Stay Sharp • For Practitioners Already Running AI
Open something you shipped yesterday — an email, a lesson outline, a sales-page draft, a client brief, whichever one is closest. Read the first sentence.
Then ask yourself one question — and answer it before you read anything else on this page:
Did that sentence start in your head, or in the AI’s?
If you can’t tell, that is the page. Keep reading.
If you can tell — and it started in yours — close this tab. This isn’t your page. You don’t need what’s on it.
The companion course to AI-Powered Solopreneur — and the one that answers the question almost no one is asking yet. Eleven interlocking frameworks distilled from 202 research papers on AI and professional practice, assembled across 16 modules in 4 units plus a capstone, ending in one living document: your AI Practice Design. Roughly six and a half to eight and a half hours, self-paced, anchored by a single sentence.
Stay Sharp is the AI Practitioner Sovereignty System: a governance discipline of eleven frameworks — distilled from 202 research papers on AI and professional practice — that you apply to the AI you already use, so extended AI use sharpens your judgment instead of eroding it.
AI extends what you’ve built. It cannot be what you haven’t built. And it will erode what you stop building.
You’ve been using AI for a year, maybe two. Output has gone up. Hours have come down. You took the workflow course, or you figured it out yourself, and AI is genuinely running in your practice. That part is real.
And then, a couple of weeks ago, in a moment that lasted maybe ninety seconds, something quieter surfaced. You went to draft something the old-fashioned way — without AI, just to see — and the cursor blinked longer than it used to. Not because you didn’t know what to say. Because the part of your brain that used to start the saying felt… cooler. Less rehearsed. Slower to engage.
You closed the document, opened the AI, and finished the draft in four minutes. You moved on.
That ninety seconds is the second question. Nobody warns you about it. The better the AI gets, the harder it is to notice. Your output keeps improving. Your clients don’t complain. Your week feels lighter. And quietly, one accepted suggestion at a time, the cognitive exercise that built the judgment your clients pay you for stops getting done.
Here is what that shape looks like in real practitioner work — and what every framework in this course is engineered to surface before it becomes structural.
An oncologist with twenty-five years of experience told a researcher last year: “My intuition is rusting.” She had not stopped being effective. Her diagnostic accuracy was as strong as ever. But the rapid, experience-built pattern recognition that used to fire on its own — the perceptual reflex that distinguished a competent diagnostician from a great one — was less active than it used to be. She was still getting right answers, but through a different path. The path that built the answer-finding capability was no longer the one she walked.
What practitioners actually report, across professions
# Oncologist, 25 yrs:
“My intuition is rusting. I get the right answer—
but the diagnostic reflex feels colder.”
# Course creator, 5 yrs:
“I can’t remember the last time I outlined a module
without asking the model what the structure should be.
The frame used to arrive on the walk to the desk.”
# Indie coach, 8 yrs:
“Session prep takes half the time. My real-time read
on what the client is actually saying feels half a step
behind where it was.”
# Developers (controlled study, 2025):
“We feel 24% faster with AI. We’re measurably 19% slower.”
— A 43-point gap between perceived and actual.Four practitioners, four professions, one mechanism. Capability that grew through repetition and challenge stops growing when the challenge is handed to a tool. Existing capability, without continued exercise, atrophies. The output keeps arriving on time. The substrate underneath it goes thinner.
The oncologist noticed her erosion at month eighteen — by which point the perceptual capability had already weakened.
I noticed mine on a Tuesday afternoon, drafting an email I’d have written in two minutes flat eighteen months ago. The cursor blinked longer than it used to. I have thirty-six years of building production systems behind that cursor — twenty-four of them at TD Bank, a dozen books, the operational instincts that paid for my house. If anyone should have felt that delay as background noise, it was me. I didn’t. The delay was real, and I knew exactly what it was.
No client lost. This time.
That moment clarified the same thing the four practitioners above eventually clarified, each at their own moment. AI tools that genuinely amplify output also conceal a second mechanism — one the dashboards never show because the dashboards measure the surface, not the substrate. The Right Column is the column nobody is reading. It exists. It is unmanaged. And unmanaged consequences become structural.
There is a name for what the cursor was telling me, and what the oncologist’s reflex was telling her, and what the consultant’s missing first hypothesis was telling him. Authorship Drift. The slow, asymptomatic migration of the originating moment of professional thought from inside the practitioner to inside the tool. The output keeps arriving on time. The author of the output quietly stops being you.
That was the moment I stopped asking how to use AI better and started studying what AI was doing to the practitioners who used it most.
01The Mechanism
After a year of reading the research and field-testing the frameworks — my own practice plus the practitioners I’ve compared notes with — every signal of the second question traces back to one of four named mechanisms. Each is documented in the research. Each is invisible from inside the working day. Each compounds with the others. None of them are prompting failures.
Capability grows through repetition and challenge. When AI handles the hard parts of a task — the strategic hypothesis, the diagnostic reflex, the original framing — those parts stop developing in you. The erosion is asymptomatic: speed goes up, volume goes up, apparent quality goes up. The substrate underneath — judgment, pattern recognition, domain intuition — goes thinner. The Lag Window between when erosion begins and when it becomes detectable runs three to twelve months. By the time it surfaces, it is chronic.
You ask AI a question. It returns a coherent answer. The answer is plausible. You move on. The cognitive step you used to take — weighing the answer against your domain knowledge, the specific client context, the things you know that the model cannot — happens less, and then doesn’t happen at all. Not because you decided to skip it. Because the answer was already there, and the step felt redundant. The Belief Offloading Spectrum runs from verify-everything at one end to accept-by-default at the other, and movement along it happens by accumulation, not by decision.
Not what you think about. How. Every extended AI interaction leaves traces — a framing adopted from the model’s structure, a confidence level adjusted upward, a reasoning path that follows the model’s logic rather than yours. Individually invisible. Cumulatively, over months, they constitute a shifted baseline. Drift is by definition below the threshold of awareness until it is significant — which is why it needs a scheduled diagnostic, not a vague intention to notice.
Despite extensive AI adoption, practitioners overwhelmingly conceal it. Each one fears that admitting use signals dependence rather than expertise. So everyone hides. The result is a market where AI use is ubiquitous and invisible, and where output quality — the signal that used to distinguish practitioners — no longer differentiates. What survives as a signal is how you govern: how you think in real time, handle novel situations, navigate uncertainty. The practitioners who will be trusted most are not those who hide AI most successfully. They are those who are most transparently expert in governing it.
They compound. Erosion makes drift harder to detect — you cannot calibrate your thinking baseline against a substrate that is itself shifting. Drift accelerates offloading — shifted evidence standards lower the bar for what counts as a sufficient answer. Offloading deepens erosion — the cognitive step you stop taking is the step that kept the capability alive. And hidden competence quietly turns the whole thing into the new market normal.
That compounding is the Right Column. Each mechanism is survivable on its own; your professional skill patches over it. Together, they create a structural drift that widens with every assisted draft. The system that closes them runs the other direction: each capability you protect seals the next erosion, each diagnostic you run surfaces the next drift while it’s still small, each governance disclosure converts hidden competence into demonstrated authority. The system gets sharper the longer it runs.
02A Tuesday With the System Installed
You open the day with a brief written in your own terms before any AI tool is loaded. Ten minutes of framing the problem on your own — the first stage of the Cognitive Mode Compass. The framing is yours, not borrowed. The hard part of the work happened before the first prompt; the AI now amplifies a thought you actually had. Erosion paused this morning — the substrate did its job.
A client deliverable. You consult the Trust Visibility Calculus row for this workflow: AI runs as Researcher (not Oracle), output gets a rubric-based review (not vibe check), final framing stays yours. The Delegation Archetype is explicit, written down, and it has not migrated since you last reviewed it. No gate erosion. The boundary you set last quarter is the boundary you operate today, because it is on paper, not in memory.
The AI produces a phrase that flows nicely but slightly resolves the ambiguity in the client situation — smoothing where smoothness is wrong. You catch it in three seconds, because the Intent–Execution Gap module trained you to read for exactly that shape. You rewrite the sentence in fifteen seconds rather than letting it ship into a draft you’ll review next week. The Practice Gap log gains one entry. Quality migration surfaced the moment it tried to form.
A voice memo from a peer in your mastermind: how are you using AI these days? You record your reply on a walk — the archetypes you’ve assigned, the cognitive modes you protect, the drift checks you run monthly — without translating. The terms arrive on their own, the way onboarding call and scope amendment arrive when you talk about your client process. You aren’t reciting the curriculum; you’re using it. The peer’s reply, twenty minutes later, says you’re the first person they’ve heard describe AI use as a discipline rather than a tool stack.
First Monday of the month. You open the four-point Metacognitive Drift Diagnostic and spend twenty minutes on it. Problem-framing: still mine. Evidence standards: tightened back up after last month’s loosening. Confidence calibration: one entry to log. Independent judgment: ran the unassisted-draft drill on Friday, and the cursor blinked less than it did six weeks ago. The Practice Gap drill is doing what it was designed to do. The system is sharper today than it was last month, because every iteration COMPOUNDS — every loop closes a small drift before it sets.
That is the day you build, framework by framework, across 16 modules and a capstone — text-based, self-paced, anchored by exercises you complete in your own practice between drafts. Eleven named frameworks. Four units plus orientation plus capstone. One living document at the end.
03Pattern Recognition
Every solo operator at your level has done this. You built your expertise over a decade or two, and along the way you developed governance for the things that mattered. You did not let your judgment drift. You did not let your reading of clients atrophy. You did not let your standards leak. When the capability that pays you was at risk, you protected it — with practice, with discipline, with a habit that ran whether or not you felt like running it.
The coach who keeps her perceptual edge sharp does it on purpose. She runs unassisted client formulations every week because she knows that thirty hours of AI-prepared frames will dull a muscle that needs use. The course creator who can still outline a module from a blank page does that on purpose too — he keeps an unassisted block on his calendar every Friday morning, and he does not give it up. Neither of them got there by accident. They built a governance habit for the capability that paid them.
The same discipline applies here. The Right Column is not a new category of professional risk. It is the same category your career has always managed — the difference between the work that’s easy to ship and the capability that makes shipping it valuable. You have governed before. You are about to govern one more thing.
What you want is not a productivity course. You don’t need one. What you want is the governance layer for the one tool in your practice that’s changing what you are while you use it — with the same precision you already bring to the rest of your craft.
04The Frameworks
You finish Stay Sharp with eleven interlocking diagnostic and design tools organised across four units — not notes from a course, not theory you forget in three weeks. Each framework is presented in the same eleven-section module template (Word, Core Idea, Model, Worked Example, Bad/Improved, Discovery Exercise, Rubric, Peer Activity, Fast Track, Connection Map) so once you learn the rhythm, you read faster. The capstone pulls the artifacts forward into a single living document: your AI Practice Design.
If you’re skimming
A Diagnostic Map by hour 1. An Assessment Map by hour 2.5. Practice Design Draft 1 by hour 6. A Maintenance Protocol you run monthly thereafter, plus the final AI Practice Design at the Capstone. Four artifacts, eleven frameworks, one living document.
Two diagnostic frameworks that surface the mechanisms operating beneath your day-to-day work, plus a synthesis gate that produces your Diagnostic Map — the artifact every later unit consumes.
The Hard-Parts Audit. You name the hard part of each AI-assisted task, identify whether AI is handling it, and surface the erosion signal — the moment you realise you are not sure you could still do this without help. The Surface/Substrate/Lag-Window model that explains why your dashboards will never warn you.
A scaled diagnostic that maps where you sit between verify-everything and accept-by-default, per workflow. Names the cognitive step you stopped taking — and the conditions under which the step was first skipped. Pairs with the Amplifier Paradox to produce a two-axis read on your current practice.
The Expertise-Literacy Leverage Matrix, plus the Unit 2 synthesis that produces your Assessment Map — the first usable artifact in the course. Identifies where the work compounds and where the risk concentrates.
Where in your practice does AI compound your specific expertise, and where does it merely substitute for skill it cannot replace? The two-by-two that distinguishes high-leverage augmentation from quiet hollowing — a positioning artifact you return to whenever a workflow decision needs framing.
Five design frameworks — one per Word — that convert the diagnostic into operational decisions. The Unit 3 synthesis produces Practice Design Draft 1: five specific decisions written down for your specific practice.
Why specification precision is the single highest-leverage skill in AI use — and the leading indicator of drift when it slips. You author the brief; AI executes the brief. You do not delegate the authoring.
Researcher, Drafter, Reviewer, Sparring Partner, Oracle — the five roles AI can play. Most practitioners have given AI the Oracle role without acknowledging it. The Map makes the role explicit per workflow, so the role doesn’t migrate on its own.
A row per workflow naming how much human review each output gets, calibrated to the measured failure mode. The component that keeps gate erosion from happening on its own — because it’s on paper, not in memory.
Where AI’s output deviates from what you actually meant — smoothing where smoothness is wrong, resolving ambiguity you wanted preserved. Pattern-recognition training so you catch the gap in seconds, not weeks.
Individual mode, Pair mode, AI-Assisted mode. Three modes of professional thought, with a calendar discipline for which mode runs when. Individual mode is the antidote to erosion — and the one most likely to disappear if you do not protect it.
Three maintenance frameworks plus a synthesis that converts the practice design into a recurring protocol. The system that keeps your edge as AI keeps changing.
A four-point diagnostic (problem-framing, evidence standard, confidence calibration, independent judgment) run on a monthly schedule. The early-warning system for the rest of the architecture.
Deliberate practice for capabilities AI has removed from your workflow. Specific unassisted drills, on a calendar, with a measurement. The mechanism that keeps the substrate growing.
The prisoner’s dilemma of AI transparency, and the community positioning stance that converts governance into authority. Designed to be deployed in client conversations — not just internalised — so your discipline becomes a visible competitive signal.
All eleven frameworks are tool-agnostic by design — the diagnostics and the design decisions work whether you operate in Claude, ChatGPT, or whatever ships next. This is the sovereignty layer, not a tool stack. It survives every model swap and every interface redesign because it lives in how you practise, not in which window you type.
05The Curriculum
11
Frameworks
17
Modules
4
Units + Capstone
| Unit | Core Question | What You Build |
|---|---|---|
| 0: Orientation (0.0) | What is sovereignty? | Baseline Capture — seven-question record of where your practice sits today, plus your fast-track path selection |
| 1: The Diagnostic (1.1, 1.2, 1.S) | What is AI doing to me? | The Amplifier Paradox + Belief Offloading Spectrum, synthesised into your Diagnostic Map |
| 2: The Assessment (2.1, 2.S) | Where do I stand? | The Leverage Matrix, synthesised into your Assessment Map — first usable artifact, ~2.5h in |
| 3: The Design (3.1–3.5, 3.S) | How do I structure my practice? | Specification, Delegation Archetype, Trust Visibility, Intent–Execution Gap, Cognitive Mode Compass — assembled into Practice Design Draft 1 |
| 4: The Maintenance (4.1–4.3, 4.S) | How do I stay calibrated? | Drift Diagnostic, Practice Gap drills, Hidden Competence Trap — converted into a recurring Maintenance Protocol |
| X: The Capstone (X.0) | What is my complete design? | AI Practice Design — your final living operational document, scored against the eleven-metric checklist and closed by the Sovereignty Commitment |
Self-paced. Text-based. Every module ships in the same eleven-section template so once you learn the rhythm, you read faster. Roughly six and a half to eight and a half hours from start to the AI Practice Design at the Capstone. Units 0–2 alone (the diagnostic + assessment half) run around two and a half hours and produce a usable Assessment Map you can run with while you work through the rest. Exercise answers persist server-side — stop mid-module, close the tab, return weeks later to exactly where you left off. The Capstone pulls forward everything you wrote.
06Audience
Coaches, consultants, content creators, course creators, service professionals — solopreneurs running indie practices with AI already in them. You’ve been using AI for at least a year. You don’t need the workflow course. You need the governance layer for the capability your career was built on. Self-paced. No cohort, no live calls. Install the system in your real practice at your own pace, with lifetime access to revisit any framework as the second question deepens.
No — they are deliberately the two halves of one discipline. AIS answers Question One: how do I get AI working in my business? Stay Sharp answers Question Two: what is AI doing to me while I run it? AIS gives you the Business Context Profile, the Workflow Library, the Trust Calibration Map, the Improvement Loop — the operating system that produces reliable output. Stay Sharp gives you the diagnostic for the substrate underneath that output — the part that erodes asymptomatically while the outputs keep arriving on time. The practitioner who takes AIS is the ideal buyer for Stay Sharp the moment Question Two surfaces in their own work — for some that’s weeks, for others a few months. Both courses open together on May 26; take them in sequence, or take both at once if the second question has already surfaced.
It isn’t. There is a version of this curriculum you could misread as a warning against AI — a case for going back to the way things were. That is not what it is. Sovereignty is not about using AI less. It is about using AI in a way that leaves you more capable, more confident, and more authoritative in your practice. The frameworks are diagnostic and design tools, not prescriptions. AI extends what you’ve built. The work is to make sure you keep building.
Better models produce more fluent outputs faster. Fluency is exactly the surface signal that masks substrate erosion — the better the model, the more polished the deliverable, the harder the Right Column becomes to see. Capability gap doesn’t close because the tool improved; it widens, because the cognitive exercise the tool removed becomes harder to justify reintroducing. The eleven frameworks are tool-agnostic by design — they outlive every model swap.
Getting by is not how you built the rest of your practice. You built the client process, the delivery system, the pricing structure because fine was not what you were after. You wanted reliable. The Right Column is the layer of your AI use that still runs on hope.
Not for beginners who’ve never opened ChatGPT — the frameworks assume an active professional practice with AI already in it. If your problem is “I don’t know what to type,” this isn’t the course. If your problem is “something quieter is happening to my work and I can’t name it yet,” this is exactly the course.
07The Offer
Self-paced · Text-based · 17 modules + capstone · 4 units + orientation · 9 appendices · ~6.5–8.5 hours
Founding pricing holds for the first 50 seats or through Monday, June 8, 2026, whichever comes first. After that, the price steps up to $497.
Founding pricing holds for the first 50 seats or through Monday, June 8, 2026, whichever comes first. After that, the price steps up to $497.
The 50-seat cap is real and bounded: 50 seats cumulative across the Wave 1 founding window, the standard window between them, and the September Wave 2 window. If Wave 1 sells 18 seats and the standard window sells 6 more, Wave 2 starts with 26 remaining. When the 50th seat sells, the founding-tier door closes for good. There is no second founding window.
That is the entire scarcity claim. A cap, a date, a price, a window. The truth.
The 30-day refund is unconditional and yours to call. Inside it, there is one named date the course is engineered around.
You bought on Tuesday, May 26. You worked through Units 0, 1, and 2 in the first two weeks. On Tuesday, June 9, 2026 — fourteen days after purchase, two weeks of the Diagnostic Map and the Assessment Map sitting in your practice — you open a fresh document with no AI loaded and you draft something the old-fashioned way. The same drill the course opens with.
You measure the cursor. If it blinks the same length as the moment you noticed it before this page, you have your answer: nothing measurably moved. Reply to the receipt. The refund is sixteen days inside the 30-day window. Every dollar back. You keep the Diagnostic Map and the Assessment Map for the trouble.
If it blinks shorter — even by a fraction — the substrate is already responding. That is the signal the system is doing what it was engineered to do, and you keep going. That date is the contract.
No testimonials. No case studies. This is a founding edition — you are evaluating the engineering: eleven frameworks drawn from 202 research papers on AI and professional practice, organised into seventeen modules, assembled at the Capstone into a single document you operate from. $297 is calibrated against the cost of one mis-aimed client deliverable — the kind that ships when AI smooths your judgment instead of carrying it. The price reflects that you are early. The depth does not.
From Pierre
Most AI courses make loud promises. This is the quiet one.
Two years from now, the practitioners who govern AI deliberately will look measurably different from the ones who don’t — in their judgment, their voice, the capability that makes their work worth paying for. The dashboards will not show the difference. The clients will. The peers will. You will.
That is the bet. That is the whole bet.
You have managed a capability before. You have built the discipline that keeps your strategic thinking sharp, or your perceptual reading, or your craft, or your delivery standard — whatever your edge is. You did not let it drift. You did not let it atrophy. You built the practice that kept it growing, and you ran the practice whether or not you felt like running it.
Authorship Drift is one more capability that needs the same treatment. It is not a special category. It is not a doom case. It is the cognitive exercise AI has quietly removed from your professional week, and the four mechanisms underneath that removal — Erosion, Offloading, Drift, Erasure — that compound while your output keeps improving.
The eleven frameworks close them, each in turn. Each capability you protect seals the next erosion before it sets. Each diagnostic you run surfaces the next drift while it’s still small. Each governance disclosure converts hidden competence into demonstrated authority.
The hard part stays yours.
The substrate stays alive.
The judgment stays calibrated.
There are two Tuesdays in front of you.
The first is Tuesday, May 26, 2026 — cart open, the founding window begins, fifty seats start moving. You either take one or you don’t.
The second is Tuesday, June 9, 2026 — the Day-14 Calibration Gate. You run the unassisted draft. You measure the cursor. You compare it to the moment you measured it before reading this page. If it didn’t move, the refund is right there. If it did, you keep going, and the same instrument becomes the weekly drill, and a year from now the cursor blinks the way it used to.
AI extends what you’ve built. It cannot be what you haven’t built. And it will erode what you stop building.
36 years building production systems — 24 of them at TD Bank, leading teams across regulated work, with ten-plus technical books on the side. I left in 2026 to build Curio Chat Academy and ran into a question the workflow courses don’t answer: not how to use AI well, but what extended AI use does to the practitioner using it. So I read the 202 papers and built the curriculum that turns the second question into a discipline you can run. I’m running the same governance experiment in my own solo practice, in real time — Stay Sharp is what I’m using on myself. This is how I teach it.
The Practical Questions
Everything between deciding and running the system in your real practice. If your question isn’t here, email hello@curiochat.com — an actual human reads every message.
ThriveCart redirects you straight to /account/ with your email already in the URL — the page verifies your purchase against Kit and shows the Stay Sharp modules (seventeen plus the Capstone) within a few seconds. For later logins: go to /account/, type your purchase email, you’re in. No password.
Two paths, both honored equally.
The standard path: reply to the ThriveCart receipt within 30 days. One-line request, no questions, full refund within 5–7 business days.
The Day-14 path: on Tuesday, June 9, 2026 (or the date that falls fourteen days after your purchase, whichever applies), run the unassisted-draft drill the course is built around. Measure your cursor. Compare to before. If nothing moved, that’s the falsifiable refund trigger built into the offer — same one-line email, full refund, sixteen days inside the 30-day window. You keep the Diagnostic Map and the Assessment Map for the trouble.
No. Stay Sharp is a governance system — the eleven frameworks are diagnostic and design tools you apply to whatever AI you already use. If you’re not yet running AI in your practice, this is the wrong course — try AI-Powered Solopreneur first. If you’re running on any tier of any model, the frameworks apply.
Your AI Practice Design is the deliverable — you author it, you keep it, it lives in your own notes or Notion or wherever you keep operational documents. Course content remains accessible as long as the platform runs; the artifacts you built using the course outlive the platform by design.
No. Stay Sharp is self-paced, text-based, no cohort, no live calls, no Q&A sessions. The eleven-section module template is built to read on its own. If you want live access, this isn’t the course.
Fifty founding seats, cumulative across three windows: the Wave 1 founding window (May 26 → June 8), the standard window between them, and the Wave 2 founding window in September. If Wave 1 sells 22 seats and the standard window sells 4 more, Wave 2 starts with 24 founding seats remaining. When the 50th sells, founding tier closes for good — no second founding window. Standard pricing ($497) stays available after.
Yes. Mobile-first responsive. Exercise answers save server-side, so you can run the Hard-Parts Audit on a phone during a coffee break and continue the Assessment Map at your desk. No audio version.
ThriveCart calculates tax based on your billing address at checkout. Most U.S. buyers see no sales tax; EU and Canadian buyers see VAT or GST per local rules. Receipt is itemized and suitable for expensing.
Yes — professional-development expense, deductible in most jurisdictions. ThriveCart generates a business-name invoice when you enter business details at checkout.
No. The 202 papers behind the curriculum are synthesised into the module text — you get the operational result without needing to read the academic source. Eight reference appendices (Glossary, Concept Index, Prompt Library, Cross-Domain Translation, Common Misconceptions, Portfolio Template, Measurable Objectives, Emerging Trends) give you the structured reference surface most practitioners actually use. The bibliography is not shipped as a separate reading list — if that’s a blocker, email us before you buy.
Updates ship to enrolled students at no charge. If a substantially new module joins the curriculum, founding-wave buyers keep access at no upcharge.