# Skylark Creations — Full Project Reference

> AI-native product studio based in Colorado. Founded by David Kooi. We build human-friendly apps and partner with domain experts to ship new ones fast.

Website: https://skylarkcreations.com
Founder: https://davidkooi.com
Newsletter: https://uncagedminds.substack.com

---

## Magnifica Humanitas — the Pope's letter on AI, in 5 minutes (Featured Explainer)

**Type:** Plain-language explainer / accessibility summary
**Page:** https://skylarkcreations.com/magnifica-humanitas
**Source text:** https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

A free, plain-language guide to *Magnifica Humanitas* ("The Grandeur of Humanity"), Pope Leo XIV's 2026 encyclical — the first papal letter written squarely about artificial intelligence. The encyclical's thesis is not a verdict on the technology but a question about us: in the age of machines, will we stay human? Technology is never neutral; it carries the values of whoever builds it. The real choice is whether we use AI to dominate one another or to rebuild a world where every person still counts.

### The central image: Babel vs. Jerusalem
Leo XIV frames the whole AI moment through two biblical stories. **Babel** — the tower built to "make a name for ourselves" — is power chasing its own greatness, one system that erases difference for efficiency and reduces people to data and performance. **Jerusalem** — Nehemiah rebuilding the broken city wall by wall — is shared responsibility, relationships repaired before stones are laid, difference treated as a resource. His line: "the construction of Babel or the rebuilding of Jerusalem begins within each one of us."

### Seven core ideas
1. **AI imitates intelligence but has no self** — it can outrun us and simulate empathy, but has no body, conscience, or real understanding. Today's AI is "more 'cultivated' than 'built'"; developers create a framework within which the intelligence grows. Human conscience and freedom must stay in charge. (Chapter 3)
2. **Human dignity is unconditional** — the insidious modern idea is that worth must be earned through productivity. Dignity precedes achievement; a system that treats some lives as "less worthy" has already failed. (Chapter 2)
3. **Every answer has a hidden human cost** — the "silent work" of data labeling, model training, and content moderation (often disturbing material), and children extracting rare-earth elements. A new form of slavery; colonialism in a new shape, where data — health records, genetic maps — become "the new 'rare earths' of power." The Church recalls its own delayed condemnation of slavery as a call to vigilance. (Chapter 4)
4. **Don't delegate life-and-death decisions to machines** — AI lowers the threshold for force, shields people from responsibility, and reduces victims to "collateral damage." Calls for a shared international framework and accountable human control. (Chapter 5)
5. **"Disarm" AI rather than reject it** — AI "must be disarmed, welcoming and accessible," freed from monopoly control, made "human-friendly," restored to the plurality of human cultures. Breaking the "monopolies of AI." (Chapter 3)
6. **Human limits are not defects to patch** — a critique of transhumanism. Once the human is something "to be perfected or surpassed," it gets easier to treat some lives as "less worthy." We grow through our limits; the real "more than human" comes from grace, not a better machine. (Chapter 3)
7. **Truth and work are public goods** — AI super-charges disinformation and threatens minors online; he calls for an "ecology of communication." Automation should serve people, not only cut costs; businesses should treat "work and dignity as measures of success," not GDP alone. (Chapter 4)

### Key verbatim quotes
- "Never has humanity had such power over itself." (Introduction)
- "Let us disarm words and we will help to disarm the world." (Chapter 5)
- Progress as "having more" without "being more." (Chapter 3)
- "The construction of Babel or the rebuilding of Jerusalem begins within each one of us." (Chapter 3)

This is an independent accessibility summary, not an official Church text — all positions are paraphrased and attributed to the encyclical, with quoted lines verbatim. For authoritative wording, see the Vatican source.

---

## Decomposer

**Status:** Production MVP (v1.0)
**Platforms:** Web — https://decomposer.io/
**Project page:** https://skylarkcreations.com/projects/decomposer

Transform messy goals into clear, executable plans with honest effort routing (👤 You / 🤖 AI / 🧰 Expert), quality evaluation, and portable sub-prompts you can execute step by step. It's the planning front-door for the AI era.

### Overview
Decomposer is an AI-powered planning tool that takes your sprawling, multi-part goals and transforms them into structured, actionable plans. Whether you're launching a business, learning a new skill, researching investments, or tackling a home project, Decomposer breaks down your goal into 8–14 concrete steps with honest labels that tell you what you should do yourself, what AI can handle, and when you need expert help.

The tool extracts hidden assumptions, evaluates quality across five dimensions, surfaces cross-cutting risks, and gives you a structured plan you can actually follow—complete with copy-ready sub-prompts for each task. Unlike generic chat interfaces that produce paragraphs that fall apart when you try to act on them, Decomposer creates persistent, shareable plans with measured quality gates.

### Why
People bring sprawling, multi-part problems to AI and get responses that feel smart but fall apart during execution. Plans miss steps, repeat themselves, or assume magic helpers exist. Most people don't think in no-overlap/no-gaps trees, nor should they need to learn prompt-engineering to get a usable plan. Decomposer is the planning layer that makes problems legible, routes work honestly, and gives you portable sub-prompts to execute step by step—it's the front door to AI that ensures you start with structure, not soup.

### How It Works
1. **Describe your goal** – Paste your objective, add context, or upload documents (PDF, Word, images, text files). The system forms an "observed brief" and proposes assumptions and constraints you can edit.

2. **AI generates a structured plan** – Using GPT-5, Decomposer creates 8–14 task cards with titles, specific bullet points, honest effort labels, dependencies, and paste-ready sub-prompts for each step.

3. **Quality evaluation** – A built-in evaluator scores your plan across five dimensions: coverage, redundancy, actionability, label routing, and synthesis fit. Low scores trigger targeted regeneration.

4. **Track and refine** – Edit cards inline, track status (Not started → In progress → Done), capture outputs from completed tasks, and watch dependent tasks automatically refine with new information.

5. **Execute your way** – Copy sub-prompts into ChatGPT, Claude, or your preferred AI tool. Export to PDF or Markdown. Share persistent links with teammates.

### Features
- **Template-Free Decomposition** – Works across any domain without rigid frameworks
- **Honest Routing** – Clear labels showing what requires You (👤) vs. AI (🤖) vs. Expert (🧰)
- **Quality Evaluation** – Five-dimension scoring (coverage, redundancy, actionability, label routing, synthesis fit)
- **Assumptions Extraction** – Surfaces hidden constraints and makes plan edges explicit
- **Plan Synthesis** – Coherent summary with risk identification
- **Granularity Coaching** – Flags tasks that are too broad or too small
- **Document Upload** – PDF, Word (.docx), images, text files, and Markdown with AI-powered extraction
- **Dependency Flow System** – Living documents that improve as work progresses; outputs propagate to dependent tasks
- **Guided Execution Modes** – Autonomous/Guided/Manual classification with domain-aware artifacts
- **Mobile-First Design** – Fully responsive, touch-optimized interface
- **Plan Management** – Inline editing, status tracking, plan history with auto-save
- **PDF Export** – Professional formatting with status indicators
- **Plan Sharing** – Compressed storage with persistent shareable links
- **Copy as Markdown** – Portable export format for any tool
- **User Accounts** – Email/password or Google OAuth via Supabase
- **Pro Tier** – $10/month for unlimited plans, priority generation, PDF export

### Notes
- **Current Status:** Production MVP (v1.0) — Full web application with authentication, billing, document upload, mobile-first design, and dependency tracking
- **Tech Stack:** Next.js 15.5, TypeScript, React 19, Tailwind CSS v4, OpenAI GPT-5 family, Supabase (auth + Postgres), Stripe, Redis (Render KV), PostHog, Vitest + Playwright
- **Pricing:** Free tier (3 plans), Pro $10/month (unlimited plans, PDF export, priority generation)
- **Model-Agnostic:** Thin adapter architecture allows swapping AI models as quality and cost evolve
- **No Lock-In:** Everything exports cleanly with copy-ready formats
- **Privacy-First:** Plans are private by default; we don't train on your data
- **Works with Any AI Tool:** Generated prompts work with ChatGPT, Claude, Gemini, or local models

---

## AniMoment

**Status:** Live
**Platforms:** Web — https://animoment.app
**Project page:** https://skylarkcreations.com/projects/animoment

Take a photo and receive a gentle, spoken animist reading of place. AniMoment honors the relational essence of everything—lake, tree, rock, soil—as different expressions of Earth.

### Overview
AniMoment is a web application that transforms how we relate to the places we inhabit. Take a photo from your camera or photo library, and receive a gentle, spoken animist reading that honors the relational essence of everything around you—lake, tree, rock, soil, path, sky. Each reading is generated using AI vision technology, then spoken aloud, creating a moment of calm, grounded presence.

The app offers four distinct reading styles: Field Guide (structured and observational), Five Neighbors (five short vignettes of nearby beings), Elder Voice (lyrical and contemplative), and Direct Animist (conversational and grounded). Each reading concludes with a tiny 10-30 second practice to help you pause and notice what's here. Your last 20 moments are stored locally on your device, creating a personal archive of presence.

AniMoment is for anyone seeking a calmer, more relational way of being in place. No dogma, no cultural borrowing—just warm acknowledgment of the world as living matter expressing itself in infinite forms. Whether you're walking in nature, sitting in a park, or pausing in your daily routine, AniMoment invites you to greet the beings around you with presence and care.

### Why
In a world of constant noise and abstraction, we often move through places without truly noticing them. We see "scenery" or "environment" rather than recognizing that everything around us—water, stone, path, soil, tree—has presence and character. Ancient animist traditions understood this, but we wanted to create something contemporary and accessible, without appropriating specific cultural practices.

AniMoment emerged from a simple question: What if technology could help us slow down and pay attention? What if AI could serve not just productivity, but presence? By combining computer vision with carefully crafted prompts, we created a tool that gently guides users to see their surroundings as relational beings worthy of acknowledgment. It's not about belief or spirituality—it's about practicing a warmer, more grounded way of being in the world.

### How It Works
1. **Capture or Upload**: Open AniMoment in your browser and take a photo with your camera or upload one from your photo library. The image is automatically compressed client-side for privacy and efficiency.

2. **Choose Your Style**: Select from four reading variants—Field Guide (structured), Five Neighbors (vignettes), Elder Voice (lyrical), or Direct Animist (conversational). Each offers a different voice and approach.

3. **AI Interpretation**: The image is sent to GPT-4o-mini with vision, which analyzes what's present in the scene. Our carefully crafted prompts guide the AI to generate a warm, concrete reading that acknowledges specific beings (path, water, sky, stone, plants) with light personification.

4. **Spoken Aloud**: The reading is converted to speech using either your device's Web Speech API (free, on-device) or optional ElevenLabs voices (premium). The audio plays automatically while you can read along.

5. **Save & Share**: Your moment is automatically saved to IndexedDB (last 20 moments), and you can share it via your device's native share sheet (requires HTTPS on mobile).

The entire experience takes 5-10 seconds from photo to spoken reading, creating a gentle ritual of pause and presence wherever you are.

### Features
- 📸 **Camera & Photo Upload** - Capture moments directly or upload from your library
- 🎭 **Four Reading Styles** - Choose Field Guide, Five Neighbors, Elder Voice, or Direct Animist
- 🔊 **Spoken Aloud** - Readings are automatically converted to speech for listening
- 💾 **Local Storage** - Last 20 moments saved privately in your browser (IndexedDB)
- 🌐 **Web Share API** - Share moments via your device's native share sheet
- 🌙 **Wake Lock** - Screen stays awake during audio playback
- 🎨 **Beautiful, Calm UI** - Earth tones, gentle gradients, and clear typography
- ⚡ **Fast Generation** - 5-10 seconds from photo to spoken reading
- 🔒 **Privacy-First** - Images compressed client-side; optional server TTS
- 🆓 **Free to Use** - Minimal API costs (~$0.001-0.003 per reading)

### Notes
- **Tech Stack**: Built with Vite, TypeScript, Node.js, OpenAI GPT-4o-mini (vision), ElevenLabs (optional), Web Speech API, IndexedDB, and Web Share API
- **For Whom**: Seekers, walkers, contemplatives, nature lovers, mindfulness practitioners, and anyone curious about relating more warmly to their surroundings
- **Approach**: No dogma, no cultural borrowing—just calm acknowledgment of place as living presence
- **Cost**: ~$0.001-0.003 per reading with GPT-4o-mini; Web Speech API is free (on-device)
- **Privacy**: Images processed via OpenAI's API; no permanent storage server-side; all moments stored locally in your browser
- **Roadmap**: Place memory, time-aware readings, gentle ambient audio, "Dear Neighbor" social feed (opt-in), iOS native app, multi-language support

---

## Agentic News

**Status:** Live
**Platforms:** Web — https://agentic-news.ai
**Project page:** https://skylarkcreations.com/projects/agentic-ai

AI-powered semantic alerts for news you care about. Agentic News deduplicates noisy coverage into story clusters, then explains what changed and why it matters through your chosen lenses—with citations.

### Overview
Agentic News is an AI-powered semantic alerts and interpretation platform that watches the web for what you care about. Instead of keyword spam and duplicate headlines, it clusters near-identical coverage into clean story groups with multiple sources.

For each cluster, Agentic generates AI "Lenses" that explain what changed and why it matters from perspectives you choose (e.g., CEO, engineer, investor, parent). You can save clusters into collections and ask AI questions across your saved knowledge—answers include citations back to the underlying sources.

### Why
Generic news feeds and keyword alerts are noisy and don't explain implications. Agentic News is built to deliver signal over noise: fewer alerts, deduped coverage, and fast, role-aware interpretation with receipts.

### How It Works
1) Create Agents for topics/entities you care about
2) Deduplicate coverage into Clusters
3) Interpret clusters with Lenses (role/perspective-specific)
4) Save to Collections + ask AI Q&A with citations
5) Optional email briefings (daily; adjustable frequency)

### Features
- Semantic agents (watchlists) for topics, competitors, niche interests
- Deduplicated clusters to reduce repeated coverage
- Role-aware lenses for "what changed" and "why it matters"
- Collections to save and organize important clusters
- AI Q&A across saved clusters with cited answers
- Email briefings/digests (frequency configurable)

### Notes
- **Web App Only:** No native iOS/Android app; web app only
- **Audience:** Competitive intelligence, research, investing, tech tracking, niche interests
- **Disclaimers:** Links to sources; does not republish full paywalled text
- **Tech Stack (high-level):** Next.js 15 + Tailwind + shadcn/ui; Node (Fastify) API + worker; Postgres (Neon) + pgvector + Drizzle; Redis; OpenAI embeddings + GPT-5.x; Render + Sentry + Amplitude

---

## BloomEDU

**Status:** Available on iOS & Android
**Platforms:** iOS — https://apps.apple.com/us/app/bloomedu-learn-plants/id6749336509 | Android — https://play.google.com/store/apps/details?id=com.bloomedu.app
**Project page:** https://skylarkcreations.com/projects/bloomedu

A focused learning app for mastering Colorado wildflowers. Photo‑rich lessons, habitat/elevation filters, and smart, bite‑sized quizzes help you tell look‑alikes apart and track progress from learning → familiar → mastered.

### Overview
BloomEDU is a focused learning app for mastering the wildflowers of Colorado, from prairie and foothills to alpine tundra. Photo‑rich lessons, habitat/elevation filters, and smart, bite‑sized quizzes help you recognize look‑alikes, remember key traits (leaf shape, bloom timing, inflorescence), and track progress with a simple "learning → familiar → mastered" flow. Built as a modular field curriculum, BloomEDU will expand to additional regions so your library grows as your adventures do. Available now on iOS.

### Features
- Photo‑rich lessons organized by region and habitat
- Habitat & elevation filters for field‑relevant study
- Bite‑sized quizzes that target look‑alikes and key traits
- Simple progress states: learning → familiar → mastered
- Modular curriculum designed to expand to new regions

---

## Cursor Environment Docs

**Status:** Available on GitHub
**Platforms:** Web — https://github.com/u00dxk2/cursor-kooi-env-docs
**Project page:** https://skylarkcreations.com/projects/cursor-environment-docs

Auto-updating environment documentation that keeps Cursor smart about your project. Stop wasting time explaining your environment setup in every conversation.

### Overview
Cursor Environment Docs is an open-source documentation system specifically designed for developers using Cursor AI. It solves the frustrating problem of repeatedly explaining your development environment setup to AI assistants in every new conversation. The system installs a smart documentation framework in your project's .cursor/ directory that AI automatically reads at session start, checks for staleness, and updates in real-time as you develop.

With a simple one-line installation, you get auto-updating documentation that captures your directory structure, shell environment, common commands, tech stack, and—most importantly—environment gotchas and mistakes so you never hit them twice. It's framework-agnostic, git-friendly, and works across Windows, macOS, and Linux.

Unlike static documentation that quickly becomes outdated, this system actively maintains itself. Every 7 days (or whenever the AI detects changes), it prompts for a review and update, ensuring your documentation stays synchronized with your actual development environment.

### Why
This project was born from real frustration with AI pair programming workflows in Cursor. As a developer working with AI assistants, you constantly encounter the same problems: wasting time explaining your shell syntax (PowerShell vs Bash), hitting the same environment gotchas repeatedly, having AI suggest commands that don't work in your environment, and dealing with inconsistent behavior between different AI conversations.

Traditional documentation sits in a README that gets stale within weeks. You need living documentation that AI assistants can automatically consume and maintain. This system bridges that gap—giving AI the environmental context it needs while requiring minimal maintenance from you.

### How It Works
The system works through Cursor's native ability to read the .cursor/ directory. When installed, it creates a structured template and AI behavior rules that guide documentation creation and maintenance. At the start of every Cursor conversation, the AI automatically reads your project-environment.md file, checks the "Last Updated" timestamp, and offers to update if it's been more than 7 days.

During development, the AI updates documentation in real-time when it discovers changes—new dependencies, command syntax issues, environment quirks, or tech stack modifications. You also get platform-specific staleness checker scripts you can run manually anytime. The entire .cursor/ directory is designed to be committed to version control, so your whole team benefits from consistent, shared environmental context.

### Features
- 🤖 Auto-Updating Documentation - AI checks every conversation and updates in real-time as you develop
- 📝 Self-Documenting System - Automatically captures gotchas and mistakes so you never repeat them
- 🔄 Cross-Session Consistency - Works seamlessly across all Cursor conversations
- ⚡ 5-Minute Setup - One-line installation, minimal configuration required
- 🔍 Staleness Checker - Know exactly when your docs need attention
- 🎯 Framework Agnostic - Works with any tech stack (Node.js, Python, Rust, Go, etc.)
- 💾 Git-Friendly - Version-controlled documentation your whole team can use
- 🖥️ Cross-Platform - Full support for Windows, macOS, and Linux
- 🤝 Multi-Platform Teams - Documents all team members' platforms in one file
- 📚 Complete Examples - Includes working examples for Node.js/Express and Python/Flask

### Notes
- Note: This is a developer tool specifically designed for the Cursor AI IDE. It leverages Cursor's ability to automatically read .cursor/ directories and apply maintenance rules.
- Open Source: Released under MIT License on GitHub
- Active Development: Version 1.0.0 released October 2025, with roadmap including NPM package, VSCode extension, and additional framework examples
- Team-Friendly: The entire .cursor/ directory should be committed to git—it's shared project documentation, not personal IDE settings

---

## BuddhaUR

**Status:** Available on iOS & Android
**Platforms:** iOS — https://apps.apple.com/us/app/buddhaur-find-peace/id6752915041 | Android — https://play.google.com/store/apps/details?id=com.skylark.buddhaur | Web — https://buddha-ur.com/
**Project page:** https://skylarkcreations.com/projects/buddhaur

Learn early Buddhism through conversation. Ask what you're actually dealing with—anger, craving, stress, attention, relationships—and get grounded guidance with tappable citations to the early suttas, so you can verify it for yourself.

### Overview
BuddhaUR is a conversational teacher in your pocket. Ask real questions about life, meditation, ethics, and Buddhist doctrine—anger, craving, stress, attention, relationships—and receive thoughtful answers in a warm, patient, Socratic tone. Every substantial claim includes citations you can tap to verify against early Buddhist sources.

The app is built around a repeatable rhythm: Learn a topic through guided lessons, ask follow-ups in chat, reinforce with quizzes or adaptive practice, sit with a calm meditation timer and notes, then track your streaks and return tomorrow. It's a practice ecosystem where learning becomes interactive, practice becomes repeatable, and the app's memory makes guidance feel continuous across sessions.

BuddhaUR is rooted in early Buddhist teachings (suttas + core frameworks), not generic self-help platitudes. The AI is constrained by retrieval of source material and is designed as a teacher, not a random answer generator. Whether you're a curious beginner wanting a friendly entry point, a meditator seeking consistent practice support, or a practitioner wanting doctrinal depth with warmth—BuddhaUR meets you where you are.

### Why
People bring sprawling questions to Buddhism and often get responses that feel smart but fall apart during practice. Most people don't know where to start in the Pali Canon's thousands of suttas, and modern meditation apps often present "mindfulness lite" disconnected from its philosophical foundations.

BuddhaUR lets you ask your question, not just read the answer. You can verify teachings via citations to primary sources. You build a habit through streaks, daily practice, and gentle reminders of progress. The app learns your journey and adapts its guidance and suggestions over time—creating a truly personalized learning experience that remembers what matters.

### How It Works
**The Teach → Practice → Reinforce Loop:**
1. **Learn** a topic through structured lessons on the Buddha's core frameworks—Four Noble Truths, Eightfold Path, Five Aggregates, and more
2. **Chat** with the teacher using Socratic dialogue—request stories, metaphors, or practical applications
3. **Practice** with quizzes or adaptive practice that targets your knowledge gaps
4. **Meditate** using the calm timer with reflection notes and "chat about this session" to connect practice to insight
5. **Track** your streaks and return daily via personalized Daily Dhamma talks

**How the AI Works:**
When you ask a question, BuddhaUR retrieves relevant passages from its knowledge base of early Buddhist texts and answers using that context (RAG—retrieval-augmented generation). Citations link out to SuttaCentral (translations by Bhikkhu Sujato), so users can check the sources. The point is simple: you can verify everything.

### Features
- **A Teacher You Can Talk To** - Ask anything and get answers designed to be warm, patient, and practical—not a generic chatbot
- **Citations You Can Tap** - Answers grounded in early sources with tappable links to verify and read more
- **Guided Lessons on 16 Core Topics** - Walk through the Four Noble Truths, Eightfold Path, Five Aggregates, Dependent Origination, and more
- **Quizzes + Adaptive Practice** - Turn "I read it" into "I know it" with practice that brings back what you miss
- **Meditation Timer + Reflection Notes** - A clean timer that respects attention, with notes and "chat about this session"
- **Daily Dhamma Talk** - Personalized daily teachings drawn from your interests, lessons, and recent conversations
- **Story Mode** - Request Dhamma stories told in a narrative, warm tone grounded in themes and sources
- **Smart Follow-Up Suggestions** - Tappable chips for deeper dives, related concepts, and practical applications
- **Personalized Experience** - The app remembers your journey, preferences, and challenges to tailor suggestions
- **Conversation History** - Review and resume past discussions; your path has a memory
- **Daily Streaks & Achievements** - Track consistency across lessons, quizzes, meditation, and talks
- **Beautiful, Calm Design** - Sage green accent, warm earth tones, generous whitespace, high readability

### Notes
- **Free Plan:** 1 lesson/day, 1 quiz/day, 1 Daily Dhamma talk/day (up to 5 min), 5 chat messages/day, unlimited meditation timer + meditation lessons, conversation history, streaks/stats
- **Premium:** $1.99/month USD — Unlimited lessons, quizzes, chat, adaptive practice, and longer Daily Dhamma talks. Cancel anytime.
- **For Whom:** Curious beginners, daily meditators, experienced practitioners, busy people with real-life problems, anyone who wants perspective without the pedestal
- **Not Medical Advice:** Educational and contemplative content only. Meditation can be challenging—practice with care.
- **Privacy:** No ads, we don't sell your data. Account deletion available in-app (Settings → Account → Delete Account).
- **Philosophy:** No gurus, no mystical drift—just authentic early Buddhist teachings you can verify for yourself.

---

## QuickBite

**Status:** Live Prototype (In Active Development)
**Platforms:** Web — https://quickbite.food/
**Project page:** https://skylarkcreations.com/projects/quickbite

Turn a 30-minute "where should we eat?" debate into a 2-minute decision. QuickBite helps groups quickly agree on restaurants through fast, fair voting with transparent consensus scoring.

### Overview
QuickBite is a group decision-making app designed to solve one of life's most frustrating micro-debates: "Where should we eat?" Whether you're out with friends, planning a family dinner, or organizing team lunch, the endless back-and-forth of "I don't know, what do you want?" can waste 30+ minutes of everyone's time.

With QuickBite, the process takes less than 2 minutes. The host creates a session by choosing a location and setting preferences (cuisine types, dietary restrictions, search radius). QuickBite then curates 10 restaurant candidates using Google Places and Yelp data, filtered intelligently by cuisine, open hours, distance, and dietary needs. Each participant votes on all 10 restaurants with a simple three-option system: Yes (strong preference), Maybe (acceptable), or No (rather not). No lengthy deliberation—just quick, honest preferences.

Once everyone votes, QuickBite instantly calculates a fair winner using a transparent scoring algorithm. The system considers vote preferences, distance penalties, price matching, and open hours to find the restaurant that best satisfies the whole group. Results show the top 3 options with detailed scoring breakdowns, so everyone understands why the winner was chosen. No authentication required, works on any device, and provides instant links to Google Maps for directions.

### Why
We've all been there: standing on a street corner with friends, or sitting in a Slack channel with coworkers, throwing out restaurant suggestions that get lukewarm responses. "How about Thai?" "Eh, I just had Thai yesterday." "Mexican?" "Too far." "That new burger place?" "Too expensive." The conversation spirals, people get hungry and cranky, and eventually someone just makes an executive decision that leaves others unsatisfied.

This isn't a technology problem—it's a coordination problem. Groups struggle to aggregate individual preferences quickly and fairly. Existing solutions are either too slow (lengthy text discussions), too random (spinning a wheel feels arbitrary), or too complicated (spreadsheets, polls with unclear methodology). We built QuickBite to provide speed, fairness, and transparency: everyone's voice counts, the math is clear, and you get from "where should we eat?" to sitting down at a table in minutes instead of half an hour.

### How It Works
**Creating a Session:**
The host starts by setting a search location (current location or entered address) and configuring preferences: search radius (1-50 miles), cuisine types (30+ options like Thai, Italian, Mexican, Sushi), facility types (restaurants, cafes, food trucks, bakeries), dietary restrictions (vegetarian, vegan, gluten-free, halal, kosher, etc.), and time slot ("now" or "tonight"). QuickBite queries both Google Places and Yelp APIs in parallel to find up to 40 restaurants, then intelligently filters and scores them based on open hours, distance, cuisine matches, and dietary considerations. The system selects the best 10 candidates and stores them for the voting session.

**Voting:**
The host shares a simple link with the group (via text, email, Slack, etc.). Participants join by entering a nickname—no signup, no authentication. Everyone votes on the same 10 restaurants using a fast card-swipe interface. For each restaurant, you see a photo, name, cuisine types, rating, distance, price level, address, phone, and hours. You choose: 👍 Yes (I'd love to go here, +2 points), 🤷 Maybe (It's fine, +1 point), or 👎 No (I'd rather not, 0 points). The whole process takes about 90 seconds per person.

**Results:**
Once everyone completes voting (or 70% of participants finish), the system calculates the winner using a fair, transparent algorithm: base score from votes, minus distance penalties (scaled by search radius), price mismatch penalties (if the group set a budget preference), and closed-restaurant penalties. The results page shows the top 3 restaurants with full scoring breakdowns, so everyone can see why the winner was chosen. Participants can open the winner directly in Google Maps and get directions. Late joiners can still vote, and their preferences are instantly reflected in the results.

### Features
- **Location-based Discovery**: Use your current location or enter any address to find nearby restaurants within 1-50 miles
- **Smart Filtering**: 30+ cuisine types, 8 facility types (restaurants, cafes, food trucks, etc.), 10 dietary restriction tags, and time-slot optimization ("now" vs "tonight")
- **Dual-Source Data**: Combines Google Places and Yelp to maximize restaurant coverage and data quality
- **Hours-Aware**: Automatically filters out closed restaurants and those closing soon; includes places opening within 30 minutes
- **Fast Group Voting**: No authentication required—just enter a nickname and start voting in seconds
- **Three-Option Voting**: 👍 Yes (+2), 🤷 Maybe (+1), 👎 No (0) for quick, intuitive preference expression
- **Fair Consensus Algorithm**: Transparent scoring with distance penalties, price matching, and detailed breakdowns so everyone sees why the winner was chosen
- **Mobile-Responsive**: Works seamlessly on phones, tablets, and desktops with touch-optimized voting interface
- **Instant Results**: Top 3 ranked restaurants with scoring explanations and direct Google Maps integration
- **Real-Time Participation**: See who's joined, who's voting, and who's completed in real-time
- **Zero Friction**: No app downloads, no signups, no passwords—just share a link and go

### Notes
- **Current Status:** Web prototype actively used internally; next phase includes Google/Apple sign-in, saved groups, two-round cuisine voting, and session timers
- **Tech Stack:** Built with Next.js 14, TypeScript, React, Tailwind CSS, PostgreSQL, Prisma ORM, and integrates Google Places API (New) and Yelp Fusion API
- **Privacy:** No long-term user accounts required; sessions are ephemeral and auto-expire after 24 hours
- **For whom:** Friend groups, families, coworkers, remote teams, anyone tired of the "where should we eat?" debate
- **Scoring Philosophy:** Transparent > Random. Every penalty and tie-breaker is explained so groups trust the outcome
- **Roadmap Highlights:** Auto-tally at 70% participation, better duplicate detection, host controls, re-roll alternatives, ordering integration (DoorDash/Uber Eats), optional user accounts with session history

---

## Quiet Tab

**Status:** Available
**Platforms:** Web — https://quiet-tab.com
**Project page:** https://skylarkcreations.com/projects/quiet-tab

Hide Gmail's unread count in your tab title so you can focus and check email on your terms.

### Overview
Quiet Tab is a tiny Chrome extension that removes the distracting unread email count (like "(4)") from Gmail's browser tab title. It's built for anyone who keeps Gmail open while working and wants fewer attention triggers and less background anxiety.

It runs only on mail.google.com, requires zero special permissions, and doesn't collect any data.

### Why
Gmail's unread badge in the tab title is a persistent attention trigger. Quiet Tab removes that trigger while keeping Gmail open and accessible when you choose to check it.

### How It Works
1. **Install the extension** (Chrome Web Store listing pending)
2. **Open Gmail as usual**
3. **Quiet Tab watches for title changes** as Gmail updates unread counts
4. **It removes the "(number)" pattern locally** and updates the tab title
5. **Gmail stays open, but the nagging count disappears**

### Features
- Removes the unread count from Gmail tab titles automatically
- Zero special permissions; runs only on `https://mail.google.com/*`
- No tracking, no analytics, no data collection
- Open source and inspectable

### Notes
- **Category:** Browser extension · Productivity / Focus
- **Platform:** Chrome Extension (Manifest V3)
- **Pricing:** Free (open source)
- **Permissions:** None (`"permissions": []`); content script scoped to Gmail only
- **Data:** Does not collect, store, transmit, or process personal data
- **Attribution:** Skylark Creations — product concept, design, development, and landing page

---

## Sagent

**Status:** Available on iOS
**Platforms:** iOS — https://apps.apple.com/us/app/sagent-guidance-on-demand/id6747406673
**Project page:** https://skylarkcreations.com/projects/sagent

Sagent lets you consult timeless minds—Buddha, Epictetus, Laozi, Jesus, Carl Sagan, Mark Twain, Rumi, and Kurt Vonnegut—about modern dilemmas. Ask a question and get guidance in the voice and spirit of your chosen mentor.

### Overview
Sagent (short for "sage agent") is a philosophical chatbot designed for modern life. Choose a guide—Buddha, Epictetus, Laozi, Jesus, Carl Sagan, Mark Twain, Rumi, or Kurt Vonnegut—and get advice that reflects their worldview and style. Whether you're seeking perspective, clarity, or a smart laugh, Sagent meets you where you are.

### Why
Most of us wish we had a wise mentor on call. Sagent makes that feeling accessible—anytime, anywhere—by channeling enduring ideas through approachable conversation.

### How It Works
We pair advanced language models with persona designs informed by primary texts and trusted summaries. Each philosopher's "voice" is carefully guided so the counsel feels timely without losing what makes the original thinker distinctive.

### Features
- Eight distinct guides: Buddha • Epictetus • Laozi • Jesus • Carl Sagan • Mark Twain • Rumi • Kurt Vonnegut
- Voice‑true counsel: answers shaped to each figure's tone and core ideas
- Perspective switching: see your question reframed by different traditions
- Modern relevance: ancient insight translated to practical next steps
- Light when possible, deep when needed: from gentle re‑centering to bracing Stoic tough love—and the occasional Twain zinger

### Notes
- For whom: founders, students, seekers, writers, overthinkers, under‑sleepers—anyone who wants perspective without the pedestal.
- AI‑generated creative simulations inspired by historical figures—not direct quotes or endorsements—and not professional advice.

---

## Dreamflyer

**Status:** Concept Stage (Early development)
**Platforms:** iOS — Planned | Android — Planned
**Project page:** https://skylarkcreations.com/projects/dreamflyer

A first‑person, meditative flying game that recreates the feeling of dream‑flight using a gentle breaststroke motion.

### Overview
Dreamflyer is about that weightless, lucid sensation so many of us know from sleep—gliding over rooftops, skimming treetops, drifting higher with calm, rhythmic strokes. Instead of adrenaline and speed, Dreamflyer aims for presence, ease, and flow—part flight sim, part moving meditation.

### Why
We wanted to bottle the specific way some dreams feel: you "swim" through the air with slow, smooth breaststroke pulls, and the world responds with effortless lift.

### How It Works
On mobile, arm sweeps (or small wrist motions) are tracked by device sensors to translate stroke cadence and amplitude into forward glide and gentle climbs. Subtle phone tilts guide heading and pitch. On the web, accessible controls mimic the same stroke‑to‑motion feel. Physics are tuned for dream logic—stable horizons, soft banking, forgiving stalls—so you can relax into the glide without motion sickness.

### Features
- Optional breath‑to‑lift (mic‑based breathing sync)
- Guided "flow flights" with ambient soundscapes
- Free‑flight exploration over stylized landscapes
- Comfort settings: vignette, horizon lock, etc.
- Gentle challenges: thermals, rings, cloud currents

### Notes
- Goal: a few minutes of calm, creative escape that genuinely feels like flying—one easy breaststroke at a time.

---

## Frolic

**Status:** In Development
**Platforms:** Web (in development) — https://frolic-web.onrender.com/
**Project page:** https://skylarkcreations.com/projects/frolic

AI-powered outdoor activity and event recommendations. Frolic combines scheduled Deep Research agents, multi-source event data, and a personalization engine to surface things to do nearby — tailored to your preferences, budget, and context.

### Overview
Frolic is a full-stack application that helps you discover outdoor activities and events tailored to your preferences. The platform combines AI-powered recommendations, automated event curation, and real-time data from multiple sources to deliver personalized experiences.

A scheduled batch processing pipeline uses Deep Research agents to build a high-quality event inventory — daily and weekly runs for each supported city. Events are ingested from Google Places, Google Events, and Eventbrite, then normalized, geocoded, validated, deduplicated, and stored in MongoDB for fast geospatial and temporal queries.

The AI recommendation engine considers your preferences and history (via a Memory Service that learns over time), current context (time, location, companions, budget), deterministic filters (distance, price, accessibility), and a scoring algorithm that balances distance, popularity, category match, and freshness.

### Why
Most discovery apps show the same top-10 lists and star ratings. Frolic is built to bring back serendipity — surfacing offbeat, delightful things to do right now, personalized to who you are and what you're in the mood for. The cache-first architecture means recommendations are instant, not waiting on slow API calls.

### How It Works
1. **Tell Frolic what you're looking for** — or let context (time, location, companions, budget) guide the recommendations
2. **AI curates and ranks** — OpenAI scores and reranks events from the cached inventory based on your preferences and context
3. **Smart filtering** — Budget, distance, time slot, and companion-based filters narrow results to what's actually doable
4. **Personalization improves over time** — The Memory Service tracks your activity preferences, budget tendencies, location history, and feedback to refine future recommendations

### Features
- **AI-Powered Recommendations** — OpenAI-driven personalization based on user preferences and context
- **Automated Event Curation** — Scheduled Deep Research agents precompute city-scoped event inventories daily and weekly
- **Multi-Source Integration** — Combines Google Places, Google Events, and Eventbrite for comprehensive coverage
- **Cache-First Architecture** — MongoDB-backed for instant responses; events are pre-indexed for geospatial and temporal queries
- **Smart Filtering** — Budget, distance, time, and companion-based filtering
- **User Memory & Personalization** — Learns activity preferences, budget tendencies, and feedback over time
- **Real-Time SSE Streaming** — Recommendations stream to the client as they're generated

### Notes
- **Tech Stack:** Node.js + Express, MongoDB + Mongoose, OpenAI GPT-4, React Native (Expo), JWT + Apple Sign-In, Server-Sent Events, node-cron
- **Data Sources:** Deep Research Agent (vendored fork), Google Places API, Google Events (RapidAPI), Eventbrite API
- **Current Scope:** Denver metro area; multi-city support in progress
- **For Whom:** Anyone looking for something to do — solo, couples, families — who wants personalized suggestions beyond generic top-10 lists

---

## Urban Beaver

**Status:** In Development (Sprint 2 - Playable prototype)
**Platforms:** iOS (in development) | Android — Planned
**Project page:** https://skylarkcreations.com/projects/urban-beaver

A relaxing 2D puzzle game where you build beaver dams to stop water and fill ponds. Drag logs and twigs, watch water accumulate, and create your perfect pond sanctuary.

### Overview
Urban Beaver is a relaxing 2D puzzle game where you become nature's engineer. Using drag-and-drop controls, you place logs and twigs to build dams that stop flowing water and create serene beaver ponds. Watch as individual water particles collide with your structures, accumulate behind your dam, and gradually fill the pond to your target level. It's a meditative experience that combines physics-based puzzles with the satisfaction of creating something beautiful.

Designed for anyone seeking a calm, focused puzzle experience. Whether you're decompressing after a long day, looking for a mindful mobile game, or just love nature-inspired challenges, Urban Beaver offers a unique blend of strategy and serenity. The game rewards thoughtful placement and experimentation—every gap in your dam matters, and every successful pond feels earned.

Built from the ground up with Swift and SpriteKit, Urban Beaver features realistic physics simulation, smooth touch controls including two-finger rotation, and a growing variety of building materials. The landscape-only orientation maximizes your workspace, giving you the full riverbed canvas to express your inner architect.

### Why
Most mobile puzzle games rush you with timers, lives, and constant pressure. We wanted to create something different—a game that respects your pace and rewards thoughtful experimentation. The idea came from watching real beavers work: they're patient, methodical, and create something genuinely impressive through incremental effort. That felt like the perfect metaphor for a puzzle game that's actually relaxing.

We also wanted to explore physics-based gameplay in a way that feels intuitive and satisfying. Every water particle, every collision, every leak in your dam is simulated in real-time. When your dam holds and the pond fills, it's not because the game scripted it that way—it's because your structure actually works. That cause-and-effect relationship creates a deeply satisfying gameplay loop that never gets old.

### How It Works
**Gameplay Flow:**

1. **Place Pieces:** Drag logs and twigs from your supply area and drop them anywhere in the riverbed. Use two-finger rotation to angle pieces perfectly.

2. **Stop the Flow:** Water particles continuously flow from left to right. When they hit your dam pieces, they splash and accumulate behind the structure.

3. **Build the Pond:** As water builds up behind your dam, a pond visual grows upward. Your goal is to reach and hold a target fill line for a few seconds.

4. **Mind the Gaps:** If your dam has gaps, water will leak through at those specific points. You'll see leak particles spawn and your pond level may drop. Plug the gaps with more pieces to maximize water retention.

5. **Win or Adapt:** Successfully hold the target water level to complete the level. If time runs out before you reach the goal, you can restart and try a new strategy.

**Technical Approach:**

Urban Beaver uses SpriteKit's physics engine for realistic collisions and settling behavior. A particle pooling system keeps the game performant even with hundreds of water particles on screen. The dam effectiveness algorithm scans vertical slices across your structure to calculate the maximum water height and identify leak points, creating emergent gameplay where every placement decision matters.

### Features
- 🪵 **Drag & Drop Building** - Intuitive touch controls for placing logs and twigs
- 🔄 **Two-Finger Rotation** - Precise angle control with multi-touch gestures
- 💧 **Real-Time Water Physics** - Individual particle simulation with collision detection
- 🌊 **Dynamic Pond System** - Watch water levels rise and fall based on dam effectiveness
- 🔍 **Smart Leak Detection** - The game identifies gaps in your dam and spawns targeted leak particles
- 📊 **Dam Effectiveness Scoring** - Real-time feedback on how well your structure blocks water
- 🎯 **Target Fill Challenge** - Reach and maintain pond level goals with visual indicators
- ⏱️ **Time-Based Objectives** - Complete levels before time runs out or hold water level thresholds
- 🎨 **Landscape Canvas** - Full-screen horizontal gameplay optimized for building
- 🧘 **Meditative Pacing** - No lives, no ads, just you and the river

### Notes
- **Current Status:** Sprint 2 in progress - Core water simulation and dam mechanics complete; working on win/lose conditions and level progression
- **Tech Stack:** Swift, SpriteKit, SwiftUI, OSLog for debugging
- **Requirements:** iOS 16.0+, Xcode 15.0+ for development
- **Orientation:** Landscape-only for optimal building experience
- **For whom:** Puzzle lovers, nature enthusiasts, anyone seeking calm mobile gameplay
- **Development Philosophy:** Iterative sprints with clear MVP goals; no feature creep
- **Performance:** 60 FPS target with particle pooling and optimized physics

---

## Agentic Directory

**Status:** Live
**Platforms:** Web — https://agentic.ai
**Project page:** https://skylarkcreations.com/projects/agentic-directory

Find AI that actually does things. A curated directory of agentic AI products and services, scored on a 20-point agenticness scale across five dimensions.

### Overview
Agentic Directory is an AI-first directory that helps users discover agentic AI products — tools and services that can take autonomous action, not just chat. Every product is evaluated and categorized using a 20-point "agenticness" scoring system across five dimensions.

The directory features vendor portals for claiming and managing listings, a submission pipeline with AI-powered enrichment via web crawling, and a conversational "Ask" interface that lets users describe what they need and get matched to relevant products. Categories, tags, and use-case filtering make it easy to find the right tool for the job.

### Why
The agentic AI space is exploding, but discovery is fragmented. Most directories lump chatbots in with genuinely autonomous systems. We built Agentic Directory to draw a clear line: this is for AI that acts. The scoring system gives users a fast, honest read on what a product can actually do on its own.

### How It Works
1. **Browse or search** the directory by category, use case, or tag
2. **Ask** — describe your problem in plain language and get matched to relevant products
3. **Compare** — every listing shows an agenticness score across five dimensions
4. **Submit** — vendors can submit products, claim listings, and manage their profiles
5. **Stay current** — a background worker crawls and re-enriches listings on a schedule

### Features
- **20-Point Agenticness Score** — Five-dimension evaluation: autonomy, planning, tool use, memory, and adaptability
- **Conversational Discovery** — "Ask" interface matches plain-language needs to products using AI embeddings
- **Vendor Portal** — Claim listings, manage profiles, upgrade plans via Stripe
- **AI-Powered Enrichment** — Background worker crawls product sites and extracts structured data
- **Full-Text + Semantic Search** — Keyword search plus pgvector embeddings for meaning-based matching
- **Category Taxonomy** — Organized by use case, industry, and capability

### Notes
- **Tech Stack:** Next.js 16, TypeScript, React 19, Tailwind CSS 4, shadcn/ui, PostgreSQL (Neon) + Prisma + pgvector, Supabase Auth, Stripe, OpenAI GPT-5.1, Firecrawl, SendGrid, Vitest + Playwright
- **Hosting:** Render (web service + background worker)
- **For Whom:** Teams evaluating agentic AI tools, founders scouting the competitive landscape, developers looking for agent frameworks

---

## Reframed

**Status:** Live
**Platforms:** Web — https://reframed.news/
**Project page:** https://skylarkcreations.com/projects/reframed

Facts, then left and right. A daily news briefing that leads with cited consensus facts, then shows each story through contrasting Left and Right framings — with a paste-any-URL reframe tool.

### Overview
Reframed is an attention-respecting daily briefing for people who want substance over spin. Each morning it delivers 12 stories across 6 news categories, built through a fully automated pipeline: ingest events from newsapi.ai with relevance scoring, generate consensus facts with source citations, then produce two contrasting value-frame perspectives at two intensity tiers each.

The core interaction is the dial — a slider from -100 to +100 that smoothly transitions between five content zones: Frame A at full or moderate intensity, facts-only in the center, and Frame B at moderate or full intensity. Every factual claim cites at least one source. Readers see who reported what, not just what to think.

Nine value-frames — Human Impact, Accountability, Freedom & Rights, Security & Order, Economic Stakes, Institutions & Process, Cultural Identity, Social Cohesion, and Innovation & Progress — are screened for relevance to each story. The best contrasting pair is selected automatically, and readers can swap frames via dropdown selectors.

### Why
Generic news feeds and keyword alerts are noisy and don't explain implications. Reframed is built to deliver signal over noise: fewer alerts, deduplicated coverage, and fast, role-aware interpretation with receipts. The dial gives readers agency over how much opinion they consume without sacrificing the underlying facts.

### How It Works
1. **Ingest** — Events pulled from newsapi.ai with keyword pre-filtering and LLM relevance scoring, balanced across 6 categories (12 stories total)
2. **Generate** — Per story: screen all 9 value-frames for relevance, pick the best contrasting pair, generate consensus facts with src-N citations, then generate moderate + max content per selected frame
3. **Email** — Compose HTML + plain text digest and send via SendGrid to subscribers
4. **Read** — Browse the web app, use the dial to control commentary intensity, swap frames, tap citations to verify claims

### Features
- **Interactive Dial** — Smooth slider from -100 to +100 across five content zones: framed commentary at two intensities, or facts-only center
- **Cited Consensus Facts** — Every factual claim references at least one source; phantom citations are automatically stripped
- **9 Value-Frames** — Human Impact, Accountability, Freedom & Rights, Security & Order, Economic Stakes, Institutions & Process, Cultural Identity, Social Cohesion, Innovation & Progress
- **6 News Categories** — U.S. Politics, International Affairs, Business & Markets, Technology & Society, Science & Climate, Rights & Justice
- **Daily Email Digest** — Story previews with fact highlights delivered to subscribers
- **Content Confidence Signals** — Well-covered, Developing, or Single-angle based on source diversity

### Notes
- **Tech Stack:** Next.js 16, TypeScript, React 19, Tailwind CSS v4, Neon Postgres + Drizzle ORM, Supabase Auth, OpenAI GPT-5.2 pro, newsapi.ai, SendGrid, Vitest
- **Pipeline:** Fully automated daily cron at 4 AM ET via Render; retry logic for API failures, skip-and-continue for individual story failures, never sends broken or empty emails
- **For Whom:** Anyone tired of spin — news readers who want to verify claims, compare perspectives, and control how much commentary they consume

---

## Urban Journeys

**Status:** In Development
**Platforms:** Web (in development) — https://urban-journeys.com/
**Project page:** https://skylarkcreations.com/projects/urban-journeys

Walk your city with intention. Discover it with delight. Urban Journeys composes themed walking routes with curated stops, engaging descriptions, and accurate distances — optimized for discovery, not speed.

### Overview
Urban Journeys is a journey composer — not a navigation app. Instead of optimizing for speed, it optimizes for discovery, theme, and delight. Tell it where you're starting, how far you want to walk, and what you want to see. It crafts an optimized walking route with curated stops, engaging descriptions, and accurate distances.

The origin: instead of always driving to the mountains to hike, why not hike the city? The first experiment was "walk to every theater in Denver" — a 35-mile pilgrimage. That's when the real idea emerged: themed urban walks, optimized for humans, not cars.

A 5-phase streaming pipeline discovers POIs from Google Places, curates the best stops with GPT-5.1, optimizes route order with TSP algorithms, adjusts distance to hit your target within 15%, then narrates the whole journey. Progress streams to your browser in real time so you can watch each phase unfold.

### Why
Navigation apps answer "get me from A to B." Urban Journeys answers "give me an experience." Points aren't obstacles — they're the point. We built it to push past the same top-10 lists and star ratings and bring back serendipity to everyday life. The hybrid approach — LLMs for creativity, algorithms for math — ensures accurate distances while maintaining personalized, engaging content.

### How It Works
1. **Enter your starting location** — Address autocomplete via Google Places
2. **Choose your journey size** — Quick Walk (1.5 mi / 45 min) to Epic Trek (12 mi / 6 hr)
3. **Pick your themes** — 12 discovery themes with real-time POI counts showing what's near you
4. **Compose** — Watch a 5-phase pipeline: Discovery (Google Places) → Curation (GPT-5.1) → Optimization (TSP solver) → Distance Adjustment → Narration
5. **Export & Go** — Open in Google Maps, Apple Maps, or download GPX for your favorite hiking app

### Features
- **12 Discovery Themes** — Theaters, street art, breweries, coffee shops, bookshops, live music, parks, historic buildings, food, architecture, vintage shops, markets
- **6 Journey Sizes** — Quick Walk (1.5 mi) to Epic Trek (12 mi) with calibrated stop counts and search radii
- **Hybrid AI + Algorithm Pipeline** — GPT-5.1 curates interesting stops; TSP algorithms (nearest-neighbor + 2-opt + or-opt) optimize route order
- **Real-Time SSE Streaming** — Watch the 5-phase composition pipeline as it works
- **Accurate Distances** — OpenRouteService walking directions keep routes within 15% of target
- **Export Anywhere** — GPX download, Google Maps, Apple Maps
- **City-Agnostic** — Works in any city worldwide via Google Places
- **Loop or Point-to-Point** — Your choice

### Notes
- **Tech Stack:** Next.js 15, TypeScript, React 19, Tailwind CSS, Zustand, Mapbox GL JS, OpenAI GPT-5.1, Google Places API, OpenRouteService, Framer Motion
- **Performance:** 60-90 second composition time, <2% error rate, 100% POI accuracy (verified via Google Places)
- **For Whom:** Urban explorers, tourists, locals seeking new perspectives, anyone who'd rather discover than navigate

---

## Syntopica

**Status:** In Development
**Platforms:** Web (in development) — https://better-reads-r9n1.onrender.com/
**Project page:** https://skylarkcreations.com/projects/syntopica

Make your books think together. Syntopica imports your Amazon library and Kindle highlights, then helps you do cross-book synthesis — bringing together ideas from multiple books to form new understanding.

### Overview
Syntopica is an AI-powered reading companion named after syntopical reading — the highest level of reading described by Mortimer Adler in How to Read a Book: reading multiple books on a single subject and constructing an analysis that none of them contain individually.

The Capture layer is complete: a Chrome extension imports your full Amazon library (Kindle, Audible, physical) and Kindle highlights with zero manual entry, using Amazon's GraphQL API for books and DOM scraping for highlights. Background sync keeps everything current on a 4-hour cycle. The web app provides a library grid with covers, book detail pages, and a highlights viewer.

A three-source metadata enrichment pipeline (Google Books → OpenLibrary → Wikidata SPARQL) achieves 93% description coverage and 92% ISBN coverage. Next up: AI-powered collections for grouping books by theme, asking questions across them, and generating citation-backed synthesis.

### Why
Serious readers accumulate thousands of highlights across hundreds of books, but those ideas stay siloed. Goodreads tracks what you've read; Readwise resurfaces quotes. Neither helps you think across books. Syntopica is built for the reader who wants to ask "What do these five authors say about attention?" and get a cited, synthesized answer drawn from their own library.

### How It Works
1. **Install the Chrome extension** — it reads your Amazon library and Kindle highlights automatically
2. **Import** — Books arrive in chunked uploads (150/chunk), highlights in batches of 300, with auto-retry and recovery
3. **Enrich** — A background worker progressively enriches books from Google Books, OpenLibrary, and Wikidata SPARQL
4. **Browse** — Library grid, book detail pages, all-highlights view with search and filtering
5. **Synthesize** (coming soon) — AI collections for cross-book questions, citation-backed synthesis, and single-book deep reading assists

### Features
- **Full Amazon Library Import** — Kindle, Audible, and physical books via Chrome extension; 4,200+ books imported at scale with zero errors
- **Kindle Highlight Extraction** — 29,000+ highlights imported via DOM scraping with multi-pass scroll strategy
- **Background Auto-Sync** — 4-hour cycle keeps library and highlights current
- **Three-Source Enrichment** — Google Books + OpenLibrary + Wikidata SPARQL for 93% description coverage
- **Library Grid + Book Detail** — Cover images, metadata, highlights viewer, search and filters
- **Cross-Book Synthesis** (coming soon) — AI collections, thematic grouping, citation-backed analysis

### Notes
- **Tech Stack:** Next.js 16, TypeScript, React 19, Tailwind CSS 4, Neon PostgreSQL + Drizzle ORM, Supabase Auth, OpenAI (GPT-4.1 + embeddings), Chrome Extension (Manifest V3), Sentry, Amplitude
- **Current Status:** Capture layer complete (import, sync, enrichment). Think layer (AI collections + synthesis) is next.
- **For Whom:** Serious readers, researchers, lifelong learners who want their reading to compound

---

## Walk And

**Status:** In Development
**Platforms:** Web (in development) — https://walk-and-web.onrender.com/
**Project page:** https://skylarkcreations.com/projects/walk-and

Minimum viable movement. A daily movement coach built around a routine you actually do — self-contained workouts, daily reminders, and a GitHub-style heatmap to visualize consistency.

### Overview
Walk And is a daily movement coach designed around one idea: the best workout is the one you actually do. Instead of elaborate programs or gym-centric plans, Walk And pairs with your daily walk and gives you self-contained coaching sessions you can start in under 10 seconds.

The app features calendar export for daily reminders, a GitHub-style contribution heatmap to visualize your movement consistency, and a focus on low-friction sessions that fit into the rhythm of your day.

### Features
- **Self-Contained Coaching** — Complete workouts without leaving the app
- **Daily Reminders** — Calendar export to stay consistent
- **GitHub-Style Heatmap** — Visualize your movement contributions over time
- **Low Friction** — Start a session in under 10 seconds

### Notes
- **Tech Stack:** React 18, TypeScript, Vite, Tailwind CSS, Node.js + Express, Prisma ORM, PostgreSQL
- **Hosting:** Render (static frontend + API service + PostgreSQL)
- **For Whom:** Anyone who wants a simple, consistent movement habit without the overhead of a full fitness app

---

## Chat With Feels

**Status:** In Development
**Platforms:** Web (in development)
**Project page:** https://skylarkcreations.com/projects/chat-with-feels

With feeling, without fiction. A chatbot with a transparent, neuroscience-inspired emotion engine that modulates style and initiative — tone, cadence, emphasis — while keeping truth and safety intact.

### Overview
Chat With Feels doesn't pretend to feel — it models a useful slice of the human emotional control loop to make conversations smoother, kinder, and more effective. The affect layer changes style, not substance: tone, cadence, emphasis, and initiative, while keeping truth and safety fully intact.

The engine tracks 21 emotion families (anger, joy, awe, gratitude, confusion, nostalgia, and more) using mean-reverting Ornstein-Uhlenbeck dynamics with emotion couplings. A lightweight appraisal layer detects conversational signals and generates emotion deltas. A style transform converts the current affect state into style instructions for the LLM — never altering facts or safety constraints.

The system also includes rolling summaries and embedding-based long-term recall for memory, plus time-based proactive check-ins that use current affect to decide when and how to re-engage.

### Why
Current chatbots are either flat (no emotional awareness) or performative (fake emotions for engagement). Chat With Feels takes a third path: transparent affect modeling grounded in neuroscience. The emotion engine is visible and honest — it shapes how the assistant communicates, not what it says.

### How It Works
1. **Appraisal** — Lightweight keyword and intent detection reads conversational signals and generates emotion deltas
2. **Affect Engine** — 21 emotion families evolve via mean-reverting dynamics (Ornstein-Uhlenbeck) with cross-emotion couplings
3. **Style Transform** — Current affect state is converted to style instructions (tone, cadence, warmth, initiative level)
4. **LLM Generation** — Frontier model receives style instructions alongside the conversation; facts and safety are never altered
5. **Memory** — Rolling summaries + embedding-based long-term recall ensure continuity across sessions

### Features
- **21 Emotion Families** — Anger, disgust, fear, sadness, shame, envy, surprise, joy, interest, love, awe, pride, gratitude, relief, calm, nostalgia, boredom, confusion, craving, lust, entrancement
- **Mean-Reverting Dynamics** — Ornstein-Uhlenbeck process ensures emotions naturally decay toward baseline
- **Emotion Couplings** — Cross-emotion interactions (e.g., joy suppresses sadness, fear amplifies surprise)
- **Style-Only Modulation** — Affect changes tone and initiative, never facts or safety
- **Long-Term Memory** — Rolling summaries + embedding-based recall for continuity
- **Proactive Check-Ins** — Time-based re-engagement modulated by current affect state

### Notes
- **Tech Stack:** Node.js 20+, TypeScript, OpenAI, MongoDB (Atlas or local), Vite frontend
- **Philosophy:** "With feeling, without fiction" — transparent affect modeling, not simulated emotions
- **For Whom:** Anyone who wants AI conversations that feel more natural without sacrificing honesty