# Rewrite memo — agetech **Word count (body):** ~1,240 words · **Hero variant:** V2D · **Anonymisation:** Tier B --- ## 1. The story I found The existing case has the argument — but it's distributed across 11 sections and framed as a build log. The actual story is: a concept needed a decision-point test that static mockups couldn't provide. Whitney built a working AI from scratch, tested it with seven caregivers, discovered the structure — not the flow — was wrong, rebuilt the interaction model mid-study, and the rebuilt version converted the executive holding the budget. The through-line a reviewer should remember three weeks later: **building the prototype found the flaw that no amount of concept refinement would have found, and fixing it mid-study is what made the beta investment defensible.** That argument was mostly buried in the closing sections of the existing case, sandwiched between process documentation. The rewrite moves it to the front and makes it the spine. The reframe I found: v1 was built on the assumption that an AI caregiving service should work like a thorough intake interview — gather everything first, deliver value at the end. The research showed that caregivers are already managing a crisis; they need tools that deliver value immediately, not another system asking them to give before it gives. This assumption was diagnosed in the existing case but framed as a design decision (v2 is different from v1) rather than a named reframe (we were wrong about something fundamental). The rewrite makes it a standalone section. --- ## 2. What I kept - **"v1 asked. v2 helped."** The existing case's single sharpest sentence. Kept verbatim; moved to lead Chapter 04. - **The end-stage dementia session.** The most emotionally resonant and diagnostically important beat in the case. Kept with the same weight; moved to close Chapter 03 as the final evidence that v1's failure went deeper than sequencing. - **The "I just gave you all that information" quote.** Strongest pull-quote in the case. Kept in Chapter 03. - **The executive demo.** The case's proof point. Kept in full, moved to its own chapter (05). - **The comorbidity rabbit hole.** Kept not as a failure story but as evidence of disciplined scoping — building to a purpose, not building to fidelity. Moved to Chapter 02. - **The five non-negotiable rules.** Kept as a structural beat in Chapter 04; the link to specific v1 observations is preserved. - **The "freaked us out at the time" quote from the executive.** Kept as the honest coda to Chapter 05 — it names the forecast's remaining content problem without smoothing it over. --- ## 3. What I cut and why - **"I joined eight months in. Scope was: build the thing, then learn from it."** (Section 2 in the original.) This is the longest piece of cut content — a full section establishing Whitney's AI fluency and her role's constraints. It was doing defensive work (anticipating "who are you to build this?") rather than argument work. The two essential facts (joined mid-engagement, scope included building and testing) are folded into Chapter 01 in two sentences. The AI fluency context isn't needed — the work demonstrates it. - **"A constraint worth naming about the testing itself."** The client privacy policy explanation (participants watched rather than typed directly). One honest sentence in Chapter 05 covers it without stopping the argument. - **"What the second cohort showed"** (as a standalone section). The findings from cohort 2 are real, but the paragraph-length version in the original was thin. The essential signal — participants leaned in, nobody asked about the value exchange — is folded into Chapter 05 with appropriate compression. - **"Four things that traveled with me out of this project."** This was a lessons-learned closing — exactly the anti-pattern the spec calls out. The real scaling content (model as production spec, testing mandate carried forward) was present in this section and in the preceding "client got the signal" section, but framed as personal takeaways. I extracted and relocated it to the Scaling section, reframed as what the next phase inherited. - **"Two intertwined problems"** and the subheadings "What to ask" / "How to ask it." These were internal section headers, not argument-shaped. The content they organized is preserved in Chapter 02; the headers are gone. --- ## 4. What I added - **The argument paragraph.** Didn't exist in the original. The case opened with a StatRow and then a section titled "A concept with eight months of work behind it." The rewrite opens with an 80–120 word thesis paragraph that states the case's argument before anything else. - **A formally named reframe section.** The pipeline-vs-tool assumption was present in the original as framing for the v2 redesign. The rewrite elevates it to a standalone reframe section with the explicit pattern: "We built [X] assuming [Y]. The research showed [Z]. The fix had to happen at the model level." - **A formally named scaling section.** "The client got the signal they needed" in the original was the case's closing section and did double duty — outcome summary + what-traveled-forward. The rewrite separates these: the executive demo becomes Chapter 05 (evidence), and the what-the-next-phase-inherited content becomes the Scaling section. - **V2D hero with chapter spine.** The existing case had no hero structure. The spine (7 chapter titles readable as a sequence) is new and designed to let a skimming reviewer understand the arc before committing to the read. --- ## 5. Hero variant choice: V2D (spine-led) This case's argument runs through a sequence: build → test → discover → rebuild → prove. That sequence is the case. A reviewer who sees the chapter spine in the hero — eight months in / building the AI / seven sessions showed the structure was wrong / pipeline to tools / executive stopped evaluating — understands the arc before reading a word of body copy. V2B would put the title "The prototype that proved the brief wrong" in the hero position; that's a strong title but it names the conclusion without showing the move. V2D shows the move. For a methodology case with a midpoint pivot, V2D earns the space. --- ## 6. Open questions for Whitney 1. **The executive demo quote.** In the original case, the pull quote is attributed to "Note from the session" — *"We can't get him to stop."* Is this quotable in a published case? If so, is "Note from the session" the right attribution, or should it be "facilitator's session notes"? This beat is the case's strongest proof point; the attribution affects how reviewers read it. 2. **The fall 2026 beta timing.** The original MDX was written in April 2026 and references "a two-month beta this fall." Is that still accurate as of publishing? If the beta has since run, the scaling section should reflect what actually happened rather than what was planned. 3. **The v2 service blueprint.** The raw case material contains `Aging _ Blueprints Working - Aging, Service Blueprint, V4, 04_26.svg` — a significant visual artifact. Is there a blurred/redacted version suitable for the case, or does it need to stay in the asset-needs list? The case would be meaningfully stronger with it in Chapter 01 as the establishing visual.