How AI Agents Are Creating an Infinitely Scalable Workforce

2026-06-18 · 4 company · 5 thematic

sentiment

4 bull0 bear0 neu

speakers

neema ghamghami sorat

Neema Ghamghami Sorat is the CEO and co-founder of Blend, a digital lending platform processing ~20% of US mortgages and home equity loans. He previously worked at Palantir after graduating from Stanford, where he studied computer science. He grew up in Ohio after his family immigrated from Iran.

joe lonsdale

Joe Lonsdale hosts American Optimists. He co-founded Palantir Technologies and founded venture firm 8VC. He previously hired Scott Wu at an early company and references his experience building Palantir's forward-deployed engineer model and government technology partnerships.

quote

“End of last year, you know, our average engineer was doing six pull requests per month in December. That's already doubled 4 months later. The best people are doing 80 to 120 pull requests a month.”

—neema ghamghami sorat

now playing · $BLEND

$BLENDbullish· medium· posneema ghamghami sorat

Internal AI agents double engineer productivity to 80-120 PRs/month at Blend

Blend tracks AI-driven productivity gains: average engineer PRs/month doubled from 6 to 12 in 4 months, with top performers hitting 80-120. Ambient intelligence agents auto-fix bu…

$BLENDbullish· high· posneema ghamghami sorat

Blend's Autopilot AI agent cuts mortgage closing from 29 to 21 days in first deployment

Autopilot, launched March 2024, acts as an agentic workforce handling unstructured mortgage documents and rules in real time. First customer analysis shows closing process reduced…

$BLENDbullish· high· posjoe lonsdale

Blend trades at 3x revenue despite profitable AI agent deployment in mortgage lending

Blend is profitable with increasing margins, growing ~10% annually, and has deployed AI agents (Autopilot) to 70 of 300 lenders with $10M+ early pipeline. The market cap of ~$350M…

$GOOGL···bullish· lowneema ghamghami sorat

Gemini 2.5 Pro quietly leads on agentic banking evals per Blend's model testing

Blend runs hundreds of historical evals against new models; Gemini 2.5 Pro (2024) was 'quietly really good' for agentic tasks. Blend plans to publish an agentic banking index trac…