@ninepixelgrid I could not agree more. Tried Claude Code for a while, came back to Cursor. And they have made some pretty massive improvements to it lately. Feels like they really want my business too, while Claude is annoying I'm using up their inference.
AI agents are about to negotiate leases, book travel, and close loans on your behalf.
But they hit a wall the moment financial verification is required. A hard pull. An SSN. A 3-7 day underwriting process. The whole flow breaks.
Here's the thing: traditional credit bureaus were built for a different era. They measure past debt behavior — a proven, reliable signal. But they weren't designed for real-time, agentic transactions.
The next evolution of financial identity looks like this:
→ Bank-verified income + cash flow via Plaid (consumer-permissioned data)
→ An AI model that outputs repayment probability for a specific loan amount, not a generic score
→ Zero-knowledge proofs so a lender can verify "this person can afford $2,400/mo" without ever seeing your bank account, SSN, or income number
→ User-controlled — lives in a wallet on your phone, shared only with consent
No PII stored at a central repository. Near-zero breach liability. Fully FCRA and GDPR-friendly by design.
And when AI agents transact on your behalf, they call this like an API. Financial clearance as a primitive for the agentic economy.
Who is building this?
Me: Hey, this csv is a lot of data, make sure you use python or something so you don't hallucinate your way through this.
Claude: I'm good with CSV data - numbers and patters are my strength...
Bash phython 3
@cathrynlavery I love OpenClaw. I think it completely changed how we think about AI. But a lot of other tools have copied the functionality now. I might come back to it once the platform is more stable and mature.
I have been a pretty heavy Cursor user for a while now. Decided to give Claude Code a try. It is good, but doesn't feel any more powerful or capable. And the limits are minuscule using the same model. What am I missing?
Anyone else find chatGPT 5.4 extremely long-winded? When a few paragraphs would suffice it keeps going and going, often just reframing what it has already said.
@elonmusk Big upgrade over the prior version. It’s cool seeing the multiple agents collaborate. Helped catch up to similar models so Grok remains my daily driver for everyday tasks and questions.
Seeing plenty of complaints where OpenClaw breaks itself after an update.
This happened to me last night where I couldn’t even talk to the agent to fix itself.
I finally thought to plug Cursor into it, and it was able to read all the files/logs, quickly identify what was broken and fix it.
This will be my go to resolution as the tech is moving so fast I am sure this will happen again.
@boringmarketer It is over. But it will take a long time for everyone to catch up. Companies will take years to adjust their teams, systems, and tools accordingly.