-
-
QuillCoach Logo
QuillCoach
Inspiration
QuillCoach emerged from a fundamental observation: scholarship tools optimize for matching, not for winning. Students get lists of 500 scholarships they might qualify for, but almost none help them actually succeed in those applications.
What struck us most was the hidden gap between what scholarships claim to value publicly and where they're actually headed strategically. A merit foundation might emphasize GPA, but internally they're pivoting toward innovation and social entrepreneurship. Students writing generic essays aligned with yesterday's priorities are competing at a disadvantage.
The second inspiration came from watching brilliant students systematically undervalue themselves—minimizing authentic achievements with qualifiers like "I just helped organize..." or "I only managed to..." We observed this pattern so pervasively it seemed cultural. Humility is an amazing trait that helps students collaborate and succeed but hurts them in the application process. These students possessed real strengths they couldn't see, and they lacked tools to ethically extract and amplify their value.
This convergence led to our core question: What if AI could bridge the gap between where organizations are heading and authentic student value?
What it does
QuillCoach is a strategic positioning engine for scholarship applications that operates in four phases:
Scholarship Intelligence analyzes the organization behind the scholarship—extracting strategic priorities, future direction, and organizational momentum from research and official documents.
Student Profile Analysis helps students identify their authentic strengths, flagging areas where they're minimizing genuine achievements and asking them to verify that any reframing feels honest.
Strategic Essay Generation creates tailored essays that answer the actual prompt while emphasizing the student's strengths most relevant to what the organization actually values (not what it claims to value).
Comparative Analysis shows the before-and-after impact—demonstrating how strategic positioning amplifies an authentic story without fabricating anything.
The ethical core is built into the system: reframe without exaggerating, amplify without inventing, and keep students in control of authenticity boundaries throughout.
How we built it
QuillCoach was built using Hampshire County AI's AI App Ideator, a thinking-first app generator we created to solve a critical problem in AI-assisted development. We noticed that most app generation tools encourage developers to jump straight from rough ideas to code—essentially generating apps from a couple paragraphs. This approach produces technically functional applications that miss the deeper strategic thinking required to build something meaningful. The AI App Ideator inverts this process: spend substantial time articulating vision, values, stakeholder needs, real-world impact, and business logic before generating a single line of code.
We used this methodology to build QuillCoach, which meant the technical specification became the creative crucible where we solved the hardest conceptual problems.
Thinking through the organizational intelligence challenge: Our initial concept assumed we could simply extract strategic priorities from organizational websites and documents. But when we refined the technical specification, we realized this approach had a critical flaw—it would capture what organizations claim to prioritize, not what they actually value. We revised the spec to require the AI to search for and integrate public quotes from organizational leaders, recent initiatives they've championed, and board-level strategic pivots. This distinction between stated values and demonstrated priorities became foundational to QuillCoach's entire intelligence layer.
Solving the student minimization problem: The technical specification initially framed the student profile analysis as simple achievement extraction—just list what they've accomplished. But through iterative refinement, we recognized this would amplify the exact problem we're trying to solve. Students minimize their achievements in how they describe them; simply extracting their self-written descriptions perpetuates that minimization. We revised the specification to require a two-step intelligence process: first, the AI analyzes student inputs to extract hidden value—recognizing patterns of minimization, reframing accomplishments in terms of impact and leadership, identifying strengths the student didn't explicitly articulate. Second, and crucially, we built in an authenticity verification layer where students actively confirm that these reframings feel honest to them. This puts the student back in control and prevents the AI from inflating their profile.
Designing strategic positioning with ethical boundaries: The essay generation specification required careful thought about how to integrate organizational intelligence with student strengths without crossing into manipulation. We spent significant effort refining the prompt structure to ensure the AI understood the hierarchy: the student's authentic verified strengths come first, the organizational priorities come second, and strategic emphasis emerges from the intersection of these two, not from fabricating relevance where it doesn't exist. We explicitly built constraints into the technical specification prohibiting the system from inventing experiences, exaggerating impact, or claiming connections that don't exist.
Building the verification workflow: Rather than assuming AI analysis would be reliable enough to proceed directly to essay generation, we designed a three-phase workflow into the technical specification. The Review phase presents the AI's analysis passively. The Verify phase asks concrete authenticity questions ("Did you actually initiate this initiative?" "Would you confidently discuss this in an interview?"). The Confirm phase requires explicit student commitment before proceeding. This distributed verification approach became more important than any single technical component—it's where the system's ethics actually live.
Once the technical specification precisely captured our vision—with all these refinements embedded—we used the AI App Ideator to generate the application. The generator handles the implementation details: button styling, font selection, layout decisions, component architecture, state management. But these choices aren't arbitrary—they're informed by the deep work we did in the specification. We chose serif headlines because they convey the trust and sophistication that an ethical AI tool requires. Monospace accents signal technical rigor. Bold gradients make strategic positioning feel empowering rather than manipulative. Step indicators reduce cognitive load as students progress through phases. These design choices emerge from our conceptual clarity, not from aesthetic preference.
The entire codebase—both the AI App Ideator and QuillCoach—is publicly available on Hampshire County AI's Github. The AI App Ideator itself is live on Poe at https://poe.com/AI_App_Ideator, so anyone can use it to build their own applications with the same thinking-first methodology, or remix our approach for their own ideas.
This is what distinguishes our process: we didn't generate an app quickly and then iterate. We thought deeply about the problem, refined the technical specification until it captured that thinking precisely, and then generated the application to realize that vision. The AI handled implementation; we handled strategy and ethics. This division of labor—where human thinking precedes and shapes AI generation—is what we believe makes QuillCoach genuinely useful rather than merely functional.
Challenges we ran into
Detecting minimization without overfitting: Teaching AI to distinguish genuine humility from false modesty required building the student-verification layer as the core safeguard. The AI makes tentative identifications; students confirm authenticity.
Organizational research hallucination: When prompting Claude to extract strategic direction, there's real risk of fabricated quotes or invented initiatives. We solved this with strict prompt boundaries that clearly indicate confidence levels and avoid fabrication.
Balancing optimization with ethics: There's a slippery slope from helping students emphasize strengths to helping them manipulate. We built this directly into system design—no fabrication allowed, explicit student confirmation required, transparent explanations of strategic choices, and comfort zone controls.
Multi-context prompt engineering: Integrating organizational intelligence, student strengths, specific essay prompts, and strategic emphasis without losing coherence required layered context injection with clear boundaries between each type of information.
Demonstrating impact at hackathon scale: The system's real value emerges over time seeing scholarship wins. We solved this by building comparative analysis that shows before/after essays with explicit strategic explanations and hypothetical success probability.
Accomplishments that we're proud of
Building a system that treats responsible AI as a core feature, not an afterthought. The authenticity verification system demonstrates sophisticated thinking about ethical boundaries—we don't just prevent the worst failures, we actively help students recognize their own value.
Creating genuine multi-layer intelligence: organizational trajectory analysis combined with authentic value extraction and strategic essay generation is harder to replicate than any single component.
Solving the reframing paradox: we demonstrated that strategic positioning and authentic voice aren't antagonistic—they're complementary. Understanding what an organization truly values lets students choose which authentic aspects to emphasize.
Building explainability into every decision: every strategic choice has a transparent "why" attached, making the system trustworthy rather than a black box.
What we learned
The AI complexity pyramid: Strategic optimization exists in layers—from surface matching through organizational intelligence to authentic value detection to strategic alignment with integrity. Each layer requires different AI techniques and human oversight.
Authenticity safeguards architecture: Preventing false inflation requires bidirectional verification where students confirm reframings, comfort zone management where students set their own boundaries, and distributed checkpoints throughout the workflow rather than a single gate.
The power of understanding actual priorities: When you truly understand what an organization values (not what it claims to value), strategic positioning becomes honest rather than manipulative. You're answering the real question, not the stated one.
Most importantly: The most powerful AI tools don't maximize outcomes at the cost of ethics—they align outcomes with integrity. Students don't need help fabricating themselves. They need help seeing themselves clearly and positioning their authentic selves strategically.
## What's next for QuillCoach
QuillCoach is launching through three simultaneous growth tracks, each building toward sustainable scale.
**Local Launch (Immediate)**: We're embedding QuillCoach into Hampshire County's existing educational infrastructure—the Hampshire County Public Library's STREAM Club AI programming and the Family Support Center's upcoming AI event for teens. Rather than pushing the app to individual users, we're running hands-on consulting sessions where our volunteers guide students through the application using the Poe interface. This approach serves dual purposes: we get real user feedback in a supportive environment to refine the product and technical implementation, and we build a core group of student advocates and community partners who understand the value of what QuillCoach offers. These early adopters become our foundation for broader expansion.
**Regional Expansion (Before December Break)**: We're marketing QuillCoach across DC, Maryland, and Virginia through educational organizations, scholarship networks, and student support programs. The timing is strategic—winter break is when students actually sit down to work on scholarship applications. Regional users will access QuillCoach directly through their own Poe accounts, giving us visibility into adoption patterns, user engagement, and which features drive the most impact. This phase generates the data and case studies we need to demonstrate market demand.
**National Scale (Active Partner Search)**: We're actively seeking a technical co-founder and operations partner to launch QuillCoach as a standalone platform. The core innovation—our thinking-first approach to strategic positioning, the organizational intelligence layer, the authenticity verification workflow—can live in any AI-enabled system. We're offering a validated product with regional traction, open-source code ready to build on, a clear market opportunity in the scholarship space, and a proven methodology (the AI App Ideator) for responsible AI development. The right partner would handle technology infrastructure, API economics, user acquisition, and go-to-market strategy. This is an opportunity to build something that scales nationally while maintaining the ethical principles that make QuillCoach genuinely helpful rather than manipulative.
These tracks run in parallel and reinforce each other. Local success generates evidence of impact. Regional traction demonstrates demand and reveals optimization opportunities. Both become assets in conversations with potential co-founders. Hampshire County AI's core mission remains supporting AI training and development locally, but QuillCoach represents an opportunity to scale that mission—and the thinking-first methodology behind it—far beyond our region.
Log in or sign up for Devpost to join the conversation.