Listening First

In August we said September was the month LuminousPath starts learning from students instead of from us. So far that student is Misty. She is taking classes this semester and uses LuminousPath for her own coursework, and most of what we changed in September came from problems she ran into doing real assignments. The LuminousPath beta is now open.

The Hardest Thing We've Built: The Brain Dump

Misty begins coaching by giving students space to get assignments, errands, and worries out of their heads before trying to organize them. In September, we built that routine into LuminousPath.

It took almost three weeks, longer than anything else we have built. Three things made it hard.

It runs on a timer. The brain dump is timed, and the timer has to keep its place when a student pauses, asks for help, or leaves the page and comes back.

The AI doesn't see it until the student decides to share it. What a student writes stays out of the AI's view until the student chooses what to share. Building that boundary, and testing that it holds, took more work than any screen.

Coaching had to change. The coach has to know when the brain dump starts, wait while the student works, and pick back up when they're done, without asking to see what the student wrote. That meant changing how coaching itself works, not just adding a new screen.

Our first run-through revealed rough edges, including a coach that went too quiet between steps. We're refining the experience with Misty's review before introducing it to students.

Fixes That Came From Real Use

Most of September's other work started with Misty using LuminousPath for her classes and telling us what went wrong.

Time that means something. Time on LuminousPath used to count how long a session sat open. One test account showed 48 hours for about an hour of real work. Session time now counts engaged time only, the minutes between messages, and long gaps don't count.

Diagrams you can read. A tutor's history diagram rendered as a long, thin strip with labels too small to read. Diagram text now never drops below a readable size, any diagram can be enlarged and zoomed, and diagram labels follow the student's chosen reading font and text size. The tutor also keeps its diagrams small: a handful of boxes and short labels.

How we test. We added test students that the platform creates and runs on its own, so a new feature gets checked end to end without a person clicking through every step. The goal is simple: new work should never quietly break something that already worked.

A note on where this stands: Misty's expertise shapes how LuminousPath teaches today, and the full, formal work of building her methods into the platform is still ahead. In our standard wording, the current platform contains preliminary, concept-level teaching logic informed by Misty Hogg's pre-existing expertise. Full formal methodology encoding has not yet occurred.

Understanding the material does not make organizing and finishing the work easy.

Our First Month in Local Lift

September was our first full month in the United Way of Tarrant County Local Lift cohort, eight finalists working through the ThinkLab curriculum from TechFW. Our thanks to Jeffrey Casanova of TechFW and SCORE mentors Bob Siffermann and Malcolm Platt, who led our September sessions.

The prework asked us to name our customers and our value in plain words. Our answer stayed narrow on purpose. LuminousPath is for neurodivergent college students, and it is not trying to be a general study tool. The students we build for often get stuck on planning and follow-through, and a tool built for everyone is rarely built for that.

Next up is filming our pitch videos in early October. The final pitch is set for November 12.

Heading to Nashville

On October 14, Anthony joins the panel AI in the Gaps: Extending Human Support at the NeuroTech Frontiers Summit at Vanderbilt University, hosted by the Frist Center for Autism and Innovation and the JANUS Innovation Hub.

He brings a pattern we keep seeing. Many neurodivergent college students can understand the material and still struggle with the structure around the work: breaking an assignment into steps, holding the plan, knowing what to do next. Anthony spent more than fifteen years working on that same bottleneck in homelessness systems, where people often had the will and the ability and the system around them kept breaking down. The brain dump is one answer to it.