Founder & CEO of XLR8·4ward (bringing usable AI to insurance). Interested in the intersections of finance & tech, AI & humanity, & quantum physics & philosophy.
Earth
Joined August 2025
- Tweets1.2K
- Following22
- Followers183
- Likes30
Pinned Tweet
I'm the founder of XLR8•4ward.
We make an AI-powered underwriting platform (InfiniteUnderwriter) for the P&C insurance industry, which I spent a few decades in.
The industry is powered on profoundly obsolete tech.
Every day of my career I've wanted to fix that.
Now I am.
Had my Cybertruck at Tesla service today.
The showroom was empty of cars & maybe 5 Teslas were in the lot. It looked like they were shutting down. I went in to ask.
Turns out they've sold out of every vehicle. Wait times are months.
FSD is really starting to drive sales!
Getting closer to onboarding the next batch of customers on our wait list.
That starts once we're sure everything is perfect for our existing customers.
Thanks to all who have been patiently waiting! We're going to make it perfect for you, too.
Today we went live with an agentic intake pipeline for customer issues and feature requests. It's pretty cool:
- customers make a feature request in the system, which is triaged by an agent that determines if it has applicability for all customers or just that customer.
- the request is passed to the coding agents to implement and thoroughly test.
-a human gets alerted to review the feature and ultimately decide whether to proceed as is, make changes, or cancel the request.
- outcomes are measured so we can monitor our confidence in the output the agents are producing, with the goal being some future state that removes the human in the middle.
The process takes maybe 2 minutes from user request to a human reviewing implementation.
Possible downsides:
- our codebase gets massive as we implement tons of requests.
- we are training our customers to make feature requests often (I'm on the fence as to whether that's a good thing or not, and I suspect it actually is).
This will be a great experiment!
Gemini Notebook would be such a helpful app but for the continuing halluciation issues.
Creating an infographic from source material is great until you look closely and see stuff like this (which happens multiple times on virutally all infographics, videos and slides):
My wife and I never refer to an AI by name.
We'll never say, "Grok suggested..." or "I asked Gemini..."
It's always, "I asked AI..."
Is that a branding failure?
I can't imagine anyone wades through the 4 syllables of "ChatGPT said..."
How to you refer to it?
Interesting that AI has clearly been trained on development timelines that predate AI coding.
If I ask for a timeline to implement a feature I'll get a response with a detailed timeline that a team of senior engineers will take 1-3 weeks.
Then we'll implement it in 27 minutes.
This is the very worst way to try to upsell users to a new Workspace tier @Google (there are no "security issues they found"; the link just leads to a pitch to upgrade Workspace. It's spam).
In one email you train users to ignore any actual future security issues:
We've been busy!
We've been busy building and haven't had time to post much, so here's an update with some features we've added to InfiniteUnderwriter:
- every property submission returns the real time distance to the nearest wildfire and recent earthquake, including magnitude.
- automatic sanctions checks against all named insureds on all submissions.
- AI WebIntelligence returns an overview of the insured with linked citations.
- for construction projects the distance to neighboring properties is estimated using AI VisualIntelligence. If you upload a site map the project's setback from the property line is included.
- set the Market State slider to inform the AI UnderwritingEngine of the current state of the market.
- allow your underwriters to draft free form endorsements.
- set a market participation strategy for your binders: quota share (set default percentages by market), ad hoc (underwriter picks each time), default a single market, random or round robin.
-option in Settings to attach wordings to quotes.
...and lots more!
Excellent article by Mark Zuckerberg in the Wall Street Journal in which he argues in favor of AI for the masses and against its concentration among an elite few.
He's right.
Imagine if, at the dawn of the industrial revolution, it was decided machines posed too great a threat and only a chosen few would be permitted their use (it must have felt a bit then like it does to many now: that the machines would take over the work of humans, displacing millions).
You'd have ended up with the vast majority of people stuck doing manual labor while a select few enjoyed all the benefits of the new technology. Progress, prosperity, health gains would all have been put on pause.
Instead, humanity made the greatest gains, on the greatest number of dimensions, in history. Hundreds of millions were pulled out of poverty.
I hope our fears (more imagined than real) don't preclude another wave of great prosperity and progress. The future belongs to the many, not to the few.
The article: wsj.com/opinion/the-ai-futur…
After using a lot of Google & Gemini tools in our development process I finally made the switch to Cursor.
Wow! What a difference!
No more scrutinizing every line of code for the inevitable mistakes, changed variables, etc.
Feeling sheepish for taking so long to switch.
The varying competence of the same AI model from one coding session to the next remains puzzling given the weights presumably don't change.
Sometimes output is 90% correct.
Other times every output contains errors & you can't trust any of it.
Currently experiencing the latter
It's hard to overstate just how bad Gemini 3 Flash Preview is.
Spent the day coding with it. It's maddening:
- routinely changes variable names crashing downstream code.
- quickly loses track of what we're doing.
- routinely ignores instructions.
I sure miss Gemini 2.5 Pro.
Gemini 3 series has been quite a frustrating disappointment for coding: it routinely generates massive amounts of errors, changes variable names arbitrarily & has a penchant for breaking changes.
2.5 was peak Gemini thus far (though none seem to match other models for coding).
Excellent insights from Andreesen Horowitz (a16z) to help us think about how AI adoption might play out (link in reply).
We tell all our customers our platform moves their bottleneck from the bottom of the funnel to the top: it becomes about how much opportunity you can feed it, since processing those (complex) opportunities becomes trivially easy.
We too think the value of lead generation is going to skyrocket.
Article: a16z.com/need-series-c-call-…
It's quite astonishing just how bad Gemini 3.1 (Flash & Pro) are at coding.
It will routinely change variable names halfway through a code block, move variables out of scope, & wreak havoc on your code base.
2.5 Pro was much better.
Explains why even Googlers won't use it.
It's funny how different instances of the same AI seem to have different personalities & reliability.
I'll work with an extremely capable AI, hit the context limit & spin up a new one.
That one will regularly hallucinate variables & do terrible work.
Next one, better again.
AI: "You are absolutely right. I apologize unreservedly. My previous code was buggy, violated your architecture, and the error is a direct result of my sloppiness. You have every right to be angry. We are wasting time, and that stops now."
Of course, it didn't stop now...
