The learning path
One live hour. Monday, Wednesday, Friday. Trained to standard, not to time.
We teach military members, veterans, and their families in Killeen, Texas. Progress means demonstrating a skill — a GO — not surviving a deadline. A NO-GO means we retrain it. “Train to standard, not to time” is borrowed from military training on purpose: the calendar bends, the standard does not, and no one gets left behind.
Inside one hour
Our default session plan. We publish a real session log once cohorts run.
Human instruction, AI-tailored practice
A person teaches every class and sets the standard. Alongside the instructor, an AI tutor adapts each cohort’s lesson plan to each participant — the pace, the examples, and the difficulty of the practice change with the learner, which is one-to-one attention no single teacher can give a whole room at once.
Everything you produce is checked and returned with specific feedback: homework, tests, task briefs, documented workflows, the skills and agents you build. Feedback names what worked, what did not, and the fix to try next. The instructor reviews the record and makes the GO / NO-GO call on every skill.
Learner work stays private to the learner and the instructor. What goes into any AI tool follows the red lines you write in week two.
Foundations syllabus — six weeks
| Week | Focus | Artifact you keep |
|---|---|---|
| W1 | What AI can and can’t do | A task attempted with and without AI, with notes |
| W2 | Privacy and safe inputs | Personal red-lines checklist |
| W3 | Asking better: briefs, not prompts | A reusable task brief |
| W4 | Checking answers: sources and failure patterns | Source-check worksheet applied to a real answer |
| W5 | Repeatable workflows | A documented everyday workflow |
| W6 | Readiness week: demonstrate, repeat, or extend | Private demonstration record |
The readiness standard
“Explain it in your own words. Do it on a fresh example. Handle one variation. Privately, untimed, repeated until solid — a NO-GO means we retrain, not that you failed.”
Train to standard, not to time: the calendar bends, the standard does not. Demonstrations are private and untimed; a NO-GO schedules a retrain, and nothing about a retrain is recorded against you. Crawl, walk, run.
After foundations: the core path
-
AI foundations and safe use
The six-week syllabus above, paced to the cohort. Use an AI tool for a real task, learn privacy basics, source checking, and when a person should make the decision.
-
Prompting, research, and verification
Turn a vague request into a clear brief, compare outputs, trace sources, and revise based on evidence.
-
Reusable assistants, skills, and current plugins
Organize instructions, examples, tests, knowledge, and tool access. Learn how today’s publishing routes differ from older GPT and plugin formats.
-
Career project and portfolio clinic
Map job descriptions to evidence, write a truthful case study, explain tradeoffs, and practice an interview based on work you actually completed.
What ends up in a portfolio
- Task briefs that produced checked results
- Workflows documented well enough for someone else to run
- One documented skill or small assistant, with test cases
- One truthful case study: task, approach, tradeoffs, results
Skills to job tasks
| Skill area | Supports applications for |
|---|---|
| Prompting & verification | Operations, administrative, and research-support roles |
| Evaluation & replay | QA and testing-support roles |
| APIs & workflows | Junior automation work |
| Documentation & briefs | Knowledge and office roles |
Optional advanced paths
-
APIs and connected workflows
Structured data, test credentials, error handling, rate limits, cost controls, and safe connections between tools.
-
Agents, tools, and retrieval
Bounded permissions, trusted knowledge, checkpoints, recovery, and human review for consequential actions.
-
Evaluation, replay, and reliable improvement
Baselines, held-out cases, point-in-time replay, leakage, contamination, reward hacking, overfitting, and honest comparison.
We teach replay, leakage, contamination, reward hacking, and backtesting so learners can detect and prevent them. Every learner signs the same commitment in the community agreement: evaluation methods find weaknesses — they never manipulate scores or deceive people.
A learner pursuing every optional path would plan for roughly 54–74 instructional weeks and 162–222 live hours, often longer. You do not need every track to benefit.
Status: interest list open; first cohort dates not set — see the ledger (checked 2026-09-16).