Watch the full episode

WeWork was once valued at 47 billion dollars. It filed for Chapter 11 bankruptcy in late 2023, with reported liabilities of more than 18 billion dollars. The company that emerged needed an operator, not another visionary. It got one in John Santora, a 47 year veteran of real estate firm Cushman and Wakefield, who became chief executive in June 2024 as WeWork exited bankruptcy. He describes himself, as reported from the event, as a suit and tie, buttoned up real estate person. That description is the strategy.

The WeWork CEO John Santora turnaround in numbers

Speaking at Business Insider’s The Long Play event on 29 October, Santora laid out a recovery that would have sounded fanciful three years ago. WeWork has now reported eight quarters of EBITDA profitability. It is projecting a fourth consecutive quarter of cash flow success. Reported revenue for 2025 is roughly 2.3 billion dollars. Treat every one of those figures as reported, not audited. They come from event coverage, not from a WeWork financial statement. However, the direction is consistent across the coverage, and it points one way: a real estate business finally being run like a real estate business.

That last sentence is the twist. WeWork did not die because coworking was a bad idea. It died because a founder sold a real estate business as a consciousness raising movement, and nobody in charge knew how to run the underlying business. Offices still need leases, landlords, occupancy rates and margins. Vision does not pay rent. Discipline does.

Give and take: the whole playbook

The playbook has two halves, and both matter. The first half is the standard. Professional dress in shared spaces. A clear signal to landlords and members that adults are in charge again. The second half is the concession. One work from home day a week. No clock watching on arrival or departure times, as long as the work gets done and customers are happy. Santora reportedly even wears sneakers himself on Fridays. The dress code is not about suits. It is about signalling seriousness to the people who pay the bills.

This is why the give and take framing is load bearing. Standards without give feel like control, and staff will route around them. Flexibility without standards feels like drift, and customers will notice before managers do. Santora paired one visible standard with one genuine concession, and that pairing is what makes the standard stick. Compare it with the rigid five day mandates failing elsewhere, which buy attendance and lose goodwill. Compare it also with fully remote cultures drifting apart, which buy freedom and lose cohesion. WeWork’s version buys both, one day at a time.

In addition, the model answers the return to office debate with something rarer than a policy. It answers it with a trade. Employees accept a visible standard because they receive visible trust in return. Managers accept flexibility because they receive visible professionalism in return. That trade is easy to describe and hard to fake, which is exactly why it works as a signal.

Why discipline is the culture

Founders overrate vision and underrate operational discipline, and WeWork is the most expensive proof of that ever produced. Neumann sold a story. Santora runs a business. The difference shows up in the numbers above, but it also shows up in the building. A member walking into a WeWork today meets a company that looks like it takes its own business seriously. Three years ago, that was not a safe assumption.

Discipline is not the enemy of culture. It is what makes culture affordable. A company that cannot pay its landlords cannot protect any culture at all. Santora understood that sequence. First, make the business viable. Then, and only then, argue about what kind of workplace it should be. Most turnaround stories reverse that order, and most of them fail for it.

What founders should steal from this

Steal the pairing, not the dress code. One visible standard that says “we are serious”, paired with one genuine flexibility concession. The standard must be visible, because invisible standards change nothing. The concession must be genuine, because a fake concession breeds more cynicism than no concession at all. Together, they tell staff and customers the same story: this company keeps its promises in both directions.

WeWork’s next test is whether discipline survives success. The moment the company starts believing its own comeback story, it is back where its founder left it. Santora’s boring playbook works precisely because it refuses to be a story. It is a routine. Routines do not make headlines, but they do make payroll. For a company that once believed its own headlines all the way into bankruptcy, that is the most radical change of all.

The post WeWork’s Boring CEO Is Winning Where Its Visionary Founder Failed appeared first on CEO Medium.

For example, the market reaction tells you how seriously investors took the story. Shares fell as much as 20% to 52.11 rupees, the day’s lower circuit. However, they recovered to close at 57.17 rupees, down 12.2%. The fall came on a day the bank reported its highest ever quarterly profit. That profit was 218 crore rupees (2.18 billion rupees, about $22.9 million). In addition, total business crossed 1 lakh crore rupees (1 trillion rupees) in the September quarter. A record quarter did not save the CEO. Instead, a story did what numbers could not.

The market reaction and the perceived conflict

Vasudevan led Equitas since it began as a microfinance institution in 2007. In addition, he oversaw its conversion to a small finance bank in 2016. In other words, a small finance bank is a lender that serves people and small businesses underserved by mainstream banks. He spent two decades at the lender, after two decades at Cholamandalam, and built it from microloans into a listed bank.

The perceived conflict, stated fairly, is this. His daughter, Varshini Vasudevan Pathangi, co-founded Unico Housing Finance, an unlisted non-bank lender, in 2023. In fact, a non-bank lender, or NBFC, lends money but does not hold a banking licence. Vasudevan says the arrangement was disclosed to the bank’s board and to the Reserve Bank of India, India’s central bank. Furthermore, he says the bank had no dealings with Unico beyond one short-term deposit. That deposit was made about two and a half years ago and repaid at maturity. Moreover, he maintains there was no real conflict, and that the arrangement sits within the Banking Regulation Act.

In investor meetings over the past three years, he was repeatedly asked about the perceived conflict. He explained the disclosure, the single deposit, the regulatory position. However, the perception persisted. In his resignation letter he wrote: “Sometimes perceptions are stronger than reality.” In addition, investors had told him the “overhang” was causing the shares to underperform. Per media reports, the Reserve Bank had earlier this year asked the bank’s board about the relationship. This happened when his tenure was being extended. However, Vasudevan himself told BusinessLine that “the regulator is not in the picture at all.”

What the Equitas CEO resignation teaches founders

Founders treat disclosure as the fix for a conflict. It is not. In fact, disclosure tells people the facts. However, it does nothing about what they believe. Once a narrative attaches to a leader, every explanation sounds like a defence. As a result, the cost of staying eventually exceeds the cost of leaving. For example, Vasudevan chose exit, and that choice is the warning. Therefore, if you wait until investors tell you the overhang is hurting the stock, you have already lost the argument.

The board has asked Vasudevan to stay until a successor is found. In addition, it has asked its nomination and remuneration committee to identify internal and external candidates. Furthermore, any appointment will require Reserve Bank approval. This is not his first attempted exit. In 2022 he planned to step down to pursue philanthropy. However, the board talked him out of it. The difference between the two exits is the lesson. The first time he wanted to go. The second time the market wanted him gone.

What founders should do

Map every family, friendship and side venture overlap before investors do. Build real separation instead of relying on transparency alone: recusal from related decisions, independent oversight, or a clean break. If an overlap survives three investor meetings, treat it as a governance problem, not a communications problem. Transparency tells the truth. However, structure changes what people can believe.

Vasudevan built a bank over nearly two decades and left it at a record profit. He left not because he did something wrong, but because he could not kill a story. That is the part founders should remember. You do not lose your company the day you do something wrong. Instead, you lose it the day people stop believing the explanation.

Sources and further reading: Reuters: India’s Equitas Bank tanks after CEO steps down over conflict of interest. Hindu BusinessLine: Equitas SFB MD & CEO P N Vasudevan resigns citing perception of conflict of interest. Market data: EQUITASBNK.NS. More from CEO Medium: CEOs, Leadership & Management. Bausch + Lomb’s CEO Is Out, Nobody Will Say Why, and Forecasts Just Went Up. Lindt Fired Its CEO and Called Back the Old One. The Board Should Be Next.

Watch the full episode

The post He Did Nothing Wrong. He Resigned Anyway. That Is the Lesson appeared first on CEO Medium.

Watch the full episode

The DoorDash Instacart deal is about Walmart, not groceries

Start with the numbers that explain why two rivals would combine at all. Instacart’s market value was about $13 billion at Friday’s close, so a $17 billion valuation is a premium of roughly 30 percent. On Monday, Instacart shares finished 7.9 percent higher, while DoorDash shares fell 6.1 percent. Investors, in other words, think Instacart is getting a good price and DoorDash may be paying it.

The strategic logic is not groceries. It is Walmart and Amazon. Wedbush analysts wrote on Monday that a combination of strong restaurant delivery plus a marketplace of grocers “can be a critical combination against Walmart and Amazon in particular, who continue to push in this market”. EMarketer forecasts United States online grocery sales growth of 12.9 percent this year, with online sales accounting for 15.5 percent of overall grocery sales. Walmart leads those sales with 30.9 percent market share, ahead of Amazon at 23.6 percent, with Instacart and Kroger trailing both. Neither DoorDash nor Instacart can outmuscle those two alone. Together, they might.

Why now: Instacart is being squeezed

Instacart did not come to the table from strength. The Information reported last week that Instacart was exploring a sale, generating interest from DoorDash and Uber, and the Journal reported that discussions with Uber are no longer active. Bernstein analysts noted they were surprised at how quickly a possible deal was moving, and that their initial reading was that a deal was unlikely “considering it was [Instacart] approaching DASH/UBER and not vice versa”. Instacart went looking for a buyer. That matters.

The squeeze is visible in the trading. Instacart’s sales have grown more slowly than DoorDash over the past year, according to eMarketer analyst Zak Stambor, who warned that Instacart “risks getting squeezed between retailers building their own digital channels and larger platforms that can bundle grocery with other services, giving consumers more reasons to keep coming back”. That is the whole story in one sentence. A standalone grocery marketplace is a feature. A bundle of restaurants, groceries and retail, delivered by one network, is a habit. Instacart has the grocers. DoorDash has the network and the habit. Each has what the other lacks.

What the combination would actually do

Put the two businesses together and the shape is obvious. DoorDash, led by co-founder and chief executive Tony Xu, brings restaurant delivery, a vast courier network and a push into delivery of other goods from other retailers, plus autonomous delivery vehicles and purchases through ChatGPT. Instacart, run by chief executive Chris Rogers, brings an established grocery marketplace and deep retailer ties. Bernstein analysts called the joined companies an “indispensable solution”, while noting any deal would require some combination of debt, cash on hand or equity.

There are real questions underneath the enthusiasm. Wedbush flagged product and merchant overlap, and how the Instacart brand integrates into the broader acquirer’s brand. Those are the questions that kill mergers after the announcement, when two product teams discover they built the same thing twice. On regulation, Wedbush said it does not expect material pushback, given the strength of Amazon and Walmart in the market. That is a reasonable reading: a combined DoorDash and Instacart would still be chasing the two giants, not replacing them.

The founder lesson: bundle or be squeezed

Founders should read this deal as a warning about the endgame in platform markets. When the giants enter your category, a good single product stops being enough. Walmart built its own grocery delivery. Amazon built its own. The platforms that survive are the ones that bundle enough services that leaving costs the customer something. Instacart learned that a marketplace alone, however good, can be squeezed from both sides at once.

Tony Xu made the same point from the other side this week, talking about AI shopping agents. Agents can search, order and pay, he argued, but they cannot carry anything to a door. DoorDash owns the end to end job: getting an item to a customer in the condition they expect, on time, every time. That physical network is the moat, and it is the thing Instacart does not have. If this deal closes, the delivery industry will have consolidated from a handful of rivals into a genuine duopoly of scale against Walmart and Amazon. If it falls apart, Instacart will have advertised to the whole market that it is for sale, and the squeeze will only tighten. Either way, the standalone era of grocery delivery is over.

Related: Entrepreneurship and startups and Gently does it: how an AI-powered delivery firm is decentralizing logistics.

Photo by MART PRODUCTION on Pexels.

The post DoorDash Wants to Buy Instacart for $17 Billion. The Real Target Is Walmart appeared first on CEO Medium.

Watch the full episode

Why OpenAI Google chips matter more than another model launch

Forget model benchmarks for a moment. Who controls the hardware underneath the models increasingly decides the AI race. OpenAI now acts like a company that knows it cannot afford a single supplier. OpenAI is one of the largest buyers of Nvidia graphics processing units. The chips help it train models and run inference. Buying from Google does not replace that relationship. It gives OpenAI leverage inside it and a second source to point to in every negotiation. That also offers a hedge if supply tightens.

This is also not a cold start. Reuters reported in June 2025 that OpenAI had begun renting Google chips. In October 2026, OpenAI announced its top models would be sold through Google Cloud. Its exclusive distribution arrangement with Microsoft expired in April. The relationship has been deepening in plain sight for months. A chip purchase would be the next logical step, not a sudden reversal.

Rivals in search, partners in silicon

Google and OpenAI compete head on in chatbots and search. That is what makes this story strange and interesting at the same time. Google’s Tensor Processing Units, or TPUs, have gained traction as an alternative to Nvidia’s market leading chips. Companies look for cheaper options and chips better suited to running models. Google began selling TPUs directly to some customers this year. Previously, it rented them only through its cloud.

OpenAI is not the only buyer testing that door. Oracle said on October 27 that it will begin offering Google’s TPUs for the first time. That came alongside an expanded Nvidia deal at multi gigawatt scale. It also came with new inference offerings from Cerebras and SambaNova. Oracle Cloud president Mahesh Thiagarajan put it plainly: “Different AI workloads have different compute needs.” That sentence is the whole market in miniature. Training still belongs mostly to Nvidia. Inference is the business of answering user prompts at scale. It is up for grabs, and Google wants a share of it.

Nvidia is not being dumped. It is being negotiated with

It would be wrong to read this as OpenAI breaking with Nvidia. The evidence supports diversification and leverage, not a divorce. Nvidia still dominates training. OpenAI still buys its GPUs at enormous scale. No source has suggested otherwise. What has changed is the assumption that one supplier is enough. When the biggest buyer in the market starts shopping elsewhere, every other buyer gets permission to do the same. Every supplier gets a warning.

Investors will get a fresh data point within hours. Alphabet reports quarterly results after the market closes on Wednesday, October 28. Attention will focus on whether heavy AI investment keeps driving cloud growth. Another very strong quarter looks likely for Google Cloud. Any comment on TPU sales or OpenAI will draw close attention. Microsoft reports the same evening. By Thursday morning, the market will have a much clearer picture. That picture will show whether the AI infrastructure boom is paying for itself.

The lesson for founders: audit your single point of failure

Here is the part that matters to readers who will never buy a chip. OpenAI, the most powerful AI company in the world, is diversifying its most critical input. So a startup with one vendor for its most critical input is far more exposed than it thinks. Models get copied in months. Chip supply, power contracts and compute access decide who ships and who waits.

Founders should do this audit now, before they need to. List the inputs your business cannot run without. For each one, ask what happens if the price doubles, the lead time triples, or the supplier changes its terms. Then find a second source while you still have negotiating power. Negotiating from weakness costs far more than a backup supplier ever will. The next great AI rivalry will not play out in model benchmarks. It will play out in foundries, power grids and procurement meetings. OpenAI just showed everyone which game it is playing.

Related: Leadership and management, Anthropic’s IPO filing, and Texas Sues OpenAI and Sam Altman Over ChatGPT Safety.

Photo by Jeremy Waterhouse on Pexels.

The post OpenAI Is Buying Chips From Google. The AI War Just Became a Supply Chain War appeared first on CEO Medium.

That is a 44 point gap. It comes from Boston Consulting Group, not from one company’s HR blog. BCG surveyed 645 CEOs at companies with at least $100 million in revenue, plus 500 nonexecutive employees, for the second edition of its CEO Insomnia Index, released on October 14, 2026. The report is back in leadership feeds this weekend, and the reason is simple. Every founder is now running an AI transition, and this survey measures what that transition feels like from below.

Watch the full episode

The CEO employee trust gap, in plain numbers

First, look at the AI numbers, because they are the sharpest part of the story. Four in ten employees say their CEO has not been transparent enough about how AI will affect their roles and careers. Yet only 15% of CEOs worry that their approach to AI could erode trust. One side is anxious. The other side does not know it should be.

Second, the gap is wider than AI alone. Nearly 70% of CEOs are confident employees trust them to prioritize employee well-being when they fund critical business priorities. Only 36% of employees agree. That is a gap of more than 30 points on a question that should be easy to get right: do you put people first when money gets tight?

Third, most CEOs do not seem worried at all. In fact, 58% say they are not concerned about rising employee disgruntlement, and many point to efforts they already have in place. Meanwhile, 60% say near-term pressures consume a disproportionate share of their time, up 3 points from the previous survey. In other words, leaders have less time to build trust at the exact moment trust matters most.

Trust is not soft. It is effort.

Here is the number that should change how founders talk about trust. Among employees whose trust in their CEO’s handling of AI grew, 50% are more willing to go above and beyond at work than they were three months ago. Among employees whose trust eroded, fewer than 10% say the same. That is a difference of more than 40 points.

So trust is not a vibe. It is discretionary effort, and discretionary effort is output. A team that believes its leader will tell the truth works harder on a hard change. A team that does not believe it conserves energy, waits, and watches.

Be careful with what this survey does and does not prove. BCG reports what CEOs and employees said about themselves. It does not prove that AI causes distrust, or that trust alone causes extra effort. However, the association is stark, and it points in one direction: people give more when they feel honestly led through change, and they give less when they feel managed around it.

This is not one survey

The BCG numbers do not stand alone. Glassdoor’s Worklife Trends 2026 finds senior leadership ratings down sharply since their pandemic peak, with only a slight recovery in 2025. In employee reviews, mentions of misalignment rose 149% and mentions of distrust rose 26%. That language matters. Misalignment means people no longer believe the plan. Distrust means they no longer believe the person explaining it.

Businessolver tells a similar story from the other side of the desk. In its 2026 State of Workplace Empathy research, 33% of CEOs now describe their own workplace culture as toxic, up 25 points in a year. Among employees, 40% say the same. Even more telling, workers in cultures they call toxic report far less AI training than workers in cultures they call empathetic: 33% versus 87%. So the trust problem and the AI readiness problem may be the same problem, wearing two labels.

Add one more BCG finding, because it explains why the gap persists. About a third of CEOs, 34%, say they have considered stepping down in the past six months. At the largest companies in the study, that figure rises to more than 40%. Leaders are tired, pressed for time, and still expected to narrate a transformation they have not finished deciding. Silence then feels easier than honesty. Unfortunately, silence reads as a plan the team is not supposed to see.

What teams actually need to hear

Most AI announcements fail in the same way. A CEO announces a transformation as if the announcement were the strategy. The team hears a transformation with no role map, no timeline, and no honest answer about headcount. Then everyone does the math in private, and the private math is always worse than the real plan.

What teams actually need is three sentences. First, what changes. Second, who it affects. Third, when they will hear more. For example: “We are testing AI tools in customer support this quarter. No support roles change before March, and we will share the results with the team before we decide anything.” That is not a promise you cannot keep. It is a timetable people can hold you to.

And when you genuinely have not decided, say so, with a date attached. BCG’s own advice makes the same point: when answers are uncertain, communicate more, not less, rather than leaving employees to fill the silence with their own assumptions. “We have not decided yet, and here is when you will know” builds more trust than silence, because it treats the team like adults.

What founders should do this week

Do not launch another engagement survey and call it a trust check. Instead, try a smaller, sharper exercise.

First, ask the team directly what would make them trust the direction more. Ask in writing, anonymously if needed, and ask about AI specifically: which roles, which timelines, which decisions feel unclear?

Second, publish what you heard. Not a summary that smooths the edges, but the actual themes, including the uncomfortable ones. People trust a leader who repeats back hard feedback far more than one who thanks them and moves on.

Third, name one AI decision you will explain before you make it. Pick something real, like how AI will change one workflow, or how you will judge whether a pilot worked. Then explain it before the decision lands, not after.

Finally, measure again in 30 days. Ask the same question, in the same words, and compare the answers. Trust grows when people see their feedback change something. It dies when feedback disappears into a deck.

For more on building a culture that can survive hard change, read Creating Brave Cultures Through Transformational Leadership (link above). For the brain science behind why uncertainty drains performance, see How to create a brain-friendly workplace culture (link above). Both point to the same lesson: people do their best work when they understand the change and believe the person leading it.

The bottom line for founders

The CEOs who close this gap will not be the ones with the best AI tools. Tools are now easy to buy. Instead, the winners will be the leaders whose teams believe them when they say what comes next.

That belief is earned in small, repeatable acts: a clear role map, a real timeline, an honest “we have not decided yet” with a date for the answer. Do those things, and the 44 point gap starts to close. Skip them, and no AI rollout will save you, because the people who must make it work have already stopped believing the plan.

Photo by Tran Nhu Tuan on Pexels

The post 9 in 10 CEOs Think Their Team Trusts Them. Fewer Than Half Do. appeared first on CEO Medium.

That contradiction is the story. Amodei is not a founder who stumbled into a valuation. He is a founder who warned, in plain language, that the technology his company builds could endanger humanity. However, he then filed to sell it to the public. The price would make Anthropic one of the most valuable companies on earth.

The Anthropic IPO filing in numbers

Anthropic submitted its S-1 to the Securities and Exchange Commission on Monday, October 26, after filing confidentially in June. In addition, the filing puts hard figures on a company that private investors valued at $965 billion earlier this year.

Revenue in the first half of 2026 was $16.3 billion, up from $1.16 billion in the first half of 2025. The net loss over the same period was $80.7 billion, compared with a $13 billion loss a year earlier. Reports put the IPO raise at around $100 billion. That would eclipse the record set by SpaceX’s June offering. Meanwhile, the Wall Street Journal has reportedly pointed to the week of November 16 as the target.

Anthropic is five years old. Amodei founded it with his sister Daniela after both left OpenAI. In addition, the company says its top customers are diversified. According to filing summaries, no single customer topped 10 percent of revenue in the first nine months of 2026.

Why Amodei’s warning matters more now

Weeks before the filing, Amodei argued for an industrywide slowdown on the most advanced models. He also called for independent safety evaluators embedded at the top labs. However, critics said the plan had conflicts of interest. The labs themselves would pay for their own scrutiny.

Founders should read the filing and the warning together, because that is how the market will read them. The warning gives Anthropic a moral frame no rival can copy cheaply. It says: we are the lab that tells you the danger. In turn, the IPO converts that frame into capital on a scale no private round could match. Whether that is principled or convenient is the debate the next month will settle. The debate itself is free marketing for the offering.

What founders should take from this

First, losses no longer disqualify a company from the biggest stage in capitalism. An $80.7 billion half-year loss is not a footnote in this filing. Meanwhile, it sits next to a growth curve so steep that investors must price 2027 and 2028, not 2026.

Second, a founder’s public position is now part of the prospectus. Therefore, investors, regulators, and rivals will quote Amodei’s safety stance back at him while Anthropic is public. Founders who take a public stand should assume it becomes a permanent part of their company’s story.

Third, timing is a weapon. Anthropic filed first among the frontier labs, ahead of OpenAI. Whoever lists first sets the terms of the conversation. Therefore, every lab that follows must match Anthropic’s numbers.

What to watch

No one has set the final valuation or offering price yet. However, the reported November 16 target week could move. OpenAI’s own expected debut means this is not a single IPO. In other words, it is the opening of a new public market for AI labs. In fact, Anthropic is writing the first prospectus.

Amodei built Anthropic on a promise that safety and capability could grow together. The Anthropic IPO is about to test whether investors believe that promise at $2 trillion. Founders watching should pay less attention to the valuation and more attention to the test. Ultimately, the result will set the price of ambition for every AI company that follows.

Featured image: Photo by Alex Knight on Pexels.

Related reading:

Entrepreneurship & Startups

Anthropic IPO Delayed to November: What the $2 Trillion Listing Means for Founders

Texas Sues OpenAI and Sam Altman Over ChatGPT Safety

Sources:

New York Post: Anthropic files for blockbuster IPO that could value AI giant at $2 trillion

KTVU / Associated Press: AI giant Anthropic files papers to sell shares to the public for 1st time

The post Anthropic Just Filed for a $2 Trillion IPO. Its CEO Spent Weeks Warning About AI Risk. appeared first on CEO Medium.

Polymarket confirmed nothing publicly. A spokeswoman declined to comment to the New York Post, which broke the story, and sources say a separation agreement is being finalised. But the outline is clear enough: an internal probe, worried investors, a regulator circling, and a board that decided the company’s chief marketer had become its biggest liability.

Watch the story on CEO Medium News:

Watch the full episode of CEO Medium News

What the probe found

The trouble started with growth tactics. Reporting on the internal probe describes a campaign that ran from December 2025 to May 2026. Paid creators published 1,105 promotional videos as part of it. Creators filmed hundreds of the videos on replica websites built to look like Polymarket’s real platform. The videos showed about $1.9 million in simulated wagers presented as genuine action. One batch of 118 clips displayed nearly $900,000 in fabricated winnings. Placed for real, those bets would have lost more than $166,000. The campaign pulled in over 140 million views, and some creators never disclosed they were paid.

The Wall Street Journal exposed the staged videos in June. The Commodity Futures Trading Commission opened an investigation, which is ongoing, and no wrongdoing has been established. According to the Post’s sources, the company’s own lawyers had already told Modabber to stop the campaign. That was before any of it became public. The people paid to see the risk warned him internally. The campaign ran anyway.

The probe reportedly looked wider. More than $2.5 million in payments to creators ran through a personal PayPal account over thirteen months. At least $350,000 of it went to identified influencers, some of whose posts carried no sponsorship disclosure. Whatever the intention, money moving through a personal account is what governance failure looks like on paper.

Why the Polymarket CMO fired story is a governance story

Firing a CMO over a marketing scandal is routine. Firing your oldest friend is not, and by the Post’s account, Coplan nearly did not do it. Insiders say disciplining Modabber was always tricky because of the bond. They say an irate Coplan threatened to fire members of the legal team pushing for the termination. And they say he later accepted the split was best for the company.

The board’s patience ran out first. Polymarket is weighing a public listing, and investors have pushed for stronger financial controls. Directors concluded in recent weeks that Modabber was blocking the company’s attempt to shed its scrappy startup image. The August restructure of the marketing team, which brought in new leadership over him, now reads like the warning shot.

None of this required hindsight to see. A loyalist nobody can discipline is not a culture asset. He is an unpriced risk, and the bill arrives at the worst possible moment. In this case, a federal probe is open and IPO bankers are reading the coverage.

Trust is the product

There is a reason this scandal lands harder at Polymarket than it would at a normal consumer app. A prediction market sells one thing: prices people can believe. The odds are the product. Marketing that fabricates wins is not an aggressive growth tactic that went too far. It is counterfeiting inside a business whose entire value is that the numbers are real.

That is also why the comparison with ordinary startup hype does not hold. A meal kit company exaggerating its subscriber growth is cheating its investors. A market exaggerating its winners is cheating the market itself. And that market is seeking regulators’ blessing for a public listing. The CFTC probe will decide the legal questions, and it has established nothing so far. The board has already answered the governance question.

What founders should learn

Loyalty is not a management structure. The people who were there on day one need performance reviews, spending controls and someone who can tell them no, exactly like everyone hired later. If a co-founder or an old friend sits beyond discipline, you have not built a loyal team. You have built a blind spot with a title.

Second: marketing that cannot survive daylight is not marketing, it is liability with a media budget. If the tactic needs a replica website and undisclosed payments, the growth it produces borrows against the company’s credibility, at an interest rate regulators set.

And the hardest one: if your board has to force you to fire your friend, you have already failed at the part of the CEO job that a CEO cannot delegate. Coplan, at 28 one of the youngest billionaire founders in the country, built something genuinely new. The open question after this week is whether he learned the lesson, or simply lost a friend.

Photo: Superbass, CC BY-SA 4.0, via Wikimedia Commons

The post Polymarket Fired Its CMO Over Staged Betting Wins. His Boss Is His High School Friend. appeared first on CEO Medium.

Watch the story on CEO Medium News:

Watch the full episode of CEO Medium News

On Wednesday, a New Mexico judge ordered Meta to pay $25 billion to the state, the largest penalty the company has ever faced over the Cambridge Analytica scandal. The number is enormous. The reasoning is worse. Judge Francis Mathew did not describe a rogue team or a policy slip. He described a plan.

What the Meta $25 billion fine actually says

A jury found in September that Meta violated New Mexico’s Unfair Practices Act more than 43 million times. The violations misled users about how Meta handled their data. Each violation can carry a penalty of up to $5,000, which put the theoretical ceiling near $219 billion. Mathew set the figure at $25 billion.

The case reaches back to 2018. That year, Facebook disclosed that a personality quiz app had harvested data from as many as 87 million users. It fed the data to Cambridge Analytica, the political consultancy that built voter profiles for the 2016 Trump campaign. New Mexico sued in 2021. While 48 other states chose to settle their claims, New Mexico went to trial alone. It sat through two weeks of evidence, and won.

The judge also found that the misrepresentations went far beyond a marketing page or a help centre. They reached the most senior levels of the company. That includes chief executive Mark Zuckerberg and former chief operating officer Sheryl Sandberg.

Deception at the core of the business model

One sentence in the ruling is why this story belongs on a leadership page and not just a legal wire. Mathew wrote that Facebook’s misrepresentations about users’ control of their data “were not isolated incidents, but rather a systematic pattern of deception at the core of Facebook’s business model,” part of “a coherent strategy designed to mislead the public.”

Read that as a founder. Courts fine companies for broken rules all the time, and companies pay and move on. This ruling says something different: that the gap between the promise and the practice was the product strategy. The promise was control. The practice was collection. The company told users they held the keys while it monetised what walked through the door.

That is the version of the story that should worry every operator. After all, most companies do a milder version of it. The privacy page says one thing. The growth dashboard rewards another. Nobody writes the deception down as strategy, which is exactly how it becomes one.

The trust gap has a market price

Every chief executive claims trust is the company’s most valuable asset. The claim is cheap. Trust is not a value on a wall. It is the distance between what you say and what you do. And the people who measure it cannot see inside your building.

Meta’s fine is simply the market price of a wide gap, calculated by a jury at $5,000 a violation. Attorney General Raúl Torrez, who brought the case, joked that the state may have “paid for the construction of schools for the next 100 years,” and state officials say the money is dedicated to an education fund. Founders should hear the other side of the joke. The tuition is real. Whoever holds the company when the gap is finally measured pays it.

There is a practical discipline buried in this. Write the policies you could defend in a courtroom, not the ones that read well in a press release. Audit the distance between your public story and your data practices before a regulator does it for you. New Mexico has now published the playbook, and attorneys general in the other 49 states know how to read.

The appeal does not end the lesson

Meta says it will appeal, and the number may well shrink. Chief legal officer C.J. Mahoney called the ruling one that “makes it all but impossible to do business in the state of New Mexico.” He argued that the lawsuit “seeks to impose a tax on speech.” The company may be right about the law. It is probably wrong about the lesson.

A penalty that a company can price in is not a deterrent; it is a licence fee. Meta has paid licence fees before, including a $5 billion federal settlement in 2019 over the same scandal. The uncomfortable question this verdict asks is whether $25 billion is finally large enough to change the internal maths. Or Meta absorbs it too, as the cost of doing business the way the business has always been done.

That question is not really about Meta. Every founder reading this runs a smaller version of the same calculation. How much daylight can exist between the story and the operation before someone with subpoena power measures it? New Mexico just showed the measuring instrument. The companies that survive it will be the ones that closed the gap while it was still cheap.

Photo: Anthony Quintano, CC BY 2.0, via Wikimedia Commons

The post Meta’s $25 Billion Lesson: Deception Was the Business Model appeared first on CEO Medium.

The interesting part is the growth underneath it. Blacksmith says its annual recurring revenue has grown 17 times since January. More than 10,000 startups and enterprise companies now run their builds and tests on its platform, and the company has processed over 90 million CI jobs. Customers include Mercury and Granola, two startups that know infrastructure when they see it.

What the Blacksmith Series C is really buying

To see why investors moved so fast, start with the job Blacksmith does. Every piece of software has to be built and tested before it ships. That loop, known as continuous integration or CI, runs every time an engineer proposes a change. For years it was a background chore. Then AI coding agents arrived, and the volume of code needing checks exploded.

Anthropic reported last month that its own CI volume grew 25 times in six months as Claude took over most of its code output. Blacksmith’s co-founder and CEO Aditya Jayaprakash hears the same story from customers. Developers who once pushed three or four pull requests a day now push orders of magnitude more, he says, and their CI infrastructure cannot keep up. Most teams still run those checks on general-purpose cloud machines that were never designed for the job, so they wait in queues while expensive engineers sit idle.

The pain is not abstract. Builds queue behind other builds, so a ten-minute check becomes an hour of waiting. Teams over-provision cloud machines to shorten the queue, then pay for capacity that sits idle overnight. Or they under-provision, and watch engineers context-switch while tests crawl on older hardware. When the same conversation opens the same way at almost every large engineering team, as Jayaprakash describes, the market has effectively written the pitch for you.

The bottleneck moved from writing code to trusting it

Jayaprakash founded Blacksmith to make CI, in his words, “less painful”. The sharper version of the pitch is about trust. “As agents write more of the world’s software,” he said this week, “the question that matters is whether it can be trusted.” Every untested change is a small gamble, and companies now place thousands of those gambles a day. Whoever checks the work reliably, at speed, owns a toll booth on the road AI code has to travel.

There is a budget logic working in Blacksmith’s favour too. Engineering salaries are the largest cost in most software companies, and CI delays tax exactly that line. A tool that hands senior engineers their waiting time back does not need a speculative return on investment story. It pays for itself in hours, which is why infrastructure like this keeps selling through every change of fashion at the model layer.

That position explains the round’s shape. Premji Invest joined as a new investor, while Peak XV Partners, GV and Y Combinator followed on. The company has now raised $125 million in total, and it plans to spend the new money on hiring and on expanding its compute footprint. In other words, the capital goes straight into the bottleneck it sells relief from.

Raise when refusal becomes embarrassing

Notice the tempo. Series A, Series B and Series C landed inside roughly a year, and each round followed a visible inflection rather than a calendar. The Series C arrived two months after the Series B because demand for the story outran the last cheque. Founders often treat fundraising as a season to endure. Blacksmith treated it as a market to read: when your growth rate becomes the headline, the round prices itself.

The investor list tells its own story about where the company started and where it is heading. Y Combinator backed it early, and Peak XV Partners, the firm that split from Sequoia Capital in 2023, stayed in as Sequoia itself led the new round. Premji Invest joined for the scale phase. Syndicates like that do not assemble around a feature. They assemble around a company they believe can own a layer of the stack.

There is a contrast worth holding onto. Elsewhere this week, investors handed Flapping Airplanes, a research lab with no product on sale, a $1 billion round at a $6 billion valuation, a bet on pure potential. Blacksmith raised the opposite way, on usage that anyone can count. We have written before about AI companies raising fortunes on agent promises, and the pattern is consistent: the market funds both stories, but the revenue story gets to choose its terms.

What founders should take from this

First, sell into the pain a bigger revolution creates. Blacksmith did not need to build a coding agent. It needed the agents to succeed, because every line they write needs testing. Ask where the flood created by the current wave pools, and build there. Second, boring is a moat when it is measured. Reliability rarely trends, but 17x revenue growth does, and enterprise buyers renew what does not break. It also compounds quietly: every job the platform runs teaches it more about how code fails, which makes the next check faster and the product harder to displace.

Third, keep your proof countable. Blacksmith sells speed in a step everyone else neglected, and it can prove the value in job counts and customer names. Whatever your market, find the number that proves your value without adjectives, then put it at the top of the deck. If that number is growing 17 times a year, you will not need to chase the meeting.

The broader lesson is that the AI boom has entered its plumbing phase, and plumbing pays. Founders who validate, secure, monitor or bill for AI output are standing where the money now flows. More on the companies building that layer sits in our Entrepreneurship & Startups hub.

Photo: Sarah Pflug via Burst (Shopify)

The post Blacksmith Hit a $1 Billion Valuation by Testing AI’s Code appeared first on CEO Medium.

Nine months ago, the same company launched with a $180 million seed round that valued it at $1.5 billion. Since then, its valuation has quadrupled, and total funding has reached $1.18 billion. Index partner Mark Xu takes a board seat as part of the deal. For context, this is a Series A. The letter has stopped describing a stage, and it now simply means the cheque was very large.

Inside the Flapping Airplanes Series A

Bloomberg first reported the round, and TechCrunch confirmed the shape of it on Wednesday. The $6 billion figure is a post-money valuation, which means it already counts the new billion. Valor, A* and Atreides also appear among the backers, alongside the seed investors who returned. In other words, almost every top-tier firm that saw the company early has doubled down, and several that missed it have bought their way in.

That syndicate matters because the company offers so little else to diligence. There is no product to test, no pricing page to study and no revenue line to model. Investors have underwritten a team, a research direction and a claim about how AI should be built. Everything else is still to come.

Flapping Airplanes is not an outlier, either. It belongs to a wave of research-first outfits, often called neolabs, that raise at scale before they sell anything. The same week brought reports that China’s DeepSeek had raised about $15 billion, while legal AI company Legora banked $550 million at an $8.8 billion valuation. Money is not scarce in AI right now. Credibility is, and that is what these rounds are really buying.

The bet: efficiency beats scale

The research direction is easy to state and hard to execute. Most frontier labs improve their models by feeding them more data on more chips. Flapping Airplanes argues that the industry has barely scratched the surface of what each chip can do. Its team has spent its first year building what it calls a megakernel framework, software that bundles many GPU tasks into a single program so the hardware rarely sits idle.

In practice, the founders say one kernel can run a full training job, serve live inference and even play a video game at the same time. The framework overlaps work that waits on data with work that crunches numbers, so neither stalls the other. Asher Spector, one of the founders, put the ambition plainly when the company launched: “A model that’s a million times more data efficient is probably a million times easier to put into the economy.”

The commercial plan, for now, is deliberately vague. Spector has said the lab eventually plans to sell its technology to other companies that want cheaper AI. Until then, the performance claims rest on the founders’ own word. That gap between claim and proof is the risk every investor in the round has accepted.

There is also a timing argument underneath the cheque. Compute is the AI industry’s biggest cost, and it stays scarce because demand keeps outrunning supply. A technology that squeezes more work from the same chips does not need to beat the frontier labs at their own game. It only needs to make their existing hardware cheaper to run, which is a far easier sale. That may be why strategics as well as venture firms wanted in: Nvidia’s presence among the backers reads as a hedge as much as an endorsement.

A founding team built like a talent magnet

The team is the asset, so it deserves a close look. Brothers Ben and Asher Spector founded the lab with Aidan Smith in late 2025. Asher completed a statistics PhD at Stanford, while Ben was a computer science PhD student there after founding Prod at MIT. Prod is a student-run accelerator that Sequoia partner David Cahn has called one of the most powerful talent incubators in Silicon Valley. Smith, meanwhile, is a former founder and Thiel Fellow who worked at Neuralink during college.

Cahn, an investor, describes the trio as building “the young person’s AGI lab”. Read that as a recruiting pitch as much as a compliment. A lab with $1 billion can hire almost anyone, but a lab famous for being where the best young researchers gather hires people money alone cannot reach. In the current market, that reputation compounds faster than interest.

What founders should take from this

First, capital follows asymmetry. A small team cannot outspend OpenAI or Google, so Flapping Airplanes attacks the cost curve instead. If the efficiency thesis works even partially, every model company becomes a potential customer. Founders in crowded markets should ask the same question: which assumption is the whole industry making that we can attack from the side?

Second, networks raise money before products do. The Spectors did not arrive as strangers. Prod gave them a deep web of talented peers, and Sequoia knew them before the company existed. Most founders cannot copy that history, but they can build the smaller version of it: stay close to the communities where your future co-founders, hires and investors already gather.

Third, and most importantly, a valuation like this is a promise that comes due. Temporal, which we covered when its $550 million raise showed AI agent infrastructure pulling in serious money, paired its price tag with a $250 million revenue run rate. Flapping Airplanes has paired its price tag with a research claim. Both can work, but only one of them has customers to fall back on if the story slows down.

The honest summary is that this round tells us less about one lab than about the market around it. Investors now believe the next phase of AI belongs to whoever makes intelligence cheaper, and they will pay unicorn prices years early for a credible shot at it. Whether that belief survives contact with results is the story of the next two years. More founder strategy and startup funding analysis lives in our Entrepreneurship & Startups hub.

Photo: Wikimedia Commons, CC0 1.0 (public domain)

The post No Product, No Revenue, $6 Billion: The Flapping Airplanes Bet appeared first on CEO Medium.