Huge congrats to my client UpDoc AI and to Dr. Sharif Vakili. First-ever FDA clearance for a patient-facing clinical LLM. This is what the future of proactive medicine looks like.
UpDoc just landed the first FDA clearance ever for a medical device built on patient-facing LLMs, with an AI that manages a Type 2 diabetic's insulin between doctor visits.
What it's cleared to do, within the parameters a physician sets first:
- Call or message the patient
- Change the insulin dose off glucose-monitor readings or what the patient reports
- Order follow-up tests
- Log every decision in the patient's medical record
It followed a clinical trial the founders (both physicians) ran at Stanford Medicine.
UpDoc CEO Sharif Vakili: "It's like I've deployed a concierge doctor to follow up with the patient."
AI is bringing some much needed acceleration to a slow, reactive healthcare system
This is the exact loop we need in the firm. Every time a reviewer fixes something, the system actually gets smarter instead of just moving on. That’s real leverage. #taxtwitter
⚙️ Behind the build of self-improving tax agents with Codex
We co-built Tax AI with @ThriveHoldings around tax prep workflows so when reviewers fix any errors, Codex can trace the failure, improve the system, and test the change before it ships.
https://t.co/otI9oYp2A6
The more I watch AI move into tax and accounting, the more I think the real shift is not replacing judgment. It is removing the pile of admin sitting in front of judgment.
Clients do not care if a CPA personally chased every missing W-2 or reviewed every payroll notice from scratch. They care that the work is right, fast, and calm.
That is where firms can actually win with AI.
For CPAs, the second brain is client tax history, prior year judgment calls, notice responses, checklists, memos, and preparer notes in one place. Not fancy AI stuff. Just making sure the firm does not forget what the firm already knows.
https://t.co/bQi0EF1GTu
The biggest competitive advantage no one is talking about in AI:
Having a "second brain" internal knowledge base containing all your company data.
By 2027 every company will have one. They'd be insane not to.
Call transcripts. Slack messages. Emails. Meeting notes. SOPs. Everything.
This matters because data is the foundation of every AI deployment. The companies that get their context organized will ship AI projects faster than everyone else.
I don't think people realize how much of a massive competitive moat this is.
There are 2 ways to play this:
If you're a business owner: start building your second brain now. Get a Hermes agent and use the built-in LLM-wiki skill. Feed it any source.
It tags, organizes, and adds it to your wiki. Within 90 days it's a competitive asset.
If you're looking for a startup idea: this is one of the biggest opportunities in the AI age.
It's the AI equivalent of selling picks and shovels during the gold rush.
You can charge a fortune to organize their data because the ROI is enormous once the project ships.
Go build it.
@AmrTawfik160 ElevenLabs voices and honestly they sound way better than I expected. Two hosts named Charlie and Sarah. Charlie is quick and sharp, Sarah asks the questions you actually want answered. Feels like a real show.
My morning commute hit different since I had OpenClaw turn my Twitter feed into a daily podcast. Fresh audio every morning based on what actually happened in my feed overnight. Two AI hosts, three topics, ready before I get in the car. This stuff is genuinely wild.
OpenClaw + the X API is a match made in heaven
It's an incredible way to stay on top of what's trending in the world, plus even make better content and go viral
In this video I show you how to use OpenClaw to connect to the X API and 2 use cases that will blow your mind:
Every CPA firm already has the raw folder. It's just called the shared drive. Client notes from 2019. Email threads with half the context. Research memos nobody reads anymore. The hard part was never collecting it. It's making it queryable. #TaxTwitter
https://t.co/Z4pRHQwMfO
Karpathy's "second brain" concept in 60 seconds:
1. Three folders (raw, wiki, outputs). That's the whole architecture.
2. One CLAUDE.md schema file tells the AI how to organize everything.
3. Dump your bookmarks, notes, and articles into raw/. Don't organize them.
4. One prompt: "Compile a wiki from raw/ following CLAUDE.md." Walk away.
5. Ask questions against your wiki. Save answers back. It compounds.
6. Monthly health check catches errors before they stack.
No Obsidian or complex plugins. Just desktop folders and a schema file.
Full walkthrough + free skill that builds it for you in 60 seconds in the article below.
@mercer70638 Exactly. The boring middle is the opportunity. Every firm has 30–50 tiny manual touches around each return that nobody has mapped because they’re “just how we do it.” AI doesn’t need to replace the preparer to make that workflow meaningfully better.
#TaxTwitter we don’t need AI to fully automate tax prep yet.
The first wins are around the edges: client follow-ups, customized planning tools, client dashboards, workflow automation, onboarding, engagement letters and compliance.
That alone can make a firm meaningfully better. I’m excited to see what we can build.
It's never been easier to do everyday work with Codex.
Choose your role, connect the apps you use every day, and try suggested prompts.
Codex helps with everything from research and planning to docs, slides, spreadsheets, and more.
I built a magazine for our CPA firm using Claude and Codex.
Physical version for the waiting room. Digital version for clients.
It’s called The Lazar Ledger.
It’s about building better edges around the firm:
better education, better client experience, better follow-up, better tools, better reasons for clients to think of you before there’s a problem.