Hey everyone! Healthcare is starting off 2026 with a bang. Anthropic just made another major move in this space, and it is a reminder that AI in healthcare is not a distant future story anymore. The real shift is happening behind the scenes, where these models start plugging into the messy infrastructure that actually runs care.

🏥 Founder’s Corner: We go deeper on healthcare. Not the flashy “talk to your doctor” stuff, but the invisible highway behind care. The messy systems, the disconnected data, and why connectors might be the real unlock.

🧠 AI Education: ChatGPT Part 3 is all about working with your stuff. Files, Projects, Canvas, history search, and Custom Instructions. The features that turn ChatGPT from a one-off chat into something you can actually use consistently.

📈 10-Minute Win: A post-trade Post-Mortem Playbook. A simple loop to turn every trade into a lesson, spot your repeated mistakes, and build one rule at a time so your process improves.

Let’s dive in.

1) Apple, Google strike Gemini deal for revamped Siri in major win for Alphabet

Summary: Apple will integrate Google’s Gemini models into a revamped Siri later in 2026, deepening Apple’s AI stack while expanding Google’s footprint across Apple’s massive device base.

Why it matters: This is a “distribution wins” moment: whichever model sits behind default assistants can shape how billions experience AI — and who gets the data, the developer mindshare, and the ecosystem gravity.

2) Advancing Claude in healthcare and the life sciences

Summary: Anthropic is expanding Claude’s healthcare and life sciences tooling, including “Claude for Healthcare” (HIPAA-ready) and deeper integrations for life sciences workflows like trials and regulatory ops.

Why it matters: This is the clearest sign yet that “AI in medicine” is shifting from demos to workflow-native tools (connectors, compliance, real systems) — which is where real adoption happens.

3) Google expands AI-assisted shopping features of Gemini

Summary: Google is expanding Gemini’s shopping capabilities by partnering with major retailers (including Walmart and Shopify) and enabling “instant checkout” experiences inside the chat flow.

Why it matters: AI assistants are becoming transaction engines, not just answer engines — and whoever controls the in-chat “buy” moment could end up controlling a huge slice of digital commerce.

4) Personal Intelligence: Connecting Gemini to Google apps

Summary: Google is rolling out “Personal Intelligence” for Gemini, letting it use context from connected Google apps (like Gmail, Photos, Search, and YouTube) to deliver more personalized help.

Why it matters: The next AI leap is “context” — not just smarter models, but assistants that can safely use your information to be genuinely useful (with privacy controls determining whether people trust it).

5) Anthropic, OpenAI have taken early steps to go public - NYT

Summary: A report says Anthropic and OpenAI have started preliminary IPO work, alongside talk that SpaceX is also moving toward a potential public listing path.

Why it matters: If the biggest AI labs shift toward IPO readiness, the market pressure changes: more scrutiny, more demand for durable revenue, and faster “enterprise-grade” productization — not just research wins.

The Invisible Highway: Can Claude for Healthcare Fix Healthcare’s Infrastructure Problem?

Last week I wrote about the rise of the "Executive Patient" and how OpenAI’s consumer launch is fundamentally changing patient behavior. We discussed a world where the patient arrives with more data, more agency, and higher expectations than ever before. But there is a glaring problem with that vision. When that empowered patient walks through the doors, they are stepping into a system that is still running on infrastructure built for a different era.

For decades the healthcare industry has struggled to connect the dots. Billions have been spent digitizing records, yet we still live in a world of disconnected data islands. The payer system does not speak to the provider system and clinical data does not flow to the claims department. Our medical claims system runs on J-codes, which were introduced in 1978, long before electronic billing was introduced. Pharmacy claims run on a completely different set of electronic standards that do not connect to the medical coding systems. This lack of continuity is the root cause of the broken experiences that frustrate patients and burn out clinicians.

Enter Claude for Healthcare.

Anthropic quietly released a platform aimed squarely at the critical infrastructure of the healthcare enterprise. They are not trying to change how you talk to your doctor. They are trying to fix the broken highway that connects the entire industry.

The Billion-Dollar Disconnect

The healthcare industry has a massive infrastructure problem. Despite the widespread adoption of Electronic Health Records, clinical teams still spend 13 hours every week completing prior authorizations according to a January 2026 AMA survey. The data exists but it is trapped in static files that require human manual labor to move, read, and interpret.

We have spent years trying to solve this by layering new applications on top of old foundations. But new tools and technology can only go so far when the underlying infrastructure remains siloed. We have built digital filing cabinets when what we really needed were digital bridges. You cannot expect technology to fix the foundational workflow problems that exist in healthcare.

This disconnect has a price tag. The industry currently loses $496 billion annually to billing and insurance-related administrative waste. The result is a system where 93% of physicians report that prior authorization delays lead to negative clinical outcomes. The infrastructure cannot keep up with the volume of care, and it certainly cannot keep up with the new speed of the "Executive Patient."

Claude as the Universal Adaptor

Anthropic’s strategy is distinct because it treats AI as an operational layer rather than just a chatbot. The core value of the new platform lies in its "Connectors." Unlike a general-purpose model, Claude comes with pre-built integrations into the industry's sources of truth including the CMS Coverage Database, ICD-10 coding systems, and the NPI Registry.

This connectivity allows organizations to automate the complex workflows that usually keep clinicians glued to their screens. Elation Health provides a perfect example of this in action. They embedded Claude directly into their primary care EHR to handle chart summarization. Instead of spending ten minutes clicking through tabs to review a patient's history, the physician gets an immediate, cited summary of the patient's story. The result is that clinicians are finding answers 61% faster, which allows them to stop hunting for data and start looking at the patient.

On the operational side, Carta Healthcare is using the platform to tackle the heavy lift of clinical data abstraction. This is typically a manual process where staff scour records to report quality data to registries. By using Claude to automate this extraction, they reduced processing time by 66% while maintaining 99% accuracy. This effectively unblocks the flow of quality data without adding administrative overhead to the staff.

These examples prove that the value of AI isn't just in writing emails. It is in automating the invisible friction of the back office so that clinical teams can return their focus to the human work of care delivery.

Trust as the Ultimate Feature

The final piece of this puzzle is safety. Building a digital highway for healthcare requires more than just connectivity; it requires a foundation of absolute trust. Enterprise healthcare organizations are risk-averse by design. They cannot afford to bet on a "black box" model that might hallucinate a diagnosis or expose patient data.

This is where Anthropic is placing its strategic bet. They have engineered Claude on HIPAA-ready infrastructure with a "Constitutional AI" framework that acts as a permanent safety inspector. Unlike standard models that simply predict the next likely word, Claude is designed to show its work. It provides rigorous audit trails and citations, allowing a claims adjuster or a physician to trace every AI-generated output back to a specific medical guideline or coverage policy.

Mike Reagin, the CTO of Banner Health, validated this approach, noting that they chose Claude specifically for its "reputation for responsible AI" and the belief that safety is "non-negotiable." In an industry where a single error can lead to a lawsuit or a denied life-saving procedure, this ability to verify the "why" behind the answer is the foundation to scalable infrastructure.

The Road Ahead

If OpenAI is the catalyst for patient agency, Claude is the infrastructure that makes that agency actionable. We are moving toward a future where the back office is increasingly self-driving through AI automation, offering a critical inflection point for legacy healthcare companies.

The winners of the next decade will not be the organizations that simply buy the software. They will be the leaders who have the courage to experiment and fundamentally rebuild their workflows from the ground up. Business architecture will be redesigned as the scope of what is possible evolves with AI, and specifically the push for data-driven automation ecosystems like "Claude for Healthcare." The strategic imperative for legacy organizations is to use this infrastructure to strip away the administrative drudgery and redisperse their human staff to where they matter most: the patient experience.

I predict that within 36 months, the labor-intensive tasks that burden operational and clinical teams will be streamlined and automated through data connection highways powered by agentic AI workflows. This will impact every sector of healthcare, from administrative staff working to clear a prior authorization, to clinicians submitting for a surgical procedure. The strategic imperative is to take those recovered hours and rebuild strategies that ensure the patient journey is supported. The true moat of the future is not your proprietary data silo; it is how you connect to the broader ecosystem to remove operational strain that slows down service and erodes trust. Automation builds the highway, but it is the human experience that will keep the Executive Patient from taking the next exit.

In Part 1, you learned ChatGPT is a workspace with multiple modes and limits. In Part 2, you learned which modes help you create and understand faster. Now we solve the real beginner problem: your work gets messy. Chats pile up. Files get lost. You forget what you asked last week. Part 3 is about the features that turn ChatGPT from a one-off chat into something you can actually use over time. This is where consistency starts.

Feature Index

Group 1: Bring your own material in

File uploads

What it is: You can upload documents, spreadsheets, and images so ChatGPT can read them and help you work with them.

Why it matters: Most real tasks live in your stuff, not in your head. Uploads let ChatGPT stop guessing and start responding to the actual content/context you have.

When to use it:

• You want a summary of a document.
• You want key points extracted.
• You want a table or spreadsheet interpreted.
• You want help understanding a screenshot.

Limits and gotchas:

• Free tier has upload limits. They can vary by demand.
• Clean inputs beat messy ones. A clean file gets a better answer.
• If a file has sensitive information, consider removing it before upload.

Where to find it: Use the upload option in your chat tools. On mobile, it is usually attached to the message box options..

The beginner rule for uploads

If you want better results, give ChatGPT fewer, cleaner files.

Do this:

• Upload one file per task.
• Tell ChatGPT what you want and what to ignore.

Avoid this:

• Uploading five files and hoping it figures it out.
• Asking for insights without stating the goal.

Group 2: Projects, the fastest way to stay organized

Projects

What it is: Projects let you group related chats and files together. It is like creating a folder for one ongoing goal.

Why it matters: Without Projects, you end up with scattered chats and repeated context. Projects reduce re-explaining and make long work easier.

When to use it:

• A task that lasts more than one day.
• A repeating routine you want to improve over time.
• Anything with multiple steps and multiple files.

Limits and gotchas:

• Free tier Projects have limits, especially around file counts.
• Do not treat a Project like a dumping ground. Keep it focused.

Group 3: Canvas, the best place to draft and revise

What it is: Canvas is a drafting workspace. It is designed for writing and editing with ChatGPT, without the mess of long chat threads.

Why it matters: Chat is fine for brainstorming. Canvas is better for creating a clean final draft.

When to use it:

• Newsletter writing
• Any document you plan to copy and paste into another tool

Limits and gotchas:

• Canvas is available on web and desktop. It is not fully available on mobile yet.
• Canvas works best when you give clear structure.

A simple workflow for publishing:

1. Brainstorm in chat.
2. Move the draft into Canvas.
3. Revise in Canvas using targeted edits.

Export from Canvas

What it is: Canvas supports exporting drafts into common formats so you can move the content into your workflow.

Why it matters: It reduces copy and paste mess and keeps formatting cleaner.

Group 4: Make ChatGPT consistent with Custom Instructions

Custom Instructions

What it is: Custom Instructions is where you tell ChatGPT how you want it to behave by default. It is like setting your preferred “operating system settings” for conversations.

Why it matters: Beginners waste time repeating the same preferences. Custom Instructions removes that repetition and makes outputs more consistent across chats.

Use Custom Instructions when you find yourself repeatedly saying things like:

• Keep it short.
• Use plain English.
• Ask me questions before you assume.
• Give me a table.
• Use a specific tone.

Keep it simple. Make it specific. Do not overload it.

A strong beginner template:

• I am an AI beginner.
• I like short sentences and plain English.
• I want examples grounded in everyday life.

How I want ChatGPT to respond:

• Start with a short definition.
• Then give one clear example.
• Then list common confusion points.
• If you are missing info, ask me one question before answering.

What not to do:

• Do not use jargon without defining it.
• Do not guess or invent facts.
• If you are unsure, say you are unsure.

Group 5: Find and share your past work

Chat history and search

What it is: ChatGPT stores your past chats, and you can search them.

Why it matters: Your best prompts and outputs are reusable. History search turns ChatGPT into a personal library.

Beginner habit: Whenever you get a great output, rename that chat with a clear title so future you can find it.

Share a chat

What it is: You can share a chat with someone else using a share link.

Why it matters: It keeps context intact. It is better than screenshots. It is better than copying fragments.

Group 6: Use GPTs, when you want a guided experience

Use GPTs from the GPT Store

What it is: GPTs are custom versions of ChatGPT designed for specific tasks. You can browse and use GPTs on Free, but access may be limited by capacity.

Why it matters: GPTs can reduce setup time because the workflow is already structured.

How to use them well

• Treat them as templates.
• Still be clear about your goal.
• Still verify important outputs.

One-screen recap

• File uploads help ChatGPT work from real material instead of guessing.
• Projects keep long work organized and reduce repeated context.
• Canvas is the best place to draft and revise content you will publish.
• Custom Instructions makes ChatGPT more consistent across chats and saves time.
• History search turns good prompts into reusable assets.
• GPTs can guide you through tasks, but results still depend on your inputs.

Post-Mortem Playbook (After a Trade)

Most investors don’t lose because they “picked the wrong stock.” They lose because they repeat the same mistakes without noticing behavior patterns: chasing, sizing too big, moving stops, or exiting for the wrong reason. This workflow turns every trade into a quick learning loop — so your process improves.

Step 1 — Capture the trade in a clean template (2 minutes)

Open Google Sheets and create a new tab named: Trade Post-Mortem.

In Row 1, paste these headers:

Trade ID | Ticker | Date In | Date Out | Entry | Exit | Size ($ or shares) | Thesis (1–2 lines) | Planned Stop | Time Horizon | Exit Reason | Outcome ($ / %) | Emotion (1 word)

What it should look like (example row):

T-001 | AAPL | 01/05 | 01/12 | 182.10 | 176.50 | $500 | “Earnings run-up + momentum” | 178.00 | 2 weeks | “Stop hit” | -3.1% | “Rushed”

Optional (nice upgrade): add a chart screenshot (broker chart or TradingView free). If you do, you can upload the screenshot to ChatGPT — but it’s not required.

Step 2 — Run the “process-first” post-mortem in ChatGPT (4 minutes)

Copy/paste this prompt into ChatGPT, then paste your single trade row (or multiple rows if you want):

Role: You are my Trade Post-Mortem Coach.Goal: Help me improve decision-making. Judge process, not outcome.

Rules:

• Be blunt and specific. No hype.
• If info is missing, list the exact questions you need (max 5).
• Separate “Bad outcome but good process” vs “Good outcome but bad process.”

My trade details (paste below):[PASTE THE ROW FROM SHEETS]

Output exactly in this structure:

1. One-sentence verdict (process quality)
2. What I did right (3 bullets)
3. What likely hurt me (3 bullets)
4. Mistake tags (choose: Thesis / Timing / Sizing / Risk / Exit / Execution / Emotion / No plan)
5. If-Then upgrades (3 rules I should adopt next time)
6. Next trade checklist (6 items, yes/no)

If you have a chart screenshot and want extra clarity, add: “Here’s a chart screenshot for context” and upload it.

Step 3 — Make ChatGPT build your scorecard table (3 minutes)

Now paste this follow-up prompt:

Create a Post-Mortem Scorecard I can paste into Google Sheets.Output CSV only (no commentary).

Columns:Trade ID, Setup Quality (1–5), Entry Discipline (1–5), Risk Plan (1–5), Exit Quality (1–5), Emotion Control (1–5), Biggest Lesson (1 line), Rule Upgrade (If-Then), What I’ll Do Next Time (1 line)

Use the trade above + your analysis to fill the row(s).

Paste the CSV into a new tab called: Scorecard.

Step 4 — Lock in one improvement (1 minute)

Pick one “Rule Upgrade” and make it real:

• Add it to the top of your Sheet as: “My Rule This Week:”
• Example: “IF I don’t know my stop before entry, THEN I don’t enter.”

This is how you compound skill — one small rule at a time.

The Payoff

You now have a repeatable system that turns trades into progress. Instead of “I got unlucky,” you’ll know: Was my setup solid? Did I size correctly? Did I follow my own risk plan? Over time, this reduces emotional trading and upgrades your investing process — which is the only edge beginners can reliably build.

Transparency & Notes for Readers

• AI can’t read your brokerage account. You must paste the trade details (and optionally a chart screenshot).
• If you upload images/files, free tiers may have daily upload limits — keep it simple and paste text if needed.
• This workflow helps you improve decision-making, not predict markets.
• Educational workflow — not financial advice.