Gutenberg Lab — Morning Paper

How Valon Rebuilt the $13 Trillion Mortgage Servicing Industry by Becoming a Servicer First

Featuring

Angela Strange (General Partner, a16z) × Linda Du (Co-founder & Co-CEO, Valon) × Andrew Wang (Co-founder & Co-CEO, Valon; former Soros analyst)

Bottom line

Rebuilding mortgage servicing is not a technology problem but a change management problem.

3-Line Summary

Valon rebuilt the $13 trillion U.S. mortgage servicing market by first becoming a licensed servicer themselves, proving the technology at scale, then selling the software to incumbents. The founders spent 3-5 years obtaining licenses across 50 states, manually encoded 1960s-era federal and state regulations into code, demonstrated the technology with a $200 billion portfolio, then sold the servicing business to Carrington to pivot to pure software. Generative AI now enables previously impossible automation of complex edge cases, portfolio-specific customization, and agent-orchestrated servicing, achieving 3x efficiency and 70-80% operating margins compared to incumbents.

3 Key Points

1. 18 hours/day for 6 months to encode regulations: Andrew Wang spent COVID lockdown reading federal regulations (RESPA, TILA, FDCPA, GLBA, TCPA) and 50 states' mortgage, debt collection, foreclosure, privacy, and escrow rules, "refactoring the entire legal code base" into abstract frameworks that handle all variations. He recalls "there was a point at which you could cite almost every single state's regulation, and I can tell you what code number it was." This suffering became their moat.

2. Minimum 3-5 years for licensing: New York, Fannie Mae, and Freddie Mac all require profitability before approval. Valon started with non-QM loans (state licenses only), but California requires government approval unless you're a real estate broker. Every license demands 5+ years of experience across collections, customer service, payment processing, delinquency, and foreclosure. New York servicing approval took a record-breaking 3 years (never got origination license); the application went missing, and Andrew physically went to the mail room to track it down before COVID lockdown.

3. 10% of market in 6 months: Within 6 months of officially going to market as a software company, Valon signed over $200 million in deals. NewRez (Rithm) is transferring 4 million loans (10% of the entire market) to the Valon platform. Linda Du notes, "Andrew and I didn't decide that we hit product market fit. The market decided." All deals are outcomes-oriented: "For every dollar that our customer earns or saves, that's when we earn as well."

Editorial Perspective

In regulated industries, the moat of AI deployment is not technical superiority but the willingness to endure pain. What Valon demonstrated is the paradox that the hardest path—obtaining licenses yourself, taking operating risk, proving it works, then selling—is the only viable approach. This principle applies to law, healthcare, financial infrastructure, and any regulated domain. However, as the founders acknowledge, "Six years ago I would have told you this is a technology problem. What I know today is that this is a change management problem." Even if you win on technology, you cannot deploy without moving thousands of people with different objective functions. In the AI era, the scarce resource is not implementation capability but the ability to lead transformation.

Source: a16z Podcast "How Valon Rebuilt a $13 Trillion Industry From Scratch" (October 7, 2026)

https://a16z.simplecast.com/episodes/how-valon-rebuilt-a-13-trillion-industry-from-scratch-0mcVWvYe

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

How much should you tell your AI agent? — The convenience-humanity trade-off

Featuring: David Pierce (The Verge, host) × Allison Johnson (The Verge, staff writer)

Bottom line: AI agents perform linearly better with more data, but that convenience comes at the cost of opportunities for humans to think, choose, and experience friction themselves.

3-line summary

Two Verge journalists field-tested three AI agent products — Meta's Muse, OpenAI's Dots, and startup Instinct — that execute tasks on your behalf when you message them. These systems request access to email, calendar, credit cards, and health records. While the convenience was proven, fundamental questions remained: can you trust companies like Meta with that much data, and does removing daily friction erode something essential about being human?

Three key takeaways

1. Interface determines use case, not capability — The most constrained tool, Instinct (SMS-based), proved most useful day-to-day. Being "everywhere and nowhere" (accessible via smartwatch, CarPlay, any messaging context) beat dedicated apps requiring context-switching. Muse worked better for sit-down tasks at a computer.

2. Frictionless commerce is "too good" — a feature and a bug — Muse stores credit cards via Stripe and completes purchases when you type "go ahead." Johnson called it "problematic for my bank account, but the product working as intended." Pierce asked Muse to search for cars; it sent listings twice weekly, then asked "do you want me to buy this one?" He said yes, and it started loading the purchase. "My god. Didn't you just buy me a car?"

3. Rambling for 10 minutes beats precision commands — Johnson used Dots voice mode to redesign her personal website, talking for 10 minutes with unstructured feedback. The AI transcribed and executed well. Pierce noted this contradicts his instinct to give short, clear, turn-by-turn instructions: "Evidence suggests rambling lets AI distill better output."

Editorial perspective

The dishwasher analogy cuts to the core. "Would I gain something morally by doing my own dishes after every meal?" Probably not. AI agents are the latest step in a long automation trend, not fundamentally different. But this step is faster and demands more data than before. As Johnson put it, "We live in a society where everyone has dishwashers; insisting on hand-washing doesn't work unless you're Bill Murray." The question isn't whether to automate but which tasks and how fast. That answer is left to individual intuition. Pierce's advice — "Only give it one credit card number and keep careful watch of that credit card" — is a practical line you can draw between convenience and vigilance.

Source: The VergeCast "How much should you tell your AI agent?" (October 6, 2026)

https://www.theverge.com/the-vergecast

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --

Hip-Hop's Living Witness on How to Keep Creating Across Age, Technology, and Platform Shifts

Featuring

Joe Rogan × LL Cool J (hip-hop pioneer, actor, entrepreneur)

Bottom line

Creative success requires not talent but humility, discipline, and the belief that dreams have no deadlines.

3-Line Summary

LL Cool J, who has remained at the forefront of hip-hop for over 40 years since its dawn, shared the secrets of sustained creativity. He insists that "fame is a consequence, not a goal," sharpens his ego through jiu-jitsu, and trains his mind through reading. Even in the AI era, he argues that the "human spark" is what gives art its value, and technology is merely a tool.

3 Key Points

1. Mastery is transferable across domains

LL states: "Once you know what success looks like and what it takes, you can apply that to other missions." Martial arts, music, business, life—the stages of mastery (struggle, discipline, breakthrough) are universal. This is a counterpoint to modern "niche specialization" culture.

2. Hip-hop was aspirational, not ostentatious

"It wasn't an open display of wealth because it wasn't at the wealth level... It was 'I have $50 in my dresser drawer. I got a gold chain. I couldn't eat six months ago.'" Gold chains were a toughness signal ("I can wear this in this neighborhood"), fundamentally different from today's "portfolio rap."

3. The internet is a script; real life is improvisation

"If you've been living on the internet, you've been dealing with a script the whole time. Then you get in real life and you're awkward." He quotes Mike Tyson: "People got too comfortable saying stuff and not getting punched in the face for it." Online validation doesn't translate to real-world competence—a structural insight explaining Gen Z's social anxiety epidemic.

Editorial Perspective

LL's narrative carries the perspective only someone who has navigated four platform shifts can possess. From vinyl to streaming, MTV to TikTok—technology changes, but the core of art remains the "human spark." His answer to AI-era anxiety is not to fear technology but to keep believing in one's own creativity. It's also a quiet rebuttal to the culture of giving up based on age.

Source: The Joe Rogan Experience "#2561 - LL Cool J" (October 6, 2026)

(unknown)

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --