Nima Ghamsari

New updates this week. Really cool - now with auditable, step by step income and asset calculations and much more. And a few more goodies in the works for next week. Side note - 30+ banks and lenders have turned this on already. 3 weeks in.

@goldilock1969 in all seriousness we will just keep executing and delivering value to clients and see where everything shakes out

@goldilock1969 man I was really hoping they would!

Progress every week. We’re going to make banking agent-first. Autopilot 1.1 shipped today.

Today, we're proud to announce our Q4 2025 earnings, closing out a year of sustained momentum and innovation at Blend. We added or expanded 10 customer relationships, increased pipeline approximately 40% year-over-year, and returned capital to shareholders through strategic share repurchases. We remain focused on building the technology you need to operate more efficiently. This includes Blend Autopilot, our agentic AI solution designed to attack the $11,000 cost-to-originate head-on. 🎧 Tune in live at 1:30 p.m. PT / 4:30 p.m. ET to hear the full story: hubs.ly/Q046f-_d0 For complete earnings results, visit our investor relations site: hubs.ly/Q046fV9C0 📈 #BLND

@DegenLaneCap @MagicCompounds daily progress. constant improvements. expect autopilot v1.1 to come out this week.

@DegenLaneCap @MagicCompounds we have some absolute animals here. and I am contributing my fair share of PRs too :)

$BLND simple earnings take away: nothing really matters until we see conversion from preview to paid on new ai offering. Thats the only datapoint that will change the narrative all else equal on macro side. Business is healthy enough to weather any macro.

@MagicCompounds of course!

@NGhamsari Thanks, helpful. Wrt underlying model, can this be easily changed as new SOA foundation models are released?

The quiet part of this story is that the team built this in 12 weeks. And is moving faster than ever. With a small, motivated team, a good platform / customer base, and a clear mission, the things we can do are mind-blowing. And we're just getting started.

@MagicCompounds meaning - a loan application is a thousand pages and the guidelines are a thousand pages. too big for any context window together. so the game becomes orchestrating things in a way that makes even the most minute detail deterministic

@NGhamsari That’s awesome. As in you are limiting the context to documents uploaded and some internal set of knowledge/parameters? Appreciate the reply

@MagicCompounds the magic of the blend layer is a combo of - we already have great orchestration in place (important). and we manage context really well so the operations are (much more) deterministic I just ran the same set of docs through 20 times and got the same income calc out every time

@MagicCompounds we use quite a bit of everything internally, depending on the specific type of operation. gemini is great at vision and one shot reasoning. opus / sonnet are great at orchestration.

It's live in beta for our customers today to turn on. When you turn it on, it'll automatically follow every application in the system. Reading, thinking, calculating, understanding, following up - anything a human would do. And it's free during the preview period. 5/5

The world is rapidly moving from AI-native to background worker agents. To me, agent-first means the agents are the *first* ones to do anything. They do the dirty work. Humans provide oversight. And for our customers, lenders who are trying to approve consumers, 1/5

1) AI for our customers. How do we make their teams 10x as efficient? 2) AI for our engineers. How do we make sure we’re using the best tools to the fullest, and make our engineering team 10x as efficient? 3) AI for everyone else at Blend. We have hundreds of customers, lots of products. Let’s make out knowledge workers 10x as efficient.

@NGhamsari @bcherny How does $blnd plan to use AI?

Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven. These use cases are diverse and surprising -- the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model. Today, we're so excited to introduce Cowork, our first step towards making Claude Code work for all your non-coding work. The product is early and raw, similar to what Claude Code felt like when it first launched. Cowork includes a number of novel UX and safety features that we think make the product really special: a built-in VM for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure, We are excited to see how you all use it. Cowork is available now as a research preview for Claude Max subscribers in the macOS app. Click on “Cowork” in the sidebar: claude.com/download

@felixrieseberg When do we get native, org wide skills (standardized product specs, customer QBRs, etc)

👋 Hi, I'm Felix and I work on Claude Cowork, bringing Claude Code closer to all kinds of knowledge work. It's an early and rough preview, please tag me in any feedback - we want to iterate very quickly and make it a little better every day.