A map of your own week
Your recurring work, sorted on paper into three columns: what an agent could carry now, what still needs your judgment, and what should never be touched at all.
A practical class for New Braunfels residents, owners, operators, and professionals
Hand off the
right work.
Keep the judgment.
In two hours, learn how today’s AI apps, models, tools, and agents fit together. Leave able to identify a useful first workflow, protect judgment and private data, and test what the system actually did.
The next date, capacity, and price are TBA. The planned venue is Cowboys & Cadillacs in downtown New Braunfels. Joining the interest list takes no payment, reserves no seat, and commits you to nothing.
Free to joinInstant email confirmationLocal, in person
- Next class
- Date TBAJoin the interest list for updates if a session is announced
- Length
- Two hoursPlain language, demonstrations, and a live build
- Format
- In personPlanned downtown venue; final room pending
- Cost
- TBAInterest form is not a payment or reservation
Why now
The hard part was never
the software.
It’s knowing what
to trust it with.
You are not short on AI tools. You are short on a way to decide. A polished demo can hide weak scoping, missing review, and work nobody can audit.
-
What if an agent emails the wrong client?
A workflow can follow its instructions and still act on the wrong record. A person then owns the apology.
-
What if a workflow breaks in silence?
A vendor can move a button without producing a useful alert. The first visible sign may be missing work.
-
What if access is never scoped?
Handing a tool the keys to everything may make setup easier, but the blast radius outlives the convenience.
-
What if nobody notices it stopped?
Unsent follow-ups can look like disinterest unless monitoring makes the gap visible.
-
What if nobody trusts it unattended?
If a person must constantly re-check every step, the automation may have created a second job.
These are prepared failure patterns, not client case studies.
Each one starts with a scoping decision.
What you leave with
You leave with decisions,
not inspiration.
A clear-eyed read on what’s real
What agents genuinely do well at your size today, with examples from businesses like yours — and the exact point where the impressive demos quietly stop working.
The five questions to ask first
What to ask any vendor — us included — before you spend a dollar or a weekend. Enough to tell a careful build from an expensive gamble.
You take the paper home: the Map Your Week worksheet, the Doing It Right checklist, and the five questions card. Nothing else. It should stay short enough to actually get referred to.
One synthetic job, start to finish
Watch an agent
work a Friday.
Not a chatbot answering a question. This deterministic educational example uses fictional roofing leads to show a bounded follow-up workflow — every simulated step, everything it refuses to touch, and the moment it hands the work back to a person. It is not a client result or deployed service.
Sample run: Friday, 4:00 pm
Wakes up on a schedule
- L‑01Roof inspection after the April hailWaiting
- L‑02Gutter replacement quoteWaiting
- L‑03Storm damage, insurance pendingWaiting
- L‑04Second opinion on a leakWaiting
- L‑05Re-roof, asked about financingWaiting
- L‑06Skylight, no phone number givenWaiting
Synthetic draft 3 of 17 — fictional lead record
Hi there, You asked us to look at the roof after the hail came through on April 2nd, and we never got you on the calendar. That one is on us. If it is still on your list, I can have someone out Tuesday or Thursday morning. It takes about forty minutes and there is no charge for the look. If you have already had it handled, just say so and I will close it out. You will not hear from us again about it. — Sample Roofing Co.
Generated from a fictional fixture created for this deterministic walkthrough. No customer record or real business is represented.
It runs on a timer at 4:00 pm every Friday. It cannot decide to run itself, and it cannot run early.
The power is in the second column.
In this synthetic example, the tool prepares seventeen drafts. The safety lesson is everything it was never given: no send permission, no ability to edit a record, no guessing when two leads share a phone number.
Scoping that second column is the work. It is most of what the two hours are about, and it is the part almost nobody does before they wire something into their business.
Register interestWhat we make clear
From AI vocabulary
to a useful first build.
No hype glossary. We use plain language, small experiments, and one live workflow so you can tell what a tool is, what it can access, and where your judgment stays in charge.
Companies, apps, models
Who makes the model, where the app fits, and why switching models can change cost, speed, privacy, and capability.
Tokens & context
Why a model can sound confident and still miss the important detail—and how to give it the right amount of context.
Tools, plugins & MCP
What it means to connect a calendar, file system, or business tool, and how permissions turn a chat into an action.
Cloud, local, chatbot, agent
A practical map of tradeoffs: hosted or on your machine, answers or actions, automation or approval.
Discernment in plain language
Pause before sharing sensitive information. Check the source, the scope, the output, and the human who approves it.
Live vibe-coding build
The room chooses from prepared, bounded options. We use synthetic data, define acceptance criteria, build, test, fix a defect, and verify the result.
Takeaway: you leave able to explain the system to a teammate, choose a sensible first experiment, and know exactly when to stop and review.
The session
Two hours.
Built for participation.
The format is designed for discussion, plain-language demonstrations, and a whiteboard that fills with anonymized workflow shapes from the room. Final capacity and room setup remain TBA.
- Planned venue
- Cowboys & Cadillacs
125 N Castell Ave - Capacity
- TBA
- Format
- Demonstrations, exercises, and a bounded live build
- 0:00
Why you’re here
Everyone names their business and the one task they would hand off tomorrow. It goes on the whiteboard, and that list becomes the spine of the whole session.
- 0:10
What the pieces actually are
Companies, apps, models, tokens, context, and agents in plain language. We compare prepared examples and label what is observed versus assumed.
- 0:30
Where they pay off
Four categories, one concrete example each, sized for a business with one to twenty people. Then we sort the room’s whiteboard list into them, live.
- 0:45
Map a useful first workflow
Use the worksheet to identify one recurring task, its data, its risk, and the human approval point. Private customer information stays out of the exercise.
- 1:15
How this goes wrong
Five prepared failure patterns, followed by the checklist that reduces them: scoping, access boundaries, testing, monitoring, and a plan for the day the system breaks.
- 1:35
Build and verify something live
The room chooses from prepared, bounded options. We define acceptance criteria, build, run, find or fix a defect, and show the evidence rather than pretending a draft is finished.
- 1:50
What happens next
A recap, a written first step, and time for questions. Optional follow-up, if any, will be clearly separated from the class and described only after its terms are approved.
When & where
Next date
coming soon.
A first session date and room have not been confirmed. If registration opens, the approved details and contact process will be published with it.
-
Taking names
First session
Date and time will be announced
- Pending date
Planned venue
Cowboys & Cadillacs · 125 N Castell Ave
Who it’s for
A participatory room.
Deliberately.
The class is intended for people who want to think, test, and ask practical questions rather than passively watch. Final capacity is still TBA, so the interest form helps shape the room honestly.
Take a seat if you
- Own or run the business, and can act on what you decide
- Have recurring work that eats hours and doesn’t really need you
- Have heard enough about AI to be interested, and enough to be wary
- Would rather get the first one right than get it fast
- Are comfortable saying your actual problem out loud at a table of peers
Give your seat away if you
- Want to learn to build agents yourself — this is judgment, not a how-to
- Are looking for a product demo or a vendor pitch
- Need regulated-industry or enterprise compliance architecture
- Already run agents in production and want tuning advice
- Want someone to tell you AI will replace your staff. It won’t, and we won’t
Who’s running it
Practical judgment,
not AI theatre.
Robert teaches a careful approach to useful systems: define what the thing is for, what it may touch, who approves its actions, and how a person would know if it broke.
The class applies those questions to prepared examples and a bounded live build. It does not present an illustrative workflow as proof of a completed client engagement.
He also sells the thing being taught, so it is worth saying the quiet part out loud: the session is built to be useful whether or not you ever hire anyone. You leave with the worksheet, the checklist and the questions — including the ones that would let you say no to him.
No borrowed logos, no invented credentials. The work is at rblurrr.cloud — judge it yourself.
Practical details
Everything else,
plainly.
- What it is
- A practical two-hour, in-person AI class
- Where
- Planned at Cowboys & Cadillacs, 125 N Castell Ave; final room confirmation pending
- Capacity
- TBA
- Cost
- TBA; the interest form does not reserve a paid seat
- Next date
- Date and time TBA
- Bring
- One recurring task, and who owns it today
- Cadence
- TBA
Questions people actually ask
Is this a booking or a payment?
An interest list, for now. The date and venue are still being confirmed, so nothing is charged and no seat is formally held. If registration opens, its process, availability, and contact plan will be stated then.
What will it cost?
Pricing has not been approved yet. Sharing interest is free and does not create a payment, reservation, or commitment. Confirmed details will be sent before any registration decision.
Do I need to be technical?
No prior coding skill or personal setup is required. The instructor performs one bounded live build, and every technical term is explained in plain language.
Will we build an agent at the table?
There is one short live build selected from prepared, bounded options. It uses synthetic data, explicit acceptance criteria, and no external side effects. You will see it run, fail an edge case, get fixed, and pass a final check; it is a teaching artifact, not a deployed business system.
Will an agent send anything on my behalf?
Not in the class, and not in the demo on this page. We start at drafts, boundaries and explicit human approval. Anything that can act without a person signing off is a later decision, made on purpose.
What should I bring, and what should I leave at home?
Bring one recurring task and a rough sense of who handles it today. Leave confidential customer data at home — we work from examples and shapes, never from your records.
Is this a sales pitch?
The class is designed to teach practical judgment and produce a useful next step. Any optional follow-up offer will be clearly separated and described only after its terms are approved.
My business is tiny. Is it still worth it?
That is the target. Every example is sized for one to twenty people. Small businesses tend to get more out of a first agent than large ones, because the recurring work is concentrated in fewer hands — usually yours.
Register interest
Interested?
Tell us what would help.
A few required fields are enough to join the interest list. You’ll get an immediate professional confirmation email, and Robert will personally review what you share.
- No payment taken today
- No sales call unless you ask for one
- Instant confirmation sent to your email