Would you take medical advice from an AI? It turns out that a lot of people are doing just that and I gave it a try. I have a problem with both my ring fingers where the tendon catches something and the finger get’s stuck temporarily: it’s called trigger-finger. My massage therapist, Chris Howard, has been helping with cupping, but I thought an AI consultation would be a good next step. A few months ago I raised the issue with ChatGPT, but didn’t get much help. Now, it’s August 2026, so time to try again. Here’s my prompt:
I have a problem with trigger-finger in both hands. I have a wrist brace that I’ve used for wrist pain caused by computer use, but that doesn’t help because it doesn’t extend to the fingers. I’ve found that placing my hand flat under the pillow gives me less pain in the morning. Can you research and find an appropriate wrist brace, particularly for my right hand. I’m thinking that at worst, a baseball sliding glove might help, but the medical profession must make something better.
I’ll paste Claude’s response below. The answer was pretty good. I bought the cheaper Featol brace and it helps. I also bought the Oval-8 Finger Splits and they work very well and I can wear them much more often. So far – So good. A solid improvement and the tendons in my fingers are getting a chance to heal.
Here’s Claude’s (Opus 5) answer:
Your instinct is right, and so is the pillow trick — it’s actually the mechanism behind the standard treatment.
Why your wrist brace does nothing
Trigger finger is the flexor tendon catching as it passes through a tight pulley at the base of the finger (in the palm, not the wrist). A wrist brace immobilizes the wrong joint entirely. The pillow helps because during sleep your hand curls into a loose fist, which is precisely the position that aggravates the tendon and explains why mornings feel worst. The American Society for Surgery of the Hand also notes that when you lie down, fluid pools in the hands, and that swelling makes locking more likely overnight. Flat under the pillow keeps the tendon extended and lightly compressed. You’ve been improvising a night splint. Dr. ArthritisMedical News Today
For nights: a resting hand splint (this is the pillow, engineered)
This is the category you want if multiple fingers are involved in both hands. It’s a padded pan that holds the wrist and knuckles in neutral with the fingers laid out flat.
Specific options, roughly cheapest to nicest:
OTC 2385 Soft Resting Hand Splint (~$30–40) — rigid molded plastic shell with a washable fabric cover; it holds fingers, wrist, and thumb in a functional position and is sold in left/right versions.
The Comforter Splint (3-Point Products, ~$60–70) — this is the one hand therapists actually reach for. It supports the wrist and MCP knuckles in neutral with adjustable stays, while the finger joints stay free. Softly padded, which matters when you’re wearing it eight hours a night for two months.
Amazon has a swarm of near-identical Featol / BraceAbility / Fanwer resting splints in the $25–35 range. They’re stroke-rehab products repurposed for this, and they work, but the padding is hit-or-miss.
For days at the keyboard: a small finger splint, not a hand brace
A resting splint makes typing impossible. Here the evidence has shifted toward the smallest effective splint:
3pointproducts
bttnusa
stockxmedical
The Oval-8 is a thin figure-eight plastic ring worn at the middle knuckle. A randomized trial comparing it against a custom MCP-blocking splint found pain reduction in both groups, but greater with the PIP (Oval-8) splint, along with better function scores — largely because patients could tolerate wearing it far longer during the day. The older, larger MCP-blocking splint (a palm-based piece with a ring around the finger base) still has good data behind it — 77% of subjects got at least partial relief — but it blocks intrinsic grasp and people take it off.
Practical note: Oval-8s are sized per finger, and fingers swell. Buy the Sizing Kit or a multi-size pack rather than guessing a single size; each splint also rotates 180° to gain a half size.
On the baseball sliding glove
Not useful. Sliding mitts are padded for impact protection but exert no extension force — they let the fingers curl freely, which is the exact motion you need to stop. Good thinking, wrong physics.
How to actually use it
Nightly wear is where the value is, and the course is longer than most people expect: typically six to ten weeks of consistent nightly use, with some improvement showing at two to three weeks. Taper gradually rather than stopping cold. Half-hearted intermittent use is the main reason splinting “fails.”
Two things worth flagging
When ordering, resting hand splints are hand-specific — confirm you’re getting Right and check the sizing chart against your palm circumference, since these run small and a too-tight splint at night is worse than none.
More importantly: bilateral trigger fingers at 70 is worth an actual evaluation rather than self-management alone. A certified hand therapist can fabricate a custom MCP-blocking splint molded to your hand in one visit (usually covered by insurance), and a hand surgeon can offer a corticosteroid injection, which has a considerably higher success rate than splinting for cases that have progressed to locking rather than just clicking. Splinting is the right first move and costs $30 instead of several hundred — but if six to eight weeks of disciplined nightly wear doesn’t move the needle, that’s the signal to escalate. I’m not a physician, and multiple digits involved on both sides is the kind of pattern a hand specialist should look at directly.
[TL;DR] Positive news about AI in science, engineering, medicine, and math deserves attention. An early morning idea turned into a ChatGPT Site in 90 minutes.
I’ve written software since 1971. Since the mid-1980’s I’ve heard that a new computer program is going to write the code from now on instead of programmers. That’s left me with skepticism about what AI might do. After spending time with them, here are a few early impressions.
AI can write working code
I started in March of 2025 coding with the chatbots. The CLI and VS Code based tools were only beginning to appear. But given a fairly simple problem in writing SQL the chatbots were able to produce working code. ChatGPT and Gemini were able to write some code but Grok went past that and found a bug in the code it was give to to adapt and fixed it. Now all the foundational models are good at code. The harnesses, like Claude Code, Codex, Cursor, and JetBrains, have taken that further and are very good at it. I documented that experience here:
Testing still matters
The AI tools are confident even when wrong, so testing matters more than ever, not less. The good news is that debugging skills are still useful. However, I’m working on some projects were I’ve not looked at the code and have interacted with the AI by doing my own testing and then going back and requesting changes like a typical user might.
Image Generation Isn’t so Great
Given the logos of Gemini, Grok, Claude, and ChatGPT, I asked each:
Using the four logos provided make an image for the theme of AI-Assisted coding. Include a few lines of code somewhere in the image, maybe in the backround. Make sure to produce a square image.
The results were pretty bad. They all butchered at least two of the logos. Most of the output wasn’t square. The image at the top of this post is from ChatGPT and here’s a try from Grok.
Grok First Try
If you’re exploring the same territory, I’d love to compare experiences.
Hello, and welcome. I’m Andy Novick, and this is my new corner of the web. It’s going to be a place to write about whatever makes me curious, share links to my projects, and keep a record of what I’m learning along the way.
For a long time my online life was scattered across a handful of places: code on GitHub, photos on SmugMug, SQL articles on Novick Software, and an older personal blog called Andy’s Guy Stuff. This site is my attempt to pull those threads together into one home, with a blog at the center of it.
The blog is the heart of the site. For now, much of what I write here will fall under the category of AI Discoveries. These will be notes from my ongoing experiments with AI tools and AI-assisted coding. I expect it to wander into photography, programming, and the occasional thought about a book, a movie, or something happening in the world.
If you want to find my other work, the footer below has links to everything. Thanks for stopping by.
I’ve tried to keep up with what AI makes possible. Here’s a rundown of the projects I used AI for during 2025.
T-SQL coding with AI
In March, I decided to see if AI would write T-SQL, the programming language of SQL Server. None of the programing tools were geared to SQL, I used an AI chat interface. I created a scenario for a Simple ETL project somewhat similar to projects I have been doing for my consulting client. Grok did the best job and even found and fixed a bug that was in the prototype code I had hacked together. Pretty impressive and I did a YouTube about it here: https://www.youtube.com/watch?v=G7eHQKerVMk The code is published on GitHub here TSQL_Codegen1
Book Review of CO-INTELLIGENCE: Living and Working with AI By Ethan Mollick
This was posted on Linked-in. I wrote the review myself with only grammar and spelling help from AI. It got more attention than I expected. Here’s the review.
Writing books for my granddaughters
I wrote two books for the granddaughters, ages 7 and 4. The first one, Saving the Constitution, coincided with a trip to New Hampshire for the Independence Day holiday. It was about the two of them as teenagers living in Philadelphia on July 4th, 1776. They had to help Thomas Jefferson complete the wording to the Declaration of Independence. ChatGPT helped with the writing, story line, and pictures. AI struggled to keep maintain consistent character appearances across illustrations, but the results were still acceptable. The Papa Andy character came out looking like Ben Franklin in one picture and like a 40 year old farmer in another. The girls loved it.
Later in the summer when I had a plan to turn a freezer box into a rocket ship for them to play with, I wrote Astrid and Frieda go to Mars. Now in early adulthood, they had to travel together to Mars to deliver the first human born on Mars. This time I think I did a better job of prompting and the story and pictures were more consistent. ChatGPT’s Studio Ghibli style had just come out and that worked very well.
Tech Support
This is a lifesaver. I’ve been using computers for 55 years, mostly as a developer. While I have extensive programming experience, my knowledge of technical environments such as cloud services, Linux, and networking is more fragmented. Now I get technical support as good as Neal, who supported the last office where I spent a lot of time. He’s very good, for a human, on the topics he knows like Windows, networks, and storage, but he doesn’t know every topic. Grok and Gemini seem to know all topics and lay out step-by-step instructions to solve each problem or implement each new function. They make mistakes and are willing to correct them, just like Neal. Unlike Neal, they won’t do the implementation for me. Maybe next year.
Trip Planning
My wife and I took two trips out west this fall the first in October to Utah and the second in November to California. Gemini was key to making our trip to the parks in Utah very special and very easy. The key was laying out the route: Zion, Bryce Capital Reef, Arches, Canyonlands, Monument Valley, and Page, which has Horseshoe Bend and the slot canyons. Then it helped us fill in the details for each day. If felt similar to the personalized service we once received from old-time travel agents. If you’re thinking of a trip, don’t fail to ask your AI for help.
Cookoff1
We hosted the annual Christmas party for four families who share German traditions. To make it fun, this years party included a cooking contest. Cookoff1 is a Python web app built with Flask, developed in PyCharm and it’s AI tool, Junie. There are three pages: Setup, Voting, and Results. I ran it on my desktop machine and exposed only to the home network. The users got connected with either their phone or iPad and entered their votes. I brought up the results page on the living room TV for the big reveal. You can see it in the GitHub repo.
NoSave
This is a small windows app with a single multi-line textbox written in C# using Visual Studio 2026 Community Edition and GitHub CoPilot. It never saves text pasted into it. I use it as a place to paste passwords so that they’re not in the clipboard history. Windows can save a history of clipboard strings, but that leaves you vulnerable to another unrelated app grabbing them, So I’ve turned that feature off. I could use Notepad to save a password for a few minutes, but that gets saved, sometimes across reboots, so I don’t use that. NoSave just holds the password. It includes a feature to discourage screenshots, though it isn’t fully reliable. NoSave is a work in progress, there is no public repo yet.
Daily News Summary
After cataract surgery I found it uncomfortable to read the newspaper (Wall Street Journal) with any set of reading glasses, so I decided to try something different. I canceled my subscription ending a 60-year habit of reading newspapers, going back to forth grade. For about 6 years I also delivered the Standard Star, New Rochelle’s newspaper. I replaced the paper with e-mail subscriptions across the political spectrum: from CNN and Axios to The Hill and Just the News, and read them in the morning. They newsletters duplicate each other so it became a bit tiresome to read them all. I decided to try getting a news summary using Gemini. Using the chat interface, at 6:30 AM and 5:30 PM it reads my Gmail account for the day’s news and produces a pretty reasonable summary of the ten most important events. I instructed it to cite each source used for each summary, and it did so. Then it started sneaking in other sources, with citations. I’ll have to address that in the prompt and I’ll be adding summaries for AI news, sports, and business/investment news next year. Or, I could just read the Babylon Bee.
In-app AI
In the second half of 2025 many applications have introduced AI assistants, and I’ve been taking advantage of them. Photoshop is one. Removing people where I don’t want them is fantastic. Masking the sky without spending 10 minutes building a precise mask! That’s amazing. I love it. The photo club I belong to isn’t as excited and it’s been the subject of some new rules about contest submissions. I’ve also used the AI in Word, Excel, Notion, and several other programs. I’m hoping my tax prep software will include an AI assistant next year.
More Chats
I did lots and lots of chats this year. I made pretty good use of ChatGPT, Gemini, and Grok, and occasional use of Perplexity and Claude. For me, chat has replaced most search. I expect it to continue to do so next year.
2026?
What’s ahead? More programming for me. I’ll flesh out the news app and NoSave into more useful tools. I’m going to devote some time to research topics, such as diet and metabolism, using NotebookLM. Writing some more is on the table. I’m not sure about what. I’ll try and find a technical SQL with AI topic to present.
I listen to some AI enthusiasts like Moonshots and Nate B. Jones on YouTube. They’re telling us that AI is on an exponential curve that’s going to make it more intelligent at an increasingly faster pace than in 2025. AI has changed faster and faster and now the rate of change is increasing. Second derivatives to the moon! Will we get our robots? Maybe full self-driving? Hold on to your hat!
In early 2025, I was almost too late in reading CO-INTELLIGENCE. The book was completed in late 2023 when ChatGPT was the undisputed king but the field of Artificial Intelligence (AI) and Large-Language-Models (LLM) in particular is evolving so quickly that it often sounds dated. However, it has value in the long-term perspective that we all need when adapting to the changes in work and society brought about by AI. For that, I’m glad that I took the time to read it this week (March 2025).
The first valuable observation, and one that’s now almost trite, is that the AI you use today is the least capable AI you’ll ever use. I’ve seen several technology revolutions, and AI is the one that develops the fastest, with changes coming every week. PCs, LANs, and the internet all evolved with new changes every few months or years. Now, there is new material to absorb every week.
The most important element of the process that isn’t changing is us: the human. The book adds the most value by defining the human-AI interaction. All of us will have to find ways to work with the new AI tools. They are changing rapidly; we are not.
My most important takeaway from CO-INTELLIGENCE is the breakdown of tasks into four groups:
Just-Me Tasks – that need to be handled by a person
Centaur Tasks – A part human/part-machine combination hands the task from one to the other
Cyborg Tasks – Where the human and AI merge into one entity
Automated Tasks – That can be handled by just the AI
The most important chapter is the sixth, where these distinctions are drawn. In using AI, I’ve seen all four types of tasks and the breakdown is going to help me as I approach each new job.
CO-INTELLIGENCE: Living and Working with AI is worth the read. At least for now. In a few years or months the field will evolve and we’ll need a fresh analysis.
