Opendoor CEO: How I Saved a Billion Dollar Company From Bankruptcy
The Knowledge Project Podcast
Arriving to a Company Near Collapse 0:00
Keith Rabois describes flying to San Francisco to take over Opendoor, a company months from bankruptcy. His wife told him not to come home until he had a plan to break even, and had a mattress delivered to his office because she knew he would sleep there for days. He requested a full list of employees, every contract, and every payment made in the last twelve months. He found that companies fail in similar patterns: good people leave, control of innovation slips away, general and administrative costs rise, and the company ends up fighting its own customers and partners.
How the Company Had Rotted 1:32
Opendoor had quietly become a company run by professional consultants, structured to avoid standing out. When the board chair sent him a bland announcement document about his hiring, he edited it, someone reverted his edits to keep it inoffensive, and he wrote in bold that whoever wrote the original no longer worked there. The two PR consulting firms named in the document resigned loudly, saving the company a month of fees. Reviewing expenses, he found the largest external cost was a consulting firm that had pushed Opendoor to offshore jobs, bloat administrative costs, and cut engineering, which he says is exactly what happens when a company has no interest in creating real value, something he believes happens broadly once founders and large shareholders are no longer present to protect a company from being drained by what he calls professional leeches.
Rebuilding Around the People Who Stayed 4:00
On his first Monday, he told the fully remote company that everyone would be back in an office within a week or they were no longer welcome. He found that Opendoor had rewarded process over outcomes, and many people left once ownership and results became the expectation. He also discovered a small group, roughly a dozen individual contributors, who had been quietly keeping the company alive, and the rebuilt company was built around them. He changed the careers page from language about being a happy, caring workplace to a warning that the job would be hard, aiming to attract only highly committed people.
Truth Over Comfortable Lies 6:31
He describes how organizations drift into telling each other comfortable, polite lies, and argues a founder's job is to keep delivering blunt, first-principles truth. Opendoor, in his view, had developed a victim mentality, blaming the macroeconomic environment and its own partners and customers for its problems. His response was to create what he calls repelling forces, deliberately pushing out people who disagreed with his direction so new, aligned people could join. He singles out competent but not mission-aligned employees as the most dangerous type in any company, because their political skill lets them rise despite working against the company's real goals.
Changing Defaults Violently 9:01
His core turnaround principles are that defaults must be changed abruptly rather than through gradual change management, because gradual plans at Opendoor had failed for years. He insists leaders must state clearly what they expect of employees rather than relying on generic, feel-good mission statements. Meetings at Opendoor are limited to sharing information rather than debating, and disagreement is treated as a sign of respect, not hostility. He describes a personal rule from his time at Shopify called Say the Thing: raise problems immediately, criticize the issue rather than the person, and keep repeating your point until you feel heard.
AI as an Efficiency Exoskeleton 15:01
Rabois argues Opendoor is now effectively a new company because AI has built what he calls an exoskeleton around employees, making them three to four times more efficient. He notes that the last time Opendoor bought as many homes as it did last quarter, operating expenses were more than twice as high, meaning the company has become twice as efficient even as labor costs have risen elsewhere, and it now employs more engineers than when he joined because it uses AI well.
Three Mental Models He Returns To 18:00
He lists three recurring models: friction is underestimated, the map is not the train, and truth over feelings. On friction, he explains that markets like Walmart, Amazon, and Google were once badly underestimated because people didn't see how much friction was suppressing demand; removing friction from Opendoor's home-buying process, which once took eleven people and many days, has helped the company buy six to seven times more homes this week than in the same week last year. On the map versus the train, he cites economist Friedrich Hayek's idea that people mistake models of reality, like dashboards, for reality itself, and warns that executives who run companies purely on dashboards lose touch with the actual business, the way Ray Kroc used to walk into McDonald's kitchens to check things directly. He counters this by talking to customers weekly, visiting homes constantly, and building his own dashboards from raw data.
Why In-Person Work and a Single Culture Matter 22:32
He argues remote work fails for most companies, including Opendoor, and that it mainly succeeded at Shopify because of unusual investment in remote-enabling software. Startups that stay remote tend to move slower, he says, because culture is hard to build without doing difficult things together in person, the way high school football teams build camaraderie through shared struggle. He also argues that companies cannot sustain multiple internal cultures, since that destroys shared shorthand communication and forces constant realignment.
Killing Businesses That Weren't the Mission 25:01
In his first four weeks he shut down two business lines: a general contractor service for other companies and a division called OD Select that rebuilt unlivable homes from scratch, both profitable but outside Opendoor's core purpose. Over his time there he says he has started only three products, including a mortgage product and a buyer product, while shutting down a few dozen others.
What It Means to Be a Market Maker 26:01
He traces a pattern in how internet businesses evolve: first a discovery layer like Craigslist, then a trust layer like PayPal, then a market maker like Amazon that solves both problems plus underwriting risk, which is why Amazon beat eBay despite eBay's early lead. Opendoor applies this to housing by buying a person's home directly at a fair price so they don't have to wait to find a buyer, then reselling it later. The company earns thin margins on each home transaction, but its larger profit comes from ancillary services like mortgage, insurance, title, and escrow, a model he compares to Shopify, which sells its core software cheaply but makes its real money from merchants' success, and to Google, which profits from ads rather than search itself.
The Loyalty Problem in Real Estate 29:32
Kaz Nejatian points out that most businesses depend on long-term trust, the way Shopify built its model on ongoing service relationships rather than one-time payments. He calls people who take your money upfront and vanish carnies, and argues no one trusts a stockbroker who profits from a single trade. Yet he finds it strange that the largest asset class people own, their homes, is transacted this way, with no ongoing relationship between buyer and seller.
Get Things Done Before Asking Permission 32:02
Nejatian explains why Shopify's Tobi Lutke posts the phrase "do things, tell people" on office walls, reversing the usual corporate order of getting approval before acting. Most companies build a chain of command meant to say no and reduce risk, and Amazon countered this by declaring itself a default yes company. Nejatian prefers going further, minimizing the number of people who can say no at all. He gives a concrete example from Opendoor, where getting the power turned on in a purchased home once required a Salesforce form, several managers, and days of back and forth. He fixed it by giving employees corporate cards and a Slack tool that instantly returns the utility phone number and account details needed to turn power on directly.
Drifting Off Course Like a Pilot 35:32
He compares corporate decline to new pilots flying into clouds who turn slightly off course again and again until they end up fully inverted without realizing it. He treats heavy use of acronyms inside a company as a warning sign, since acronyms create in-groups and out-groups that make it hard to notice when something has gone wrong. He mentions personally turning on utilities and fixing work orders himself, comparing it to Elon Musk sleeping at the factory, which he sees as a genuine method rather than a publicity stunt.
Clear Language and Honest Decisions 37:15
Nejatian used to ask his team to read Orwell's essay on politics and the English language, which argues that words like fascism lose meaning through overuse. He applies this to corporate speech, challenging anyone to find a meaningful sentence using the word leverage, and criticizes passive constructions like "legal decided," insisting that a specific person, not an abstract team, should be named as the decision-maker.
Product Over Strategy 40:01
Asked how he evaluates a company, he says products matter more than strategy, since most companies share similar strategies and the best executor usually wins, as with Google among dozens of search engines. He judges a product by whether a thousand people genuinely love it, arguing that without that, no amount of capital can save a business.
Judgment Over Endless Testing 41:30
Nejatian argues that delaying decisions rarely reduces risk, describing a meeting where he asked colleagues to predict an A/B test result in advance since gathering more data often just avoids commitment. He cites Google testing 400 shades of blue as an example of testing replacing judgment rather than informing it, insisting that human judgment must guide decisions that tests cannot settle.
Turning Around Like an Intervention 43:30
He compares fixing a failing company to helping someone with alcohol addiction, where recovery begins with an uncomfortable moment of ownership and daily accountability, much like Alcoholics Anonymous tracking progress day by day. He credits his wife, kids, and regular prayer for helping him admit mistakes without becoming paralyzed by guilt.
Founders Facing Financial Fear 46:01
Nejatian describes founders regularly checking bank statements to count how many payrolls remain, a moment he says reshapes behavior the way Paul Graham's idea of "default dead" forces action. He uses the image of a chemical spill to explain fixing problems, saying the first step is easy, like grabbing a shovel, even though full cleanup takes much more effort.
Steve Jobs and Real Pain Tolerance 47:30
He references the moment Steve Jobs nearly bankrupted Next's hardware division, arguing that near-death experiences for a company strip away concern for appearances and reveal a higher pain tolerance than people expect. He connects this to the book The Courage to Be Disliked, saying most limitations are self-imposed, and shares that fleeing Iran as a teenager and being excluded from clubs left him unbothered by later criticism, including from the press.
Rugby and the Joy of Pain 50:02
Nejatian describes playing rugby in high school, a sport with almost no audience in North America, where players endure pain for their teammates rather than for recognition. He connects this to Isadore Sharp's idea that excellence is the capacity to take pain, agreeing that the person who can endure difficulty longest tends to win, and worries that modern comfort has left people unprepared for necessary pain.
Defining Success Through His Children 51:31
Asked what success means to him, Nejatian says he and his wife agreed early in their marriage to optimize their lives for putting a dent in the world. With four young children, ages seven, five, three, and two, he wants them to one day be proud of how their father spent his career, since his time away from them is a real sacrifice that he feels must be justified by meaningful work.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
