Costco Employees Can Enjoy Labor Day

 

Most major retailers remain open on Labor Day, often running special sales. Costco is unusual in that it closes its U.S. stores, called “warehouses” by the company, on Labor Day. At first glance, closing on a potentially busy shopping day looks expensive. But the visible lost revenue may substantially overstate the true cost. Former Costco CFO Richard Galanti says most of those sales simply shift to the days before or after the holiday. A shopper planning to buy groceries, paper towels, or a television at Costco may just go on Sunday or Tuesday instead. The relevant cost of closing, therefore, is not the revenue Costco could have collected on Labor Day, but the contribution margin on sales that disappear altogether rather than shift to another day.

The benefit, seemingly obvious, is even harder to quantify. Costco CEO Ron Vachris says the company closes so employees can enjoy the holidays with family and friends. That benefit may extend beyond the value of the day off itself. A company willing to sacrifice sales for its employees may generate morale, loyalty, retention, and recruiting benefits that never appear on an accounting statement. They instead show up indirectly through lower turnover, easier hiring, or greater effort. Costco therefore faces two hidden quantities: the cost of closing may be much smaller than the day's lost sales suggest, while the value to employees may be much larger than the day's wages suggest. If most sales merely shift to neighboring days while employees highly value having the holiday off, closing the stores can create value even though neither side of that calculation is readily visible.

GitLab Restructuring

 

In May 2026, GitLab announced a major reorganization that will create roughly 60 smaller R&D teams with greater autonomy and end-to-end ownership, while eliminating as many as three layers of management. This episode illustrates a basic tradeoff in organizational design. Larger functional or divisional groups make communication within a specialty relatively easy, but they can create weak incentives and costly communication across organizational boundaries. A programmer might finish her assigned code, for example, while responsibility for whether the overall product actually works belongs somewhere else. GitLab explicitly identifies layers of management and “handoffs that dilute accountability” as problems it wants to eliminate. Putting the people needed to deliver a product into a smaller team gives them greater authority, and responsibility, over the decisions affecting that product.

But organizing around small product teams creates a different communication problem. Engineers working on similar technologies may now sit in different teams, so knowledge that once flowed naturally within a large engineering group must travel across product groups. Teams can also duplicate solutions to common problems or make decisions that work for their product but impose costs elsewhere in the company. GitLab is betting that this tradeoff has changed: it plans to use AI agents to automate some reviews, approvals, and handoffs, while its current development model emphasizes small, cross-functional teams with end-to-end accountability. Organization form depends on which communication problem is more costly. Divisions facilitate coordination among specialists but weaken incentives for the final outcome; product teams strengthen ownership of outcomes but make sharing knowledge and coordinating across products more difficult. GitLab’s restructuring suggests that it now believes the first problem has become more important than the second.

Boeing / SPEEA Negotiations

 

Boeing and its roughly 17,000 engineers and technical workers represented by the Society of Professional Engineering Employees in Aerospace (SPEEA) are bargaining over new labor contracts. Both sides are taking actions to raise their disagreement values. After workers rejected Boeing’s contract offers, they also voted to authorize their negotiating team to call a strike. Boeing, meanwhile, has begun implementing a strike contingency plan designed to keep the business operating if workers walk out. Boeing says the plan is intended to allow it to continue aircraft deliveries and meet other customer commitments during a strike. Boeing also says it had hoped to spend some of those resources on employee incentives but, following the rejection, redirected the money toward strike preparations. Thus, each side is taking actions that make it easier to say “no” to the other side.

They are committing to playing hard rather than simply threatening to do so. Each side hopes these actions will be viewed as credible so as to affect the final agreement in their favor. A strike authorization strengthens the union negotiators’ ability to credibly threaten a walkout, while Boeing’s contingency investments reduce the cost to Boeing of allowing a strike to continue. Deliberately reducing your flexibility can improve the deal you receive. If your opponent knows that you have committed yourself to saying no, they may have to offer you more to get you to say yes.

Walmart Gives its FinTech a Foot in the Door

 

Fintech companies such as Chime and Current face a basic problem: acquiring customers can be expensive. They must advertise, offer promotions, or form partnerships to persuade consumers to download an app and begin using their financial products. Walmart’s majority-owned fintech, OnePay, has a very different starting point. According to Semafor, most of OnePay’s roughly 7 million customers have come from Walmart’s 1.5 million employees and approximately 150 million shoppers. OnePay CEO Omer Ismail argues that these customers are effectively acquired for free: Walmart already has the stores, website, app, checkout system, customer relationships, and traffic needed to put OnePay in front of millions of potential users.

Economy of scope can occur when an input created for one product can be shared to produce another product more cheaply. Walmart’s enormous distribution network was built to sell groceries, clothing, electronics, and other merchandise, but the same network can distribute financial services. For example, OnePay is integrated directly into Walmart’s physical and digital checkout channels, where installment loans are now powered by Klarna. A standalone fintech must build both a financial product and a way to reach customers. Walmart can use an asset it already owns to do both retailing and finance. This allows diversification to create value. Both businesses using the same underlying asset creates economies of scope across the businesses.

Accepting Late Bids?

 

Spirit Airlines’ bankruptcy has produced an unusual auction: Google agreed to pay $10 million for a trove of Spirit’s internal corporate data, including roughly 100 million emails and 500 million Microsoft Teams messages, beating a $7.5 million bid from AI firm Mercor. But after the auction closed, AI startup Micro1 offered $12.5 million, a sizable bump from Google’s bid. The bankruptcy court must now decide whether to consider the late offer. At first glance, the answer seems obvious. Spirit’s creditors want as much money as possible, so why not accept the higher bid?

Because changing the rules after an auction can affect bidding before the auction. A well-designed auction gives bidders confidence that deadlines and procedures are credible. If firms expect a losing bidder to get another chance after seeing the winning price, they have an incentive to hold back during the formal auction and wait to top the winner afterward. That can reduce competition and lower expected revenue in future auctions. Bidders on potential auctions for additional Spirit assets may adjust their bids downward. Conversely, rigidly rejecting a substantially higher late offer leaves money on the table today. There is an auction design tension because maximizing the highest observed bid is not necessarily the same as maximizing the seller’s expected revenue. Sometimes committing not to accept a better offer later can induce bidders to make better offers now.

WaPo: Social Security and Medicare going bankrupt

 

How AI Might Affect Demand

Artificial intelligence is often portrayed as a technology that will lower costs by replacing skilled workers, but ophthalmology provides an interesting counterexample. AI can now perform one relatively routine task, screening patients with diabetes for diabetic retinopathy, without requiring an ophthalmologist to examine every patient. In the ACCESS randomized trial, offering an autonomous AI eye exam during a diabetes visit increased screening completion from just 22% with conventional referral to 100%. Among patients receiving an abnormal AI result, 64% subsequently visited an eye-care provider. AI therefore substitutes for some ophthalmologist labor while dramatically increasing the number of patients who make it through the first stage of the eye-care supply chain. (Nature)

AI has lowered the cost of one link in the screening → referral → treatment supply chain. That cost reduction need not appear as lower overall expenses. Instead, it can appear as improved quality that increases demand: screening becomes easier, more patients are screened, and ophthalmologists can concentrate on patients who actually need their expertise. A 2026 Johns Hopkins study of 3,745 adults with diabetes provides evidence of this downstream effect, finding that AI screening in primary-care offices increased presentation to specialist eye care among an at-risk population. A recent meta-analysis of AI diabetic-retinopathy screening likewise concludes that it is generally cost-effective, particularly when it expands screening to underserved populations. (Nature)

The interesting result is that a cost-reducing technology can increase rather than decrease demand for expensive downstream services. Cheaper screening means more screening; more screening discovers more disease; and more detected disease generates referrals and treatment. Total spending on eye care could therefore rise even as the cost of producing a given level of eye health falls. The productivity gain results more in demand augmented by higher quality (shifting demand out) rather than lower price (moving along the demand curve). This may be a useful lesson for thinking about AI throughout the economy. When innovation improves one stage of a supply chain, its benefits need not appear primarily as lower prices or fewer workers. They may instead appear as better matching, greater use of complementary services, and higher-quality final output, in this case, healthier eyes.