Full-stack product development
Python services, a custom web interface, persistent data, and native app surfaces, developed together around actual workflows.
Independent software · Thomas HolbirdAn evolving project
From information
to understanding.
A personal platform for learning, research, and putting ideas to work.
well.education connects study tools, source-linked reporting, and AI orchestration in one workspace. A project built through the work of using it, questioning it, and making it better.
Explore the projectlearn01
research02
connect03
build04
The idea01
Good tools make room
for better thinking.
The interesting part starts after you have the information.
You still have to make sense of it. Find the connection. Practice the difficult part. Decide what deserves your attention and turn a decision into work you can follow.
Originally called Brain, this project began as a way to coordinate AI agents across software projects. It has grown into a personal environment for studying, reading, researching, and building, with shared infrastructure underneath.
well.education is its public identity: a place to see the system, understand its design, and follow the thinking behind the work.
One platform.
Learning systems / Research workflows / Knowledge connections / Agent orchestration
Inside the workspace02
Built around the
way work actually happens.
Different tools. A common thread:
make the work easier to understand.
Learning systemsPractice with
Practice with
a reason to return.
Course guides, question banks, and review tools bring the material and the work of learning it into the same place. Learn mode gives immediate feedback; saved attempts make it possible to revisit where understanding broke down.
Implemented across LSAT preparation, Spanish, and university coursework, with concept-level review and authored study materials.
Explore the learning systemLearning workspaceIllustrative workflow
Reasoning & retention
Understanding takes practice.
01Read the materialGuides, concepts, and worked examples↗
02Put it to the testPractice with immediate feedback↗
03Return to the gapReview attempts and revisit concepts↺
A learning loop, with the history attached.
Research & publishingA wider view.
A wider view.
A traceable source.
News Unfiltered brings source-linked reporting into a deliberate reading experience. The pipeline separates collection, editorial preparation, and publication, with original sources kept close to the resulting summaries.
AI-assisted writing is subject to editorial checks. Reporting, interpretation, and uncertainty remain distinct so readers can inspect the basis of a claim.
Read News UnfilteredA public project within the platformEditorial approach
news
unfiltered.
unfiltered.
The world, read from
more than one side.
Sources attached.
Claims checked.Reporting separated
from interpretation.
Open the published edition ↗Claims checked.Reporting separated
from interpretation.
Knowledge & evidenceFollow the connection.
Follow the connection.
Keep the context.
The Neural workspace organizes entities, relationships, notes, and evidence into explorable maps. A relationship is more useful when you can see where it came from and return to the supporting material.
Graph views, entity profiles, and evidence records give research a structure that can be revisited as new information arrives.
Explore the knowledge architectureKnowledge workspaceIllustrative map
Relationships are useful when their origins are visible.
Agent orchestrationGive the work direction.
Give the work direction.
Keep it in view.
Projects define where work happens. Agents define how it runs. Tasks carry the instructions, and streamed events show what happened along the way.
The web control plane and local worker divide coordination from execution. Task records, live logs, cancellation, and recorded outcomes make long-running work inspectable.
Explore the orchestration modelTask lifecycleIllustrative workflow
A task is more than a prompt.
- 01Define the workProject · role · instruction
- 02Execute in contextWorker · tools · working directory
- 03Observe the runEvents · status · cancellation
- 04Review the resultOutput · outcome · persistent record
Under the surface03
Many tools.
One considered system.
The interface is only the visible part. Behind it are task queues, persistent state, model runtimes, and a clear separation between public work and private activity.
01 / EXPERIENCE
A shared workspace
Study · research · knowledge · projects
02 / COORDINATION
A persistent control plane
FastAPI · SQLite · task state · event streams
03 / EXECUTION
Workers & model runtimes
Project-aware execution, local inference,
and service integrations.
04 / RECORD
Work that leaves a trail
Attempts, notes, evidence, task outcomes,
and published artifacts.
The engineering behind it04
The details
are the work.
A project like this is an exercise in connecting systems and taking responsibility for what happens between them.
Read the project case studyAI systems integration
Agent execution, local model runtimes, streaming responses, and tool access connected to the surrounding application.
Evidence and evaluation
Source attribution, saved learning attempts, and inspectable run histories make the output something you can return to and question.
Operation beyond the prototype
Deployments, authentication, worker boundaries, and regression checks are part of the product’s design and ongoing maintenance.
Stay well.educated.
Explore the work. Follow the reasoning.
Start a conversation about what comes next.