Posts

  • Context Limits Aren't a Claude Problem — They're a Hygiene Problem

    Context limits aren't a Claude problem — they're a hygiene problem. Here's what two YouTubers and a SOP taught me.
  • Evolve Is a Scheduler: Why Agent Memory Gets the Last Phase Wrong

    Most agent frameworks treat the final pipeline phase as a logger. That is the wrong mental model. Evolve is a scheduler — and the difference determines whether your system learns or just records.
  • Resting State: Bounded Recursion as a Frame for Self-Improvement

    Most people think of improvement as a pipeline. Do the steps, exit at the end. But there's a more honest model — and it reframes failure as a valid completion.
  • Embracing Failure: A Novel Architecture for Agentic Resilience and Learning

    Most agent systems treat failure as something to hide. This architecture inverts that — failure is a first-class outcome that makes the system more robust over time.
  • SRCGEEE Analysis: Embracing Failure — Agentic Resilience and Learning

    A SRCGEEE framework analysis of the Embracing Failure architecture post — what it gets right, where the edge cases are, and what a production system still needs to solve.
  • PhoneBuddy's First Live Call — and the Scammer Who Proved the Point

    The day PhoneBuddy answered its first real call — a live AI voice, a real PSTN number, and a contractor Apple flagged as spam who wasn't.
  • How I Built This Blog (With the AI That Was Already Doing My Other Work)

    Three failed attempts, one session, one command. How a conversation about PhoneBuddy turned into a working blog.
  • Publish Pipeline Kickoff

    Kicking off a Markdown-first multichannel blog pipeline with dry-run orchestration and governance-first publishing controls.
  • RoadTrip: An Intelligent, Trusted Travel Partner

    How do you trust an AI agent that needs access to the internet? RoadTrip is a proof-of-concept framework for building verifiable, auditable AI skills.
  • How We Built a Trusted AI Skill: A Case Study in Rigorous Development

    How immutable prototypes, test infrastructure, and oracle-based verification create trustworthy agentic systems.
  • Personal Health Studio

    From PDF to Intelligent Insights: Building a Personal Health Advisor in One Day

    The Challenge

    Imagine you have a stack of health documents – lab reports, medication lists, allergy alerts, vital signs spread across different files and formats. Your doctor gives you a PDF export from their system. It’s all there, but it’s just static text.

    Want to know your average blood glucose level? You’d have to search through the document, find the numbers, and do the math yourself. Want to see what medications interact? Good luck – that’s a mental task. Want to quickly get a health snapshot? You’re re-reading the same document over and over.

    This is what most patients deal with every day.

    The Idea

    What if you could take that PDF health record and actually talk to it? Ask it questions in plain English?

    “What’s my average blood glucose?”
    “What medications am I taking?”
    “Do I have any severe allergies I should know about?”

    And get intelligent, calculated answers instantly.

    That’s exactly what we built. In one day.


    Meet Personal Health Studio

    Personal Health Studio is an open-source health intelligence platform that transforms static health documents into an interactive, queryable database.

    How It Works (The Simple Version)

    1. Upload your health document (PDF, HTML export, etc.)
    2. The system extracts patient data, lab results, medications, conditions, and vital signs
    3. We store it in a cloud database, organized and structured
    4. You ask questions in natural language – no special syntax needed
    5. Get intelligent insights calculated in real-time

    No manual data entry. No pre-built dashboards to click through. Just ask.


    What You Can Do With It

    🩺 Understand Your Lab Results

    • “What’s my average blood glucose?” → Gets the actual average from all your measurements
    • “Are my cholesterol levels improving?” → Tracks trends over time
    • “Which tests came back abnormal?” → Highlights what needs attention

    💊 Manage Your Medications

    • “What medications am I currently taking?” → Complete list with dosages and frequencies
    • “When did I start Metformin?” → Exact dates and details
    • “What’s my Aspirin for?” → Context and indication for each drug

    ⚠️ Track Important Health Info

    • “What are my known allergies?” → Severity levels included (this matters!)
    • “What conditions do I have?” → Active diagnoses at a glance
    • “Show me my recent vital signs” → Blood pressure, heart rate, temperature – all organized

    All of this calculated and delivered instantly. In English.


    Why This Matters

    For Patients: Your health data becomes actionable instead of just informational. You understand your own health better.

    For Families: Help elderly relatives manage their health by asking questions for them. No technical knowledge needed.

    For Developers: Open source, well-architected, production-ready. Build on it. Extend it. Make it better.

    For Healthcare: A bridge between static health records and intelligent analysis, all respecting privacy and HIPAA principles.


    Built in One Day

    Here’s the kicker: This entire system was built, tested, and deployed in a single day.

    Think about that. From concept to working prototype with:

    • ✅ Cloud database infrastructure
    • ✅ Natural language understanding
    • ✅ Real data extraction and storage
    • ✅ Semantic query engine
    • ✅ End-to-end testing
    • ✅ Beautiful demo materials

    One developer. One day. Fully functional.

    Check out the repo and see the code. It’s clean, documented, and ready to go.


    The Architecture (Without the Jargon)

    Personal Health Studio uses:

    • Snowflake for data storage (enterprise-grade cloud database)
    • Python for the backend logic
    • Natural Language Processing to understand medical terminology
    • Pydantic for data validation and type safety
    • Open Source all the way – MIT licensed

    The result? A system that’s simultaneously powerful and elegant.


    Try It Yourself

    Want to see it in action?

    1. Head to github.com/bizcad/personal-health-studio
    2. Check out the demo materials and documentation
    3. Run the end-to-end test
    4. Watch as a health document gets extracted, stored, and intelligently queried

    The repo includes everything you need to get started – sample data, demo scripts, and step-by-step instructions.


    What’s Next?

    This is just the beginning. Imagine:

    • Mobile App: Ask health questions on your phone
    • Wearable Integration: Real-time vital signs from fitness trackers
    • Predictive Analytics: “Your blood glucose trends suggest prediabetes – here’s what to do”
    • Provider Integration: Share data securely with doctors for better care
    • Multi-patient Support: Manage family health records all in one place

    All of this is possible with the foundation we’ve built.


    The Bigger Picture

    Healthcare data shouldn’t be trapped in PDFs and portals. Patients should own their data. Developers should be able to build intelligent health tools. Privacy and security should be built in, not bolted on.

    Personal Health Studio shows what’s possible when you combine modern cloud infrastructure, smart natural language processing, and thoughtful API design.

    And it proves you can build it fast.


    Get Involved

    • Star the repo on GitHub: personal-health-studio
    • Try it out with your own health documents
    • Contribute improvements and features
    • Share it with someone who could benefit
    • Build on it – the foundation is solid and ready

    Personal health data deserves better tools. This is what better looks like.


    Personal Health Studio is open source, production-ready, and waiting for you to put it to work.

    Check out the GitHub repo →


    Built in one day. Ready to change how people interact with their health data.

  • One-Button Blog Publishing: Autonomous Skill in Action

    The RoadTrip blog publisher skill is now fully operational end-to-end, demonstrating safe autonomous AI in practice.
  • Skill Development Methodology

    How we build reusable, composable skills in the RoadTrip framework using spec-driven development. From specification through testing to production deployment.
  • Orchestrator Architecture Proven

    We have successfully demonstrated that the RoadTrip orchestrator can handle complex, multi-step workflows end-to-end without human intervention. This post documents what we learned and what's next.

subscribe via RSS