Using LLMs to handle insurance claims Also ❤️ Geopolitics 🗺️ Bikes 🚲 Trains 🚈

Havertown, PA 🇺🇸
Joined March 2008
Awesome history of insurance by @daschreiber : 🇨🇭1713 Bernoulli defines Law of Large Numbers 🏴󠁧󠁢󠁥󠁮󠁧󠁿1686 Lloys of London founded, redefines insurance ⛵️1799 Insurance Companies own most data 🚂1899 Insurance Companies own most data 🖥1999 Insurance Companies own most data 🤖2019 ❓ – at St Louis, MO
2
6
I thought the local ski area ( @springmntn ) had a super confusing map, Claude just drew me a new one.
63
Looking back on it, I think I'll always remember Claude 3.5 Sonnet, and Claude Opus 4.5. Those were the days that changed it.
51
Just hit 1,700 miles on my Tern GSD e-bike (1 year of riding). Can't say enough about how much I love this thing
1
63
Lol, MapQuest is currently the number five app in the Apple App Store for iPhone in the USA.
1
2
113
Apparently the reason for this is that MapQuest has refused to rename Lake Ontario to Lake America and has taken a hard stance on this, unlike Google Maps and Apple Maps.
25
According to my Twitter feed, the 2028 presidential election will hinge on: 🔴 Marco Rubio: 🏈 Pro College GameDay Platform 🔵 AOC: 💁‍♀️ Millennial Single Woman Platform
Me explaining to my girl why I can’t go to her sisters wedding in October
2
2
432
Interesting to bump this up against Polymarket on Aug 14th 2026.... JD Vance is no where on the timeline, looking way overvalued.
41
this was breaking my brain to find the missing month
months, ranked (nyc-specific): 1. september 2. october 3. june 4. may 5. december 6. july 7. november 8. april 10. march 11. january 12. august
115
Found this picture from July 24th, 2010, when I was coding HTML / CSS / Javascript by hand in the CMU Computer Lab... what a time
1
10
1,012
andrew furman 🗺️ retweeted
What else do I need to add?
102
297
68
4,918
263,739
Is no one talking about how Kimi K3 is only serving at 24 tokens per second? This seems not good.
3
119
Could be google glass … or Snapchat Specs could be as popular as AirPods in 10 years 🤷‍♂️
1
2
408
This is wild value if true. In the last two months, I've paid $200/month ($400 total) for my OpenAI plan. If I paid for these 7 Billion tokens at API prices, that would have cost me $60k
1
95
with Claude Fable 5 (and GPT 5.5 to some extent) there is now more need to balance sleep vs pushing agents 24/7
2
92
I used to refuse to work with people who would say "Lets re-write it in Rust" this morning (while I was on a run) .... I one-shot prompted Claude Fable 5 to rewrite one of my CLI tools that I use daily in Rust (works perfectly 👀)
1
85
My current lifemaxxing stack: 5am — Wake up, start a load of laundry, run 3 miles 6am — Make the kids’ lunches and breakfasts 7am — Get the kids up, fed, and dressed for school 8am — Bike the kids to preschool, pick up Trader Joe’s groceries 9am–5pm — Crank work   1pm — Lunch break and walk outside with my wife 5pm — Bike the kids back from preschool, maybe stop at playground if nice out 6pm — Dinner and clean the kitchen 7pm — Bath time and story time 8pm — Kids in bed, check emails, watch YouTube 9pm — Get in bed, read The Economist 10pm — Lights out Wouldn’t change a thing.
My current lifemaxxing stack: - 4:30am wake up - Read classic books - 3 hours creative work before 8am - Lift/run 6x/week - Eat single ingredient foods - Present time with fam/friends - 20-min evening sauna - Wild Roman skincare routine - 8:30pm bedtime Wouldn't change a thing.
4
5
160
Mission accomplished Pentagon: "We have successfully degraded the operational capacity of all national security supply chain risks" 🇮🇷Iran: infrastructure destroyed 🤖Claude: experiencing increased errors on Opus 4.6
1
77
Ok the local models are here Qwen 3.5 small models feel so slick on Locally AI app on iPhone
1
1
1
189
andrew furman 🗺️ retweeted
Tbh this is a much better way to experience “claw”.
ClawPhone is alive. I installed OpenClaw on a $25 phone and gave it full access to the hardware. I haven’t explored everything yet, but it’s a cool formfactor for agents to run on.
6
3
3
92
18,222
Are we all switching from Claude Code to OpenCode with GLM 4.7 ? Feels 10x faster with minimal intelligence loss
1
2
169