In this week’s Inman Tech Roundup, Mainstay’s enterprise tools go downmarket, and a VC says the playbook for seed-stage tech startups is changing. Robot mowers are getting better, and a neighborhood intelligence proptech vendor is targeting hallucinated, AI-generated facts.
Also, a reminder: The Inman team is in Las Vegas this week for Blueprint 2026, the leading conference for commercial real estate owners, operators, proptech founders, technologists and investors.
Blueprint is a subsidiary of Inman, and this year’s conference features plenty of residential real estate and brokerage action that we will be covering from the floor. Stay tuned to Inman throughout the week for updates.
Here’s what you need to know:
Opendoor alums bet on scaling down, not up
Mainstay, the residential real estate “system of record” carved out of Opendoor in 2024, has raised more than $18 million and acquired Truelist, an AI-powered listing tool for brokers and agents. New investors Parker89, Stackpoint, Alpaca VC and FJ Labs joined existing backers Khosla Ventures, Era Ventures and Clocktower Technology Ventures.
The company says it’s facilitated more than $12 billion in transactions and run over 2 million agentic workflows for 95 percent of the 60 largest residential investors. But until now, that infrastructure hasn’t reached brokerages and agents directly.
Carey Armstrong
Truelist changes that, adding a listing-data and workflow layer meant to clean up fragmented MLS feeds and extend Mainstay’s pricing and automation tools downmarket.
Mainstay was spun out of Opendoor as an independent company in August 2024, after its founding team spent more than a decade building and operating residential real estate technology and infrastructure at scale.
Former Opendoor President Dod Fraser leads the company alongside President J.D. Rasmussen and COO Michael Dulfer, both longtime Opendoor executives. TrueList founder and CEO Carey Armstrong is joining Mainstay as Chief Strategy Officer. Armstrong previously served as Vice President of Premier Agent at Zillow and co-founded and served as COO of Tomo.
“Mainstay and Truelist have been attacking the same broken system from different angles: Mainstay on behalf of investors; Truelist on behalf of brokers and agents,” Armstrong said. “Joining forces means we stop duplicating effort and start compounding it. I have enormous respect for what Dod and the team have built, and I’m excited about what we can do together.”
For more about the acquisition, check out Inman’s recent interview with Mainstay chief strategy officer Carey Armstrong.
What this means for agents
For real estate agents, Mainstay’s acquisition and capital raise signals that the pricing and automation tools built for institutional players are being packaged for smaller shops that could never afford enterprise real estate tech.
Whether that translates to genuinely useful tools or just another platform to log into will depend on how well Truelist’s listing workflows integrate with the MLS systems agents already juggle.
What 25K applications taught 1 VC
Aaron Golbin, co-founder and general partner at LvlUp Ventures, recently wrote on Crunchbase that the playbook for seed-stage startups has shifted after his firm reviewed roughly 2,500 recent applications and nearly 25,000 over the past year.
Non-dilutive growth capital is now competing with equity for companies with clear revenue visibility, Golbin says, with his firm issuing financing checks “on a near-weekly basis.” Distribution has also become something founders design before launch rather than bolt on afterward, building into marketplaces such as Shopify or embedding products into Slack, Teams or payroll workflows.
Aaron Golbin
Golbin argues “learning velocity,” not speed, is now the real differentiator. He added that the most fundable companies can describe their business in one sentence and defend what they’re choosing not to do. About 82 percent of applicants still in business a year later had a strong go-to-market plan in their deck, he said.
On artificial intelligence, Golbin sees an implementation gap. Seventy-eight percent of LvLUp Ventures applicants use AI in some form, but the firm favors founders treating it as infrastructure rather than experimentation.
What this means for proptech founders
For proptech founders, the message is that a good product isn’t enough.
Golbin’s framework rewards startups that architect distribution into brokerage, MLS or lender workflows from Day 1 rather than building a standalone tool and hoping agents find it.
It also suggests the non-dilutive capital option is worth exploring for proptech companies with predictable revenue (think subscription or transaction-fee models), since equity-only raises are increasingly seen as the slower, costlier path.
Robot mowers are cheaper and smarter now
Robot mowers have moved past the smart-home novelty phase, according to real estate agent and product tester Brandon Doyle. In a post on the National Association of Realtors website, Doyle said he has tested 15 models over two years.
The biggest shift is that newer machines such as the LUBA 3 AWD combine LiDAR, network RTK and AI vision instead of relying on one navigation method, cutting down on the stuck-mower problems that plagued earlier versions.
Brandon Doyle
Entry-level models now start under $1,000, down from a roughly $2,500 floor a few years ago, with strong mid-range options running $2,000 to $3,000. That’s a meaningful shift for agents who’ve watched buyers hesitate over large lots or sellers worry a sprawling yard reads as a maintenance burden.
Edges remain the weak point. Fence lines, garden beds and tight corners still often require manual trimming, even on machines with trimmer attachments. Doyle recommends all-wheel drive for any property with slopes or uneven terrain, and LiDAR or camera-based navigation over RTK-only systems on lots with heavy tree cover.
What this means for agents
For real estate agents, this gives you a concrete counter when a buyer balks at a big lot or a seller frets that their yard looks like too much upkeep. You can potentially point to sub-$1,000 entry models and same-afternoon setup instead of just gesturing at “smart home features.”
The practical value is in matching the pitch to the property. Know to flag all-wheel drive for sloped yards and LiDAR-based navigation for lots with heavy tree cover, so the recommendation holds up if the client actually buys one.
Local Logic takes aim at hallucinated facts
Location intelligence company Local Logic has launched a hosted MCP server that connects AI applications directly to its verified location data. The company aims to reduce the hallucinated neighborhood facts that plague general-purpose language models.
The company tested seven models on 490 neighborhood questions across 46 neighborhoods in 39 cities, scoring more than 153,000 individual claims against live data rather than a fixed answer key.
Gabriel Damant-Sirois
With access to Local Logic’s data, models surfaced up to 2.3 times more verified facts and cut contradicted claims by 2 to 4 times. GPT-5 nano hit 97.6 percent accuracy when grounded, edging out an ungrounded GPT-5 at roughly a seventh of the cost per answer, the company said.
“AI models are very good at producing plausible answers, but the location facts that influence real estate decisions are not reliably captured in their training data,” said Gabriel Damant-Sirois, co-founder and chief product officer at Local Logic. “The Local Logic MCP gives models a direct connection to measured information at the moment they need it. That allows teams to improve accuracy without depending on increasingly large and expensive models.”
The MCP is already live in Infinityy’s AI chatbot and in Local Logic’s own consumer tool, Ask Local Logic. The company says commute-routing and climate-risk data will be available through the MCP next.
What this means for agents
For real estate agents, this is less about a new consumer tool and more about the plumbing underneath the AI products they’re already using. If a chatbot or listing assistant claims a neighborhood is “walkable” or has “good schools,” the accuracy of that claim increasingly depends on whether the vendor grounded it in something like this rather than the model’s own guesswork.
It’s worth asking any proptech vendor pitching AI search or neighborhood Q&A whether they’re using verified location data or just trusting the model. The gap here (82 percent for web search versus 97.6 percent grounded) is large enough to matter for anything client-facing.
