Arjun Lalwani

cofounder & ceo @ralo_hq (YC). Building ai native mortgage broker. Prev product at google / youtube.

[Anton Osika @antonosika]

building a company in 2026

[Arjun Lalwani @arjunlalwani]

@bertgodel Congrats!! This is awesome

[Arjun Lalwani @arjunlalwani]

@arfrank I love how this chesky quote has become mainstream

[Arjun Lalwani @arjunlalwani]

@gargi_kand @AkritiKeswani

[Arjun Lalwani @arjunlalwani]

@3xykop @NatlMortgagePro DM’d you

[Arjun Lalwani @arjunlalwani]

Had a great time chatting with Katie Jensen @NatlMortgagePro about how AI is changing mortgage origination.

The big shift: AI can lower the cost of producing a mortgage by automating the manual work, while loan officers evolve into true advisors.

At Ralo, our early cost to originate is already 4x lower than the industry average.

If that trend continues, the biggest winner should be the borrower: lower costs, faster closings, and ultimately better rates.

https://t.co/UPS9jUaxLN

[Arjun Lalwani @arjunlalwani]

@anamika__x the replies confirm my theory 🤝

[Arjun Lalwani @arjunlalwani]

@R_Mohr @travisk Someone needs to study the correlation between excellent founder communication skills and company success

[Arjun Lalwani @arjunlalwani]

@VadimStrizheus How much time did you spend researching / copying vs testing new hooks? Is it 80% do and 20% research or vice versa?

[Arjun Lalwani @arjunlalwani]

@ardent__dev You run ads

[Arjun Lalwani @arjunlalwani]

The more we build AI for mortgages, the more I think the moat in vertical AI will be evals, not models.

Every weird edge case you encounter becomes a regression test.

Over time, you build thousands of domain-specific scenarios that answer a simple question:

Can our system safely handle this without a human?

A recent paper, MortarBench, is a great example. Researchers built 188 mortgage origination test cases across 47 real-world loan officer questions. Even the strongest frontier model only reached 77.1% exact-match accuracy out of the box.

The models will keep getting better.

But the companies that know exactly where they fail and have the eval infrastructure to prove when a workflow can safely be automated will have a huge advantage.

Great read: https://t.co/RpDg2hlJ89

[Arjun Lalwani @arjunlalwani]

@TTrimoreau Bad sign. Now you need to validate demand and find product market fit

[Arjun Lalwani @arjunlalwani]

@Suhail Best pizza: Ceres (get a res)

[Arjun Lalwani @arjunlalwani]

@BoBrainerd Yep, more broadly you are inspired by all the branding / decor of the new restaurants & CPG brands taking off here

[Arjun Lalwani @arjunlalwani]

@Alfred_Lin This is exactly what Pixar was to Disney

[Arjun Lalwani @arjunlalwani]

@pri_the_ninja we got the same advice. The team slide was #2 and converted much better and establishes credibility quickly

[Arjun Lalwani @arjunlalwani]

@nikunj In a deeply regulated space and we feel this everyday. Over time, you just learn to enjoy the pain :)

[Arjun Lalwani @arjunlalwani]

@_sun_yi Storytelling attracts visitors. Copywriting converts them to customers.