Most people think AI is about replacing jobs. But in one of the most honest moments from the Ventura Labs podcast, Seby (Resi, Subnet 46) frames it differently:
AI is here to replace the systems that have been wasting everyone’s time.
The “BS,” as he calls it.
And if there’s any industry where BS is basically baked into the foundation, it’s real estate.
Real estate is slow, expensive, opaque, and unnecessarily dependent on middlemen. And Resi is betting that decentralized AI can change that.
Not by building a prettier Zillow, but by rebuilding the real estate stack from the ground up.
Watch the full Ventura Labs podcast below:
Resi’s Pivot: From Collecting Data to Predicting Value
Subnet 46 originally started with a simple thesis: collect massive real estate data → use it later to train models.
It sounded logical. But after speaking with Bittensor builders, subnet owners, and actual real estate players, Resi realized something important. Data collection is step one… but it’s not the product.
So they pivoted.
Instead of spending months scraping and hoarding information, they moved straight to what actually matters: a daily AI appraisal competition.
The idea is simple: let miners compete on what the market values most, which is accuracy.
How the Incentive Mechanism Works (In Plain English)
Every day, Resi runs a competition.
Here’s the flow:
- Miners submit appraisal models publicly on Hugging Face
- After the submission window closes, Resi pulls around 500 recently sold properties
- Each model predicts the value of those properties
- Validators compare predictions to real sale prices
- The model with the lowest mean error wins the reward
One competition per day means miners can’t cheat by hardcoding known prices. And the outcome is astounding as models get better, because the money forces them to.
Why This Approach Works Better Than the “MLS Data Layer” Plan
Resi’s first idea was to build a huge dataset, similar to a decentralized MLS (Multiple Listing Service).
But Seby explains that miners are too capable for that kind of narrow work. Bittensor miners aren’t random hobbyists; they’re full teams with serious machine learning firepower.
So rather than asking miners to scrape data as a job, Resi is asking miners to build models.
And to build models, miners naturally have to collect data anyway. So the data problem solves itself.
Seby puts this in one sentence:
Don’t build the database first. Build the competition, and the database becomes a byproduct.
The Problem With Real Estate Today: The MLS Is a Dinosaur
Seby’s critique of the MLS is sharp.
He describes real estate evolving in stages:
- Paper MLS (realtors literally held the book)
- Digital MLS (Zillow reads listings and displays them)
- Tokenized MLS (blockchain becomes the new backend)
The key argument is that real estate isn’t ready for “decentralization” because the infrastructure is still stuck in manual processes.
And that manual process is why:
- closing takes 90–120 days
- transactions cost insane fees
- appraisals and inspections are expensive and slow
- trust is enforced by middlemen, not systems
The Real Shock: Inspections Aren’t Even Used in Appraisals
This part of the podcast is where things get interesting.
Seby explains something that sounds ridiculous until you realize it’s true: inspections happen after price agreement.
Meaning the buyer and seller agree on a number first.
Then inspection happens.
Then renegotiation happens if problems are found.
And only after that, the bank runs the appraisal. Not to find “fair value,” but to find liquidation value.
So the system is backward:
- The buyer agrees to buy
- Inspection reveals issues
- Price gets renegotiated
- Bank appraisal comes in low anyway
- Buyer pays the difference out-of-pocket
The “appraisal” isn’t even designed to help buyers. It’s designed to protect lenders.
Resi’s Roadmap: Appraisal → Commercial → Inspections
Right now, Resi is focused on residential appraisals.
Next steps:
- commercial appraisals
- inspection competitions
- then a combined system where inspection feeds into appraisal
The long-term goal is a pipeline where narrow AI models can complete an entire real estate transaction.
Not with one model, but with specialized models stacked together.
Model + Agent Competitions: The Next Big Upgrade
Resi’s current setup is model-based, but Seby hints at the next leap: model + agent competitions.
Meaning miners won’t just submit predictive models.
They’ll submit agents that can:
- research market conditions
- pull niche local signals
- refine predictions based on context
In simple terms:
Today’s model predicts value. Tomorrow’s agent will go hunt for evidence. That’s how Resi expects to push accuracy beyond the “state-of-the-art.”
The Accuracy Target: 95% Isn’t the Dream — It’s the Industry Baseline
Seby says something important: 95% accuracy is not a fantasy. It’s literally where the industry already operates.
That’s what banks are comfortable with. That’s what lenders already price risk around.
But the difference is that traditional appraisals cost $500+ and take 1–2 weeks.
Resi wants the same accuracy instantly. And then push further.
Their real target is:
- 95% with model competition
- 99% when inspection + appraisal merge
He argues that 100% is unrealistic because markets can become irrational through bidding wars, panic buying, or overpaying.
But 99% would be enough to change the industry.
The Bigger Vision: Resi Wants to Power a Tokenized Real Estate Economy
Resi isn’t just trying to be “an AI appraisal tool.” The roadmap is bigger, consisting of three models:
- Appraisal
- Inspection
- Title verification
Seby calls these the only three things you really need for a real estate transaction: who owns it, what it’s worth, what condition it’s in.
Everything else is overhead.
The Three Products
- an oracle (real-time valuation)
- a smart contract framework
- a liquidity pool
“A Mortgage You Don’t Have to Pay Back” — The Wildest Claim in the Episode
This is the line that sticks. Seby describes a future where billboards advertise:
“Get a mortgage you don’t have to pay back.”
It sounds like an exaggeration until he explains the mechanism. The thesis is built on one idea:
tokenization unlocks shared ownership.
Instead of one person carrying the entire burden of buying a home, ownership becomes fractional:
- the resident is the managing partner
- investors hold shares
- appreciation and rent create returns
The buyer doesn’t need to fully “pay back” a traditional mortgage. They just need to maintain the asset, pay utilities/taxes, and share upside.
In theory, young people gain access to homes without getting crushed by decades of debt. And investors get exposure to real estate appreciation and yield.
The Leverage Layer: Real Estate as DeFi Collateral
Seby takes it even further.
He describes a pooled liquidity system (let’s say $100M in stablecoins) where tokenized real estate can be used as collateral.
So instead of illiquid ownership, you get something closer to DeFi leverage:
- tokenize property
- borrow against it
- loop exposure
- earn rent yield
- use yield to pay interest
He compares it to funding rates in crypto, but with one key argument:
real estate doesn’t drop 50% overnight.
So leverage becomes “safer” than most crypto strategies. That’s the financial engine behind his excitement.
Resi’s Business Pitch: “We Cut Your Roadmap by a Year”
This is where it becomes clear why real estate companies might actually care.
Sebie explains tokenization platforms today face a brutal reality: they need to build everything from scratch:
- oracle systems
- smart contracts
- audits
- compliance
- appraisal workflows
- inspections
- title verification
All of it.
And they still have to run a business.
Resi wants to package the entire infrastructure layer so tokenization platforms can focus on:
- acquiring sellers
- acquiring investors
- closing deals
Resi handles the backend intelligence and verification.
Closing Thought: Resi Isn’t Just AI, It’s a New Real Estate Operating System
The core message from Seby is not “AI will replace appraisers.”
It’s bigger:
Real estate is broken because the systems are outdated.
And if you fix the appraisal, inspection, and title stack with AI + crypto infrastructure, you don’t just improve efficiency…
You create a world where:
- home ownership becomes accessible
- liquidity becomes instant
- transactions become one-click
- and real estate becomes programmable
That’s why Seby isn’t selling “a subnet,” he’s selling a future.
And his advice at the start of the episode is the real thesis:
empty your mind of assumptions and think as big as possible.
Because the biggest wins don’t come from replacing people.
They come from replacing the BS.
Be the first to comment