mix some color ▸
light as a ruler

The SimTable simulator

Watch a table get measured by light. Press scan: a projector throws gray-code stripes, two cameras read them back, and the sand's height is reconstructed — then swipe your reconstruction against the real elevation. The same trick a Kinect or Face ID pulls.

Run a scan ▸ ⚠ Needs WebGPU — Chrome or Edge on a laptop. Not iPad/Safari (it will just show black).
two eyes · one point

Feel the constraint

Why does stereo vision work at all? Two cameras see one point, and the match is trapped on a line — it cannot be anywhere else. Feel the constraint pull as you apply it.

Feel it ▸ Rigorous version: Camera Pose & Epipolar Geometry (live E & F matrices) + its companion notes.
NP-hard · everyone rode one

The school bus problem

Pick any of 195 real Albuquerque schools and watch buses find their routes — the multiple Traveling Salesman, which is really the Capacitated VRP your district actually solves. Five stages, each one's flaw choosing the next, every number measured live. The punchline: improving a route beats constructing a good one.

Route the buses ▸ Schools and roads are real OSM data. Student homes are simulated — and the page explains why they have to be.
you already know this syntax

ask debbie [ … ]

Every prompt that built this site — written in NetLogo. Because ask debbie [ build the page ] and ask turtles [ fd 1 ] are the same line, and you've been teaching the notation for directing agents since 1999. Includes the unedited prompts, typos and all.

Read the source of the week ▸ Not runnable NetLogo — and the page is colour-coded to show you exactly where NetLogo stops and the invented dialect starts.
where this actually goes

The jobs this is already

The jobs on a wildfire that are your students' projects, at incident scale — GISS, SITL, FBAN, LTAN, FEMO, BAER, ATGS, drone, PIO. FSPro's burn probability is "the fraction of the total simulations where fire arrives at a particular cell" — and its authors parallelized it across 16 or 32 processors because the runs are independent. That's this program's founding premise, in a journal, describing a job.

See the mapping ▸ The pure computer-vision job on a fire — reading imagery — asks for ICS-100 and IS-700. Two free courses. No fitness test.
for teachers · behind the scenes

The edu page

How this curriculum is actually built: the prompt, the affordances, the path — and the beads, domains, and pipeline behind it. The template you'd reuse for your own project.

Go behind the scenes ▸

The arc for the week

  1. Your phone is a server you own — serve a page, hold your own address.
  2. Many phones, one instrument — capture a session from every angle.
  3. Build your own simtable — a phone, a projector, any surface.
  4. The simulation on the table — the same kind of model you write in NetLogo this week, running on a surface, with people around it.