Alan Grosheider - MKS Ventures

10/30/2026

CoStar is paying $730 million to see nearly every roof in the U.S.

This week, it agreed to buy Vexcel, an aerial imagery company that captures more than 60 million images a year.

The plan is to use AI to turn those images into tens of billions of data fields about property condition.

Roof wear.
Additions that never made it into the public record.
Storm damage.
Tree limbs hanging over the shingles.

All visible before anyone sets foot on the site.

That's a real shift. It also creates a problem almost nobody is talking about.

Once a model flags a bad roof on 400 properties, someone has to send 400 qualified people to confirm it.

I built and sold a nationwide real estate inspection company. Ordering inspections was never the hard part.

The hard part was everything after the order. Finding the right vendor in the right market. Getting the bid back. Setting the date. Chasing the report.

Aerial AI won't remove that work. It will multiply it.

CoStar is paying $730 million to see nearly every U.S. roof.
Detection is about to get cheap.
Dispatch still runs on email. (We're fixing that, by the way.)

Read more...

AI imagery will make property condition knowable before anyone visits a site. That moves the bottleneck from finding problems to coordinating the people who confirm and fix them. On October 27, 2026, CoStar Group agreed to buy Vexcel, one of the world’s largest aerial imagery operators, for about ...

10/27/2026

Anthropic brought in $16.3 billion in six months.

Now it’s going public.

Most founders will read that and think, "Bigger is better."

I read the rest of the filing.

$518 billion in compute commitments over roughly a decade. Most of it locked in.

Nearly a quarter of revenue from two customers.

That’s the price of building frontier models. Decade-long contracts, capital only a handful of companies on earth can raise, and one exit big enough to make the math work: the public market.

Almost nobody reading this should play that game.

The other game looks like this.

The same week, Nous raised $17M. Its AI agent runs bills and admin for almost 250,000 UK households.

Mave raised a $28M Series A. It automates security investigations for Fortune 50 companies with a team of about 30.

$45M combined. Less than one-hundredth of one percent of Anthropic’s commitments.

Both rent the model. Both eliminate expensive work. Both have an obvious list of companies that would want to buy them.

Here’s what I tell founders at MKS Ventures:

Pick a workflow with a price tag.

Name your acquirer in year one.

Raise for the next milestone, never for the headline.

Your startup has much better odds if you build to sell.

https://mks.ventures/anthropic-ipo-lessons-for-founders/

Anthropic’s IPO filing shows $16.3B in six-month revenue and $518B in compute commitments. Why founders have better odds building an AI startup to sell.

10/25/2026

Walmart opened its store to AI agents. Amazon locked them out.

On October 21, Walmart launched a gateway that lets approved AI agents search its inventory and buy on a customer’s behalf.

Shoppers can set spending limits, require approval for every order, and revoke an agent’s access at any time.

Amazon went the other way. It blocks third-party shopping agents.

Every real estate vendor will soon face the same choice.

A national environmental firm recently sent us 40 sites across Texas that needed asbestos surveys.

Each site meant finding a licensed inspector, confirming insurance, getting pricing, scheduling access, and chasing the report.

That’s coordination, and AI agents will take over exactly that.

When a buyer’s agent looks for an inspector, it can only hire firms it can find and verify.

The appraisers, environmental firms, and inspectors who open up first will get the work. The ones who block agents will wonder where it went.

I'm betting on Walmart’s approach.

Which side is your firm on?

Read more...



https://blue222.com/walmart-now-welcomes-ai-agents-amazon-blocks-them-which-strategy-wins-in-real-estate/

10/20/2026

Four announcements in eight days. None of them were about home search.

eXp let agents plug ChatGPT, Claude, and Gemini straight into their transaction data.

Intuit agreed to buy Sequence, whose AI agents run an entire revenue workflow from contract to collection.

TrueFocus launched Title Lens to screen title and mortgage packages for fraud before closing.

Proof put a protocol into production that lets an AI agent prove who it works for and what it’s allowed to do.

Put them together, and you can see where AI in real estate goes next.

Read more...

https://blue222.com/the-next-ai-revolution-in-real-estate/

AI in real estate is moving from property search to closing deals. Four October developments show how AI agents will coordinate due diligence.

10/13/2026

1 in 7 home sales now falls apart before closing.

I've watched the same pattern kill commercial deals for years.

The agent orders the Phase I environmental assessment late. It's treated as a box to check.

The report comes back days before closing with a problem. One that was sitting there the whole time.

Now the buyer needs a Phase II. Soil and groundwater testing. Weeks of waiting. Unknown liability.

The lender gets nervous. The closing date slips. The deal dies.

The property never changed. Only the timing did.

Residential deals die the same way, just faster. The structural engineer gets called on day 8 of a 10-day inspection window, and the buyer is staring at a scary sentence with no dollar figure and a deadline.

People will negotiate a $6,000 repair. They walk away from an unknown number with a clock on it.

I wrote up 7 ways agents can keep deals alive through the inspection window. Link in the first comment.

Agents: what's the latest you've ever ordered an inspection and still closed?

https://blue222.com/1-in-7-home-sales-now-fall-apart/

Deals are falling apart at the highest rate in years, and most of them fall apart in the same place: the inspection period. After years of coordinating inspections nationally, I'm convinced the defect is rarely what kills the deal. The calendar is. A problem found on day 3 is a negotiation. The same...

10/11/2026

I overheard a meeting this morning that felt like it was from 1995.
Sadly, it was not.

I wasn’t trying to listen. I overheard it on a speaker call.

The leader was giving the team a demo of their new management tool:
A color-coded spreadsheet! What the hell?!

Here’s the thing. I’ve overheard this same meeting many times. Same group, every morning, at least 30 minutes.

They go around and tell each other what they’re up to, who’s doing what, and which customers are coming in that day.

Every bit of that already lives in a calendar or a system somewhere. They’re reading it out loud to each other. 30 minutes a day is 125 hours a year. Per person.

The scary part is how normal this is. Most companies still run exactly this way.

AI can run that entire process right now. An agent reads the calendar, the CRM, and the job list overnight. Everybody has the plan for the day before they sit down. Nobody recites anything.

You don’t need a big project to get there. Pick one simple process. Build a simple AI workflow around it. Then pick the next one.

Your competitors are going to. Some already have, and their teams had a 30-minute head start over you this morning.

If a color-coded spreadsheet is the new technology you’re rolling out to your team, you are losing. Fast.

That’s how you go the way of the dinosaur. Change or die. It really is that simple now.

What is your team still saying out loud every morning that a system already knows?

10/09/2026

Want a 10x return on an investment in Anthropic's IPO?

They need to become a $20 trillion company.

Nvidia, the most valuable company on Earth, is worth about $5.65 trillion.

Anthropic is reportedly targeting $2 trillion at its IPO. The business is real. Revenue grew 12-fold last year.

The 100x early returns are already spoken for. They belong to investors who bought in early: 2024 or before. Every round since has cut the remaining multiple by half or more.

So I ran the math an accredited investor should run before investing.

How do you get exposure to OpenAI or Anthropic today, and what do you own when you do?

How long does the money stay tied up?

And what are better alternatives for making AI investments?

https://mks.ventures/anthropic-could-be-worth-2-trillion-to-10x-your-money-it-would-need-20-trillion/

A 10x on Anthropic's reported $2 trillion IPO needs a $20 trillion company. How accredited investors get in, how long capital is tied up, and the math.

10/08/2026

AI cut a 3-hour job to 40 minutes.

The firm that did it is still stuck.

I had lunch with two accountants who use more AI than most companies I know. A tax tool. A note taker. A CRM. Practice management software. Claude.

Every tool works. None of them talk to each other. So a person carries the data from one to the next by hand.

Brokerages are in the same spot with different logos.

Compass built the missing layer for itself. In July, it put an AI assistant on top of its whole platform. Its own brands get it first.

Nobody is going to automatically build that layer for your brokerage.

So build it yourself, one boring job at a time. Own the controller. Rent the AI that plugs into it.

Here's how it works and where to start.

10/02/2026

Forward-deployed engineer job postings jumped 1,000% this year.

These are engineers who sit inside a customer's business and make AI actually work there.

OpenAI, Anthropic, Google, and Microsoft are all hiring them.

Why would the best model companies in the world need people on-site?

Because the hard part of AI lives in the customer's messy data, the permissions, and the exceptions nobody ever wrote down.

AI made the code cheap. It made that last mile expensive.

That's the bet we're making at MKS Ventures.

Build small. Use AI to automate everything internal so headcount never becomes the strategy.

Embed engineers inside client workflows, where the real work happens.

Turn every deployment into reusable product.

Then sell to the larger player who would rather buy that workflow than spend two years building it.

Buyers are already moving. CB Insights counted 266 AI acquisitions in Q1 2026, up 90% year over year.

The model has real risks. Embedded engineering can slide into a dev shop with dev-shop margins.

Here's how we guard against that in the full article.

https://mks.ventures/is_startup_investing_changing/

Is startup investing changing with AI? Why small, capital-efficient startups embedded in client workflows are what acquirers are buying in 2026.

09/29/2026

$2.5 billion to $10 billion. In one month.

That's Instinct, an AI agent that's still invite-only.

Sequoia, Benchmark, and Coatue just put $1 billion into it.

Investors are paying for AI that finishes the job. Plans the trip. Orders the groceries. Calls the business.

What stopped me was the sign-up.

No app. No form. No dashboard.

You give them your cell number. Everything after that happens over text.

They also built a concierge service that calls businesses on your behalf.

We found the same thing at Blue222. By accident.

After our March relaunch, some of our biggest clients skipped the dashboard and started sending spreadsheets with one instruction: handle it.

So we did. Every one of those projects trains Ocean, the inspection AI controller we have in beta.

Meanwhile, most of real estate is bolting chatbots onto old portals.

A chatbot can tell you when the inspection is.

Someone still has to schedule it.

Soon the first request for your service will come from someone's AI agent.

Could it reach your business today?

Here's where AI is headed for real estate and what brokerages, lenders, and vendors should do now.

https://blue222.com/stop-adding-chatbots-to-real-estate-workflows/