For behavior analysts who want to stop watching AI from the sidelines. You chat with the AI about your articles, your data sheets, your folders of PDFs, and walk out knowing how to build a working tool yourself. No FERPA- or HIPAA-protected data for this workshop.
Cameras on during demos. Share in chat. Two instructors, thirty builders.
We are not AI specialists. Neither of us went to school for computer science, and neither of us has formal training in artificial intelligence. We have learned the basics of one tool, applied it to our own work over the last eight months, and want to make sure behavior analysts know what is now possible. The tool is called Claude Code. The name is misleading: you do not write any code. You type sentences in English. The skill is describing what you want clearly — the same skill you already use to write a behavior plan.
The tooling that makes this workflow practical — describing what you want in plain English and getting a working tool on your laptop — came together in late 2025. The pieces matured around November. Eight months later, the on-ramp finally exists, and the workshop catches you up on what changed.
One more piece of timing. Anthropic doubled Claude Code usage on Pro and Max plans through July 13. Our June 16 workshop sits inside that window. The same $20 plan buys twice the working time it normally would, which makes the next several weeks the right moment to learn this tool and use it heavily.
One of us has spent the last year using this tool every week. The numbers below are from Kevin’s Claude Code dashboard.
Source: Claude Code usage dashboard, May 2026. Behavior consultant by training. No computer science degree. No formal AI training. What it takes is describing what you want clearly. That is the same skill behavior analysts already use to write a behavior plan.
On the energy and water concern that often follows numbers like these — see the FAQ.
Each of the six below is a real tool one of us built in plain English in Claude Code over the last eight months. They cover teaching and practice, clinical tools, data visualization, and lit review. You bring the problem. We help you describe it clearly enough that Claude Code can produce a working version.
Turn any concept list into printable SAFMEDS-style flashcards with one prompt. Spaced-retrieval practice for stats, definitions, and behavioral procedures. Possible live demo at workshop
Enter how many stimuli, type each name, set replications, and out comes a printable data sheet pre-populated with left-right counterbalancing and a clean scoring grid. Replaces the Word-template gymnastics most BCBAs still do by hand. Possible live demo at workshop
Drop in raw data — or rechart someone else's published figure — and turn it into a NYT-, JAMA-, or JABA-style chart. Useful when you are not yet sure which visualization tells the story best. Helpful for posters, lit-review figures, single-case-design graphs, and exploring a dataset before the chart is final. The sleep graph at left was recharted from Mindell et al. (2010). Possible live demo at workshop
One workflow, two steps. First, drag a folder of messy PDFs in; Claude renames every file with a convention you set (author-year-title across the whole folder in seconds). Then those clean files become a searchable, filterable library, sliced by outcome, design, sample, or any field you encode. Coverage maps show which slices of the literature you actually have. Kevin built one across 105 sleep-research PDFs; same workflow scales to any topic. Possible live demo at workshop
Real teaching data sheet teachers print every week, with one layer added on top. Open the same file in a browser and you can type directly into cells, click to mark scoring, save the marked-up sheet as HTML, and toggle which columns appear that day. The print version is untouched — same file works both ways. The approach scales to any print sheet you already use: keep the paper workflow your team is trained on, and add a digital layer over it. Possible live demo at workshop
Design a token economy with your family or your client and see the math before you commit. Set the points-to-dollars conversion, daily and weekly caps, and the probability the learner earns each behavior. Two columns — what earns points, what the points can buy — recalculate live as you adjust. Once it pencils out, print the configured chart and start. Possible live demo at workshop
Any of the tools above could be the live build on the day. We build in front of you in real time, on Claude Fast Mode — typos, mid-prompt corrections, course changes, all of it. Then you open your laptop and build your own. The workshop is the on-ramp. The ceiling is what you build the next week.
If you have ever run a paired-stimulus preference assessment, written a behavior plan, or graphed within-subject data, you have the background this workshop assumes. Practitioners, supervisors, researchers, and the students working alongside them are all welcome.
Eight months ago, I could not have created any of these tools and documents.
I am a behavior consultant by training. I do not have a computer science degree, and I have not taken an AI course. What I have done is take seven paid AI workshops in the last year, walk out of most of them empty-handed, and spend eight months figuring out how to actually use Claude Code in my own work.
Most people think Claude Code is for programmers. It is not. What it takes is describing what you want clearly, in plain English. That is the same skill behavior analysts already use when we write a behavior plan.
Here is the math. I paid $2,713 across seven AI workshops last year and spent eight months learning what actually unlocks Claude Code for someone like me. The answer is not a prompt library. It is not a file structure. You talk to Claude the way you would talk to a smart colleague, and you let the creativity that is already in the tool do the work. That is the shortcut. Three hours. $48. One free follow-up call with me if you need it — up to 30 minutes, within 30 days. You walk in with a problem in plain English. You walk out knowing how to build a working tool yourself. If you are still stuck a week later, you email me and we get on a call.
Derek and I built this because we want to help behavior analysts use Claude Code to make their job and their life better. To go from nothing to something. Three hours, thirty seats, two of us in the room with you.
If you have ever paid for an AI workshop and walked out without anything to show for it, I have done that too. That is why this exists.
We are not AI researchers, consultants, or specialists. We have taken paid AI workshops ourselves over the last year. We learned the basics, applied them to our own work, and built the on-ramp we wished we had.
Pediatric sleep and school-based consultation. Clinical training at Kennedy Krieger Institute and Munroe-Meyer Institute. Built data visualization dashboards across 140 sleep graphs, the personalized PDF research library, a clinical form generator, and a file renaming system that enforces a folder-wide naming convention.
Author of a behavioral economics textbook (2nd edition). Built the Paired Stimulus Preference Assessment Generator, the SAFMEDS flashcard app, and a personalized PDF research library, all in plain English in Claude Code over the last eight months.
Three hours, built around the same four ingredients: live builds on Fast Mode, you in your seat building, both of us available the whole time, and one collective show-and-tell at the end.
Both instructors build in real time on Claude Fast Mode. Typos, mid-prompt corrections, course changes, all of it. The session may be recorded for replay; the experience is built around showing up live.
Thirty builders, two instructors, all visible to each other. We rotate through breakout rooms during build time so both of us touch every group.
You open your laptop and build. Both of us are available the whole time.
At the end, volunteers share what they built. You see what is possible and what to build next.
For reference, Kevin paid an average of $388 per workshop across the seven AI workshops he attended last year. He walked out of most of them empty-handed. This one costs $48, includes one free follow-up call with Dr. Luczynski if you need it (up to 30 minutes, within 30 days), and is taught by someone who has been on your side of the table.
Claude Pro subscription required ($20 / month, separate from ticket). Cancel anytime after the session.
Kevin has attended seven paid AI workshops over the past year, spending $2,713 across them. In several of those sessions, almost a third of the workshop went to troubleshooting initial setups for attendees. That time was a tax on everyone in the room. We are not doing that to you.
A heads-up on hardware. These steps require a Windows or Mac computer that is yours personally — phones and tablets will not work, and work-managed devices (schools, hospitals, universities, many clinics) typically block the app installs Step 1 requires. If a work device is your only option, email your IT department in advance and ask them to allow installs from Anthropic PBC (Claude Desktop App) and Git for Windows. If your computer is more than five years old, expect some steps to take thirty to sixty seconds to respond; wait, do not re-click.
We pilot-tested these steps on multiple Windows machines with real, non-technical users. Even on Windows, there was real variability in what the installer, the operating system, and Claude Desktop App prompted from one computer to the next — some setup screens appeared for one tester and not the next, and new dialogs showed up that we never anticipated. The Mac path has not been pilot-tested yet, so Mac users should expect the steps to be even more of a map than a script. Read the steps as a strong map of what we are asking you to do, then use your own judgment to bridge small differences you encounter. If you get truly stuck on something the steps do not cover, email Dr. Kevin at Dr.L@universalbcs.com and we will get you through it.
claude.com/download or Ctrl + Alt to copy it to your clipboard. The chip flashes Copied in green when it works. Saves you from highlighting and right-clicking each time.
Ctrl + J (Windows) or Shift + Cmd + J (Mac) to open your browser's download history. The most recent file appears at the top with a Show in folder link that opens its exact location.
Goal Install the free Anthropic desktop app, sign in, and upgrade to Pro — so Claude can run code on your computer when you get to Step 3. No terminal, no command line.
claude.com/download and press Enter.Claude-Setup.exe (or similar) downloads to your Downloads folder, your Desktop, or any location you like — some browsers pop up a Save As dialog before the download starts so you can choose where it goes. Most browsers also show the recent download as a small downward arrow indicator in the top-right of the browser window (click it to find the file). Firefox shows a similar arrow in the same place.Claude-Setup.exe.
Claude-Setup.exe. If you still see the gate, email Dr. Kevin (Dr.L@universalbcs.com) — this dialog has multiple causes and the fix may need to be different on your machine.Claude-Setup.exe still sitting wherever you saved it (usually Downloads, sometimes Desktop). That second file is the installer — one-time use only. Open the folder where it landed and drag Claude-Setup.exe to the Recycle Bin. Leaving it around tends to cause people to re-open it later, get a "Claude is already installed" dialog, and lose ten minutes figuring out what went wrong.Follow these sub-steps to upgrade and immediately cancel the auto-renewal so you are not charged a second month:
claude.com/download and press Return.Claude.dmg downloads to your Downloads folder, your Desktop, or wherever you save it.Claude.dmg. A new window opens with a Claude icon on the left, a folder labeled Applications on the right, and an arrow between them.Claude.dmg (Downloads for most people, Desktop for some), and drag it to the Trash. The .dmg is the installer; you do not need it again. The app you launch from now on lives in Applications.Command + Space to open Spotlight Search (Mac's universal search bar). Type Claude and press Return.Follow these sub-steps to upgrade and immediately cancel the auto-renewal so you are not charged a second month:
Goal Get the workshop's working folder onto your computer so Claude can read it in Step 3.
The prep kit (~14 MB) contains fourteen highly-cited JABA papers (Baer, Wolf, & Risley 1968; Stokes & Baer 1977; Iwata et al. 1994; Hanley et al. 2003; and friends) you will rename and organize during the workshop, a sample classroom dataset, and a real parent sleep-study dataset for the data-visualization demo.
workshop-prep.zip downloads to your Downloads folder, your Desktop, or wherever you save it.
workshop-prep with a zipper icon. Same file; Windows is hiding the extension by default.
workshop-prep.zip. Choose Extract All… from the menu.workshop-prep right next to the zip file. This new folder is what you will point Claude at in Step 3 — not the original .zip file (with the zipper icon), which you can ignore from here on out.workshop-prep folder right where it is, wherever it landed (Downloads or Desktop, whichever your browser used). Step 3 will ask you to point Claude Desktop App at this folder; the exact location does not matter. Done with Step 2.workshop-prep.zip downloads to your Downloads folder, your Desktop, or wherever you save it.
workshop-prep with a zipper icon. Same file; macOS is hiding the extension by default.
workshop-prep.zip. macOS unzips it automatically and a new folder appears named workshop-prep. This new folder is what you will point Claude at in Step 3 — not the original .zip file, which you can ignore from here on out.
workshop-prep folder, right-click workshop-prep.zip in Finder, choose Open With, then Archive Utility (default). macOS unzips it cleanly.
workshop-prep folder right where it is, wherever it landed (Downloads or Desktop, whichever your browser used). Step 3 will ask you to point Claude Desktop App at this folder; the exact location does not matter. Done with Step 2.Goal Have Claude produce a Readiness Card from your prep folder, then save it as a PNG you can submit in Step 4.
</> icon. Code mode opens. (If this is your first time using Code mode, you may see a dashboard with usage stats; that is fine.)
</> Code icon at the top of the left sidebar, you are on an older Claude Desktop App build. Click your name or profile icon in the bottom-left corner of the app, choose Settings, and check for updates. Quit and reopen Claude Desktop App after updating.
git-scm.com/downloads/win in a new browser tab (it redirects to git-scm.com/install/windows — that is the right page).workshop-prep folder lives (Downloads for most people, Desktop for some), click workshop-prep once to select it, then click Open (Windows) or Select (Mac). Claude opens the folder and automatically reads the CLAUDE.md spec inside.
CLAUDE.md, you opened the wrong folder. Click Select folder… again and pick workshop-prep, not its parent or a folder inside it.
Generate my Workshop Readiness Card now, following the spec in CLAUDE.md.
Name: [Your full name with credentials, e.g., Dr. Jane Doe, BCBA-D]
Institution or Organization: [Where you work — university, lab, school district, clinic, agency, hospital, private practice]
Plan: [Pro — you must be on Claude Pro, $20/mo, before the workshop]
Interests: [Three to five short phrases separated by middle dots]
Research or Clinical Question: [One sentence in your own voice — a research question or a clinical puzzle you are trying to answer]
Favorite Scientific Paper: [Author last name, year. Title. Journal, volume(issue), pages.]
Message to Dr. L and Dr. Derek Reed: [One real sentence to Kevin and Derek]
Ctrl + V (Windows) or Command + V (Mac). The prompt fills the box.
Sonnet 4.6, Opus 4.x, or similar. Whatever Claude Desktop App picked by default is fine for the Readiness Card. Do not change it.
[Your full name with credentials]. Replace each one — including the brackets themselves — with your real answer. The labels (Name:, Institution or Organization:, etc.) stay exactly as written. The Plan: field must say Pro — if you write Free, Claude stops and asks you to subscribe before generating the card. Look at the example just below to compare blank versus filled. Casual phrasing is fine; do not feel pressure to sound formal.
Same shape as the prompt above, with one possible set of answers. Do not copy this verbatim — replace with your own work. Casual phrasing is fine.
Generate my Workshop Readiness Card now, following the spec in CLAUDE.md.
Name: Dr. Jane Doe, BCBA-D
Institution or Organization: Springfield Center for Behavioral Health
Plan: Pro
Interests: Early language · Caregiver coaching · Bedtime routines · Single-case design
Research or Clinical Question: Does a quick caregiver interview predict who will need functional communication training?
Favorite Scientific Paper: Hanley, Iwata, & McCord, 2003. Functional analysis of problem behavior: A review. JABA, 36(2), 147-185.
Message to Dr. L and Dr. Derek Reed: Excited to learn how to use AI to build my own tools for the clinic.
dataset.csv, writing readiness-card.html).
readiness-card.html inside your workshop-prep folder.workshop-prep folder and double-click readiness-card.html to open it manually.file://), paste it into Chrome's address bar, press Return, and click Issue Card again. Or use the screenshot fallback below.
Win + Shift + S, drag a box around the entire card, release; the PNG saves to your clipboard, paste it into Paint and save. Mac: press Cmd + Shift + 4, drag a box around the entire card, release; the PNG saves to your Desktop. Either screenshot is fine to submit in Step 4.
Goal Send us your PNG via the Tally form so we can verify your setup is ready and email you the Stripe payment link.
Once we approve your card, we email a Stripe payment link. One click, $48.
Goal Save the Zoom link from the redirect page Stripe sends you to after payment. That is where you will find it.
.ics file link) on the redirect page. Outlook, Apple Calendar, and Google Calendar all import it automatically. If your client does not, copy the date and time from the page and create the event manually.Steps 1 through 6 above are the hard floor for attending. Everything required is done. The add-on below is not for setup — it is a tool you will use every day after the workshop, for email, Slack, Word, and especially inside Claude Desktop App. Install it the morning of the workshop (not now — later) if you have ten clear minutes.
Wispr Flow turns push-to-talk dictation into typed text anywhere on your screen. Hold a key, speak a sentence, release — the words appear where your cursor is, with punctuation and capitalization handled for you. This is not just for the workshop. Once installed, you use it every time you would normally type more than a sentence: emails, Slack messages, notes, search bars, and yes, the chat input inside Claude Desktop App. Kevin and Derek both build with Wispr Flow running. It is what makes describing a problem to Claude feel like talking to a colleague, not typing prompts.
Sign up takes about a minute; the installer is the same on Windows and Mac. Use the Download button below and you get a free month of Wispr Pro through our referral link — no code to copy, the perk attaches automatically.
Ctrl + Alt together. A small mic indicator appears.Fn alone to start dictating.Ctrl + Option together instead.Alt + Tab (Windows) or Command + Tab (Mac) to cycle through open windows until you find it. Click to bring it forward. The "frozen" screen is just waiting for you to interact with the hidden dialog.
A fair question, and a real one. Training and running large AI models consumes meaningful amounts of electricity and water — for data-center cooling and power generation. People in our field are right to ask about it.
Honest answer: we know less than we should. Estimates from peer-reviewed research (notably Li, Yang, Islam, & Ren, 2023, "Making AI Less Thirsty," University of California Riverside) suggest a single AI prompt may use somewhere in the range of 10 to 25 milliliters of water and a few watt-hours of electricity. The much larger footprint sits in industrial-scale model training and data-center construction — not in individual professional use. These numbers are estimates, vary widely by model and data center, and the industry has only recently started publishing per-query figures.
What we do here: we use AI deliberately, not casually. We push the providers we use to publish their figures and reduce their footprint. We will not pretend the concern is unfounded. If you decide the environmental cost outweighs the professional value for you, that is a fair call.
Four commitments we operate under and recommend to every attendee.
1. Cite your sources. When you pull a claim, a number, or a framework from a paper or a colleague, cite it. Claude wrote the paragraph; the underlying idea still belongs to whoever published it first. Use the same citation discipline you would for any research-informed writing.
2. Verify every numeric claim. Large language models hallucinate citations, statistics, and even author names confidently. The skill this workshop teaches is using AI to draft and verify, not just draft. If a Claude output cites a study, open the study before quoting from it.
3. Protect client and family data. Do not paste protected health information, FERPA-protected student data, or any identifying details into an AI prompt. De-identify aggressively. Anthropic documents which plans store prompts for review and which do not — check your account settings — but the safer professional default is: assume any text you send leaves your control, and write accordingly.
4. Get consent when relevant. If you are using AI to generate content that will be used directly with the people you serve — clinical plans, parent-facing materials, intervention scripts — the people you serve get to know. Tell families. Tell teachers. Be explicit about which parts came from Claude and which came from you.
For the deeper academic conversation on bias, equity, privacy, consent, and AI literacy in our field, the work to follow is Project CHIRON, the AI literacy series for ABA professionals from Ryan O’Donnell and Dr. David J. Cox. Read them there; build with us here.