Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx
claude code is 500k LOC. very small codebase for arguably the most impressive harness of our time. if you think LOC = productivity you’re not a real engineer.
@NWischoff@NWischoff we’ve built exactly this @AddyAI_ & already have a product in market with paying customers. Recently just trained a custom mortgage underwriter model with 20,000+ underwriting guidelines. Just sent you a demo in your dm
It’s midnight of DAY 1 of the AI x Finance hackathon in San Francisco. AI engineers are building GenAI mortgage tools.
An interesting interaction captured in frame is one participant asking
“are you going to win”
and the other replied
“No we’re just having fun”
And this interaction captures it perfectly!
We succeeded in getting top AI engineers excited about mortgages. What a time to be building in this space.
We now have >$5000 in cash prizes for the AI x Finance hackathon this weekend at @AGIHouseSF
(register in replies👇)
If you're building tools for historical pattern finding, loans, or secure online payments, this is THE event for you. We have prizes based on topics so there is >1 way to win!
Thanks @DRWTrading, @AddyAI_, @mcttrading, LenderTookit for investing 💰 to enable events like this
Join us at @AGIHouseSF January 25-26 for the AI x Finance Hackathon!
Merging top bay area ML talent with commodity trading & mortgage experts to crack open a new era of trading bots, automated loans, and more!
RSVP in replies 👇
Thanks @DRWTrading, @AddyAI_, @mcttrading, LenderTookit and @demi_network for supporting
My AI prediction for 2025 is I think Generative UI will be a huge frontier! It’s not the most sexy but the most practical.
Every time you interact with an LLM, there’s a ton of parsing to show you a best UI. Claude Artifacts, Open AI Canvas, and Vercel’s V0 are doing okay at this but it sucks.
This is a problem we come across many times when designing user flows @AddyAI_ . It will be just magical for the AI to generate a user interface in real time and embed it in the chat.
I’m rooting for who ever is building this.
@SullyOmarr Modify to Decision llm + there’s a fine line between when the llm calls are finite. My take is if the calls are finite the prompts certainly have to be dynamic to be considered an agent. If not, it’s just another RPA
good to see that work in agentic workflows that folks like @josh_bickett were working on w/ self operating computer is being mainstreamed but let’s not forget this stuff has been pioneered by others for over a year.