BUSINESS

“The attack surface is exploding,” CEO Bill McDermott said.

Costco has a very limited selection compared to other retailers its size.That’s an intentional choice that allows the company to keep prices down. By selling, for example, one SKU of ketchup rather than various sizes and packaging options, Costco maximizes how many of that unit it sells, giving it better bargaining power with its vendors.Essentially, when a retailer negotiates with a vendor, something I did for many years working at my family’s steel scaffolding business, the more you buy, the lower the price. If I ordered steel by the truckload, it was cheaper than a partial load order.Costco uses that principle, and the fact that it only stocks about 4,000 unique products in its warehouse (about 40% of what a traditional grocery store carries) gives the chain leverage to negotiate lower prices. By keeping the product count low, the warehouse club makes every partner earn their shelf space, and sometimes, it means that a store staple, or a new favorite, loses its spot.That can be traumatic to members who see an old (or new) favorite disappear. Costco members, however, can get a little insight into what products may soon go away if they know the chain’s secret code.Costco tells members what items are being discontinued Costco offers two types of merchandise. First, it has core staples that it always offers.Your warehouse club will have milk, coffee, cereal, and other basics. The brands, sizes, and packaging may change, but you won’t go to Costco and find that it’s not stocking the basics.In addition, Costco has its rotating merchandise. Some of those items, like clothing, seasonal sports gear, holiday products, and more, only appear for a short window.If you missed the recent Esprit sweatshirt, which was a viral hit that sold out quickly, according to Today.com, it’s probably not coming back.Some of these non-staple items, however, are regulars. Maybe Costco offers a K-Cup flavor you love, has a candy or chocolate you like, or stocks a frozen meal, beer brand, or something else you truly enjoy.In some cases, those items get phased out. Maybe they don’t sell well enough, or perhaps the manufacturer and Costco can’t reach a deal to keep the item on sale.More Costco:Costco makes silent gas change members will loveBig changes could be in store for CostcoCostco faces a food safety problem members need to knowWhen that happens, Costco discontinues the item without telling members. There is, however, a way to know when your favorite items are being phased out, but you need to know what to look for.”Costco’s limited inventory and constantly rotating selection are part of what gives the warehouse its treasure-hunt appeal. So when you stumble across a product you love, take a quick look at the shelf tag. If the large white price sign has a small asterisk in the upper-right corner, you’ve spotted what’s known as Costco’s ‘star of death,'” Food & Wine reported.

Costco regularly changes its merchandise. Shutterstock

Costco members need to know the codeThe symbol can mean two different things.”The Death Star is typically added to seasonal products that return annually and to unpopular products that will be permanently retired,” according to Delish.Costco does not share whether the item is being permanently removed.“That doesn’t mean that it’s always going to be discontinued forever,” said David Schwartz, co-author of “The Joy of Costco: A Treasure Hunt from A to Z,” in an interview with Delish. “That could mean that it’s gonna be given a rest for three months, and then come back.”Costco did not return Good Housekeeping’s request for comment on the secret code.Costco is careful with its inventoryCostco constantly adjusts its inventory partially to meet member needs and partly to keep costs down. Global situations like wars and the ongoing White House tariffs can have an impact on what the warehouse club stocks.CFO Gary Millerchip commented on the company’s inventory during Costco’s third-quarter earnings call.”The supply chain is generally stable, and our merchants feel good about our inventory position heading into the summer. We have relatively low inventory exposure to shipping issues stemming from the situation in the Middle East, but we continue to monitor the situation closely,” he said.Costco plans out each item it stocks, according to RTM Nexus CEO Dominick Miserandino.”It’s simply an economics of space. The average retailer could have 10,000 to 15,000 SKUs, but Costco could have 3,500 to 4,000,” he told TheStreet. That’s a significant difference (and a Walmart or Target could have two to three times that many products). So, yes, they’re going to be most efficient and only stock the shelves with what they think will sell.”(Costco SKU counts vary by warehouse.)CEO Ron Vachris shared some insight on how supply chain impacts and sourcing impacts pricing.”One of your examples was you saw some inventory that we had during higher tariffs. Now we’re getting the lower-priced goods in. We may go down earlier in those to get into those lower-priced goods quicker. We’re down quick on eggs when that commodity started dropping,” he said during the Q3 call.Related: More Costco members are paying double for one reason

There are a host of challenges at work as the ‘long’ 30-year Treasury yield logs its longest stretch above 5% in 19 years

Every household budget has a repair that gets postponed.The roof leak that is still only a stain on the ceiling. The transmission noise you learn to drive around. Nothing bad happens for years, and then it does, and the bill is far bigger than it would have been if the problem had been addressed earlier.Washington runs on the same logic, except the ceiling belongs to about 71 million people.Social Security has been on a published countdown since long before most of today’s retirees ever filed a claim. The retirement trust fund is now scheduled to run short in 2032, and the automatic benefit cut that follows would hit everyone at once, regardless of age, income, or how carefully anyone planned.Lawmakers have known about the gap for decades. They have also had workable fixes in front of them for just as long, including one that almost nobody remembers, which came from a Minnesota congressman most readers have never heard of.He introduced it in 1987. New modeling released this week says it would have done most of the job.What the 1987 flat-rate COLA proposal actually didEvery January, benefits rise by a percentage tied to inflation. A retiree collecting $1,200 a month gets that percentage. So does a retiree collecting $3,500 a month, which works out to a far larger raise in dollars.Compounded across 30 years of retirement, the distance between those two checks widens on autopilot.More Social Security:Vanguard warns of Social Security traps costing retireesSocial Security may send a $0 check for one common mistakeHow much Social Security crisis will cost your retirementA bill from Representative Tim Penny (D-Minn.) flipped the math. His flat-rate cost-of-living adjustment “would pay all beneficiaries the same COLA,” according to the Committee for a Responsible Federal Budget, set at the dollar amount collected by the beneficiary at the 20th percentile.Same raise for everyone, measured in dollars instead of percentages. Bigger increase for the retiree at the bottom, slower growth for the retiree at the top.The bill was introduced in the 100th Congress and never became law, according to Congress.gov records.Penny was a six-term congressman who chaired the Democratic Budget Group and the Porkbusters Coalition, according to the budget group’s biography of him. He now co-chairs that same organization, which is how a 39-year-old bill ended up back in circulation.It surfaced again on July 21, when the budget group published fresh estimates from Karen Smith of the Urban Institute. Starting the same policy in 2027 would close half of Social Security’s 75-year shortfall, her modeling found. Set at the 30th percentile instead, it closes about two-fifths.

Lawmakers skipped a flat-rate COLA in 1987, and new estimates show the cost.Getty Images

Why the Social Security math got worse while Congress waitedThe delay is the whole story here, and it is measurable. Score the identical policy at two different start dates, and you get two very different repair jobs.Flat-rate COLA at the 20th percentile starting in 1988, roughly 75% of the 75-year gap closed, based on Urban Institute DYNASIM4 projections.The same policy starting in 2027, about 50% of the gap closed, according to the Urban InstituteA COLA cap set at the median, about 25% closed, the Committee for a Responsible Federal Budget indicated. Switching the index to a chained Consumer Price Index, about 15% closed, based on scores from Social Security’s Office of the Chief Actuary.Switching to CPI-E, the version built for older consumers, widens the gap by about 10%, based on the same actuary scores.I lined up those first two rows because the comparison is the cleanest measure of procrastination I have seen in this debate. The identical policy lost roughly a third of its repair power between the Reagan administration and now.Meanwhile, the hole kept growing. The 2026 Trustees Report, which moved the depletion date forward again, puts Social Security’s long-term deficit at 4.42% of taxable payroll, up from 3.82% a year earlier.The retirement fund’s reserves are projected to run out in the fourth quarter of 2032, after which “less than full scheduled benefits would be payable,” according to the Social Security Administration.Translated out of actuarial language by the Bipartisan Policy Center, “Current and future beneficiaries alike will see their benefits cut by 22%.”What a flat-rate COLA would mean for your monthly checkStart with what the 2032 cut looks like in a checking account.The average retired worker collects $2,071 a month after this year’s 2.8% adjustment, the SSA’s 2026 cost-of-living fact sheet confirms. My arithmetic on that figure puts a 22% cut at roughly $456 a month, or about $5,470 a year, arriving with no phase-in and no warning letter.A flat-rate COLA changes who absorbs the slowdown. At the 20th percentile setting, the bottom fifth of lifetime earners would see benefits fall 3% by 2065, while the top fifth gives up 19%, the Urban Institute estimated.Related: Bipartisan senators target Social Security’s shortfallSet the flat rate one notch higher, at the 30th percentile, and the bottom fifth gets a 1% raise while the top fifth drops 17%.What struck me most in that distribution table was the payable-benefit line. Measured against what the trust fund can actually pay after 2032 rather than what the law promises, the lowest-income quintile ends up 13% to 14% better off under either version.Old-age poverty falls, too, by an estimated 5% under the 20th percentile design and 10% under the 30th.The formula matters more than it sounds. Percentage adjustments hold the gap between a small check and a large one steady in relative terms, then widen it in dollar terms every single year.Across a 25-year retirement, that compounding quietly decides how much of the program’s spending reaches people who need it, and how much reaches people who would be fine without it.What Congress can still do before the 2032 deadlineA flat-rate COLA is not a stand-alone rescue. On its own, it buys about two years before the combined trust funds run dry.Paired with an employer compensation tax that applies the employer half of the payroll tax to all wages and fringe benefits, the combined funds stay solvent for 75 years and beyond under the Urban Institute’s 2025 baseline. Under this year’s darker projections, that package gets most of the way there.Had the 1987 version passed, insolvency would have moved out to 2071 by the budget group’s rough estimate, with three-quarters of the gap through 2100 covered and room left for gradual changes to the taxable maximum or the retirement age.That option is gone. The 2027 adjustment is already being estimated in the 3.8% range, which means another year of percentage raises compounding the spread between the largest and smallest checks.Penny got his vindication 39 years late, which is worth nothing to anyone currently cashing a benefit check.The pattern is the useful part. Every option still on the table gets weaker the longer it sits there. The 2032 date is six years out, and whatever version of this fix survives to 2032 will be smaller and harsher than the one on offer now.Related: Analyst drastically lowers Social Security COLA Estimate

Stella Lefty, a rising singer-songwriter whose father is a billionaire co-founder of Groupon, made her first appearance on the Billboard Hot 100 chart’s top 10 this week.

Trump on Wednesday amped up threats to bomb key sites in Iran if any more ships were seized.

You spend weeks tuning an AI chatbot. Answers are accurate. Stakeholders sign off, and you ship it. Three months later, the system is confidently wrong about a third of what users ask. Nobody changed the model, and nobody touched the prompts. The world moved, pricing changed, a policy updated, a product spec shipped a new version, and the underlying knowledge store didn’t move with it.This is not a hypothetical. It’s one of the most common production failure modes in enterprise AI right now, and most data engineering teams don’t have the right tooling to catch it, regardless of how the AI system retrieves the data.The failure that doesn’t look like a failure An AI application doesn’t care whether it’s retrieving from a vector store, a document index, or an API call. Whatever the mechanism, nothing in a standard retrieval pipeline checks whether what it’s serving is still correct. A stale pricing document retrieves just as confidently as a current one, because the system is scoring relevance or availability, not correctness. A record with a silently missing field passes through just as cleanly as a complete one, for the same reason.So the failure is invisible by design. Outdated or incomplete data still scores high on relevance, or passes every check a data pipeline was built to run. The model answers with full confidence because the retrieved context looks authoritative. Every dashboard you’re watching stays green. The system looks like it’s working. It’s just wrong.I’ve watched a similar version of this happen outside the AI context, in a fintech pipeline. An upstream system changed a field without notifying downstream users. The pipeline did not fail; it simply propagated bad values into dashboards because the system only checked whether the job completed, not whether the data was still correct. The issue surfaced only when a customer noticed something inconsistent. By then, the bad data had already moved downstream. Whether it’s a document that’s gone stale or a field that’s gone silently missing, the failure shape is the same: the absence of an error is not the presence of correctness, and without building proper validation layers, nothing in the pipeline could identify the problem.Why this is a data engineering problemTeams that hit this failure tend to misdiagnose it, and they tend to do it twice.Blaming the model: The first instinct is to blame the model, try a different LLM, adjust the prompt. The real problem lies further upstream, at the data engineering layer, the same instinct behind the fintech failure above: monitoring built for the pipeline, not the data.Blaming the retrieval layer: Once the model’s ruled out, the next instinct is to blame the retrieval or context layer instead and buy a better one. The timing isn’t a coincidence: as enterprises push these systems into the real production world, this gap is exactly what’s starting to surface, and the vendor response has been everywhere. AWS just entered the “context layer” race with a knowledge graph that learns from agent usage. Snowflake’s new Horizon Context and Cortex Sense target the exact symptom this piece opened with: agents giving confident wrong answers because nothing governs the business logic underneath them. Both are real responses to a real problem, but they sit one layer above it; a knowledge graph still depends on whatever feeds it.The real problem lies further upstream, at the data engineering layer. Teams check whether a job ran, not whether the data it moved is still true, an instinct that predates AI by years. Monitoring is built for the pipeline, not for the data. What’s actually missing: Data observabilityData observability is a well-known concept that doesn’t get enough attention in how it’s actually implemented. The relevant metric isn’t a percentage — it’s coverage: what fraction of critical datasets have lineage that’s actually queryable, versus only living in someone’s head.Uber built a dedicated data quality and observability platform long before retrieval-augmented generation existed. Their Unified Data Quality platform supports more than 2,000 critical datasets and detects around 90% of data quality incidents before they reach downstream consumers.Netflix solved a different piece of the same problem, building a company-wide data lineage system so anyone could answer where a dataset came from and what touched it along the way. It maps dependencies across Kafka topics, ML models, and experimentation, not just warehouse tables. Similar to Uber, the platform was built for humans and now it has become more important with the rise in AI/LLM applications.Between them, Uber and Netflix cover two of the four things worth building for. In practice, I think about it as four dimensions, each measurable on its own terms.Correctness: Does each record conform to the shape and rules it’s supposed to, right field types, no unexpected nulls, values in range. Tools like Great Expectations and Soda handle this well: automated row and column-level validation instead of manual checks after something breaks. Track percentage of records passing validation per run.Freshness: Is the data still current relative to its source, not just current as of its last check. Track time since last successful update per source, with an SLA per dataset rather than one blanket threshold, since some sources need hourly refresh and others don’t.Consistency: Does the same fact read the same way everywhere it’s stored or indexed. This fails silently, it only shows up when two systems fed by the same source start disagreeing. A periodic cross-check between downstream destinations, flagging mismatch rate above a threshold, is enough to catch it early.Lineage: Can you trace any output back to its source and every transform it passed through, the same question Netflix built its system to answer. None of this requires infrastructure most data teams don’t already have. I know because I’ve built it, not just argued for it.At Socure, client data arrived in whatever shape the client felt like sending it, and occasionally, quietly wrong. The challenge was building a system where incorrect data could be identified before it propagated downstream. The same principles applied: Validate what arrived, understand where it came from, and prevent bad data from becoming someone else’s problem.Great Expectations became part of that foundation: schema and range validation at ingestion, per-source SLAs for freshness, cross-system checks for consistency, and file-level lineage. All of it sat behind a write-audit-publish pattern, where data landed in staging, was validated, and only moved downstream if it passed the required checks.The result showed up downstream: better accuracy across the board, in reporting, in the ML models, and in AI retrieval built on top of that same data.What to do Monday morningIf you’re running retrieval-based AI systems in production, the diagnostic question isn’t which model to try next or which retrieval architecture to migrate to. It’s four narrower questions: Is the underlying data validated against the standards required by its consumers?What’s the oldest piece of content currently being served with high confidence?Would two chunks of the same source ever disagree with each other in the same retrieval result?Could you trace where it came from if it turned out to be wrong?If you can’t answer those questions, then the gap lies in the pipeline between your source systems and whatever your agent reads from. That’s a data engineering fix, not a model swap or a vendor migration.Whether you’re building reporting pipelines, ML systems, or AI agents, correctness, freshness, consistency, and lineage are what make data trustworthy. AI simply exposes weaknesses that have existed in data engineering all along. 

A federal judge paused the merger, ruling states had raised “serious questions” about antitrust law.

Bank of America CEO Brian Moynihan turned heads during a NewsNation debate, warning that sticky inflation may leave the Federal Reserve with no choice but to raise interest rates again later this year. Moynihan argued that pipeline costs, energy pressures, and persistent household expenses will likely keep inflation elevated well into 2027 and 2028, with the soft-landing narrative running out of steam.Morgan Stanley, though, just entered the debate with a completely different reading of the same economy.Though BofA sees renewed tightening risk, Morgan Stanley believes recent data point in a very different direction. It’s a unique timing, as markets are already positioning around the next Fed move.Morgan Stanley rejects Wall Street’s renewed rate-hike panicMorgan Stanley just pushed back against a popular view that the Federal Reserve will have to resume tightening later this year.The most prominent proponent of the view is Bank of America, whose CEO recently doubled down on three quarter-point rate hikes in 2026, lifting the policy range to 4.25% to 4.5%, as stubborn inflation lingers into 2027 and 2028.More Fed:Cooler inflation delivers big win for Fed interest-rate betsFed’s Waller issues stark warning on inflation, interest ratesGoldman hints at Fed’s next interest-rate bet under WarshMorgan Stanley says that the latest inflation data point the other way. June headline CPI fell 0.42% month over month, while core CPI declined 0.02%. On a year-over-year basis, headline inflation slowed to 3.46%, while core CPI eased to 2.57%.The bank estimates that June headline PCE fell 0.08%, while core PCE rose just 0.17%. It expects inflation to run closer to a 2% annualized pace during the second half.That backs up Morgan Stanley’s out-of-consensus call for no Fed moves in 2026, followed by two quarter-point cuts in 2027, even as markets price one or two hikes.For context, markets assign a combined 71.8% probability to one or two hikes and an 87.7% probability to at least one hike by December, according to Investing.com.

Morgan Stanley expects cooling inflation to keep the Federal Reserve on hold.SimpleImages

Why Morgan Stanley thinks BofA may be wrong Morgan Stanley makes a couple of claims on the economy, but the most analytical of them concerns tariffs.Morgan Stanley forecasts that tariffs have bumped the overall price level by 62 basis points, which is close to its model’s 70-basis-point estimate. However, it lays out the case that tariffs have contributed nothing new to core-goods inflation since February.If that pass-through is essentially finished, Morgan Stanley sees another 60 to 70 basis points of goods disinflation over the next year. Cumulative tariff inflation is also expected to flatten out 0.62 percentage points rather than pushing upward.That, in many ways, goes against the “pipeline costs are still coming” argument from BofA. In my previous coverage, I acknowledged that BofA’s thesis weakens sharply if gasoline, services inflation, and wages cool, and Morgan Stanley is actually offering the hard evidence that this cooling might already be occurring.Could AI become the inflation problem the Fed does not need to fight? Interestingly, Morgan Stanley pointed to an unusual handoff developing within the inflation data.Tariff pressures are fading, but AI-backed demand might be replacing part of it. Case in point: Computer software and accessories have risen more than 20 percentage points since December, while strength in “other video equipment” underscores higher semiconductor costs. The AI boom has been linked a bit too much, I’d argue, to capital expenditures (CapEx).The steep increase in prices for the software, chips, and hardware needed to support that buildout is, in many ways, being ignored.Tariffs eventually wash through the economy, but the big issue with AI demand is that it can persist as businesses grow data centers, software budgets, and computing capacity. In effect, the inflation source might be shifting from imported costs to sustained domestic investment demand.Recent third-party evidence also supports that argument. Several company-level examples that offer a lot of clarity on the situation.Uber: Exhausted its entire 2026 AI budget within four months as token-based software costs surged, Fortune reported.Apple: Raised MacBook and iPad prices as AI demand tightened memory and storage supplies, according to Reuters. Dell: Repriced PCs and servers almost daily to offset rapidly rising component expenses, Channel Dive confirmed.Samsung: Increased Galaxy S26 and foldable-phone prices as higher memory costs pressured margins, according to Reuters.Microsoft: Raised Xbox console prices by $100 to $150, citing sharply higher memory and storage costs, XboxWire noted.Are markets already doing the Fed’s tightening for it? Morgan Stanley doesn’t feel the Federal Reserve needs to respond immediately with a rate increase.Its financial conditions index suggests markets have already delivered tightening equivalent to roughly a 39-basis-point Fed hike since Middle East hostilities began. Of that, about 21 basis points have occurred since the June Fed meeting alone. Higher Treasury yields and a firmer U.S. dollar were the primary catalysts, while resilient stocks and contained credit spreads offset some part of the restraint.Monetary policy also works through these same channels.Rising bond yields raise borrowing costs, a stronger dollar weighs on exports and multinational earnings, and tighter financing conditions can efficiently slow down investment and demand without the Fed formally lifting rates.Morgan Stanley’s argument is not about dismissing inflation risk, though, and it’s actually more about avoiding unnecessary over-tightening. AI demand is keeping goods inflation elevated, but markets might already be imposing enough pressure to justify patience rather than another hike.What does Morgan Stanley’s call mean for investors and consumers?Naturally, Morgan Stanley’s outlook favors long-duration assets that typically benefit when rate expectations fall.Growth stocks, REITs, homebuilders, and longer-dated bonds stand to gain immensely if the Fed stays on hold in 2026 and starts cutting in 2027. Banks might see a lot less support from growing loan yields, though lower rates could eventually ease credit stress and improve borrowing demand.However, the risk is that markets might price in relief too quickly.If AI-related demand continues to keep software, semiconductor, and hardware prices elevated, inflation might remain uneven, even if we see tariffs and energy pressures fade away.For consumers, a Fed pause prevents another immediate bump in variable borrowing costs, covering credit cards, adjustable-rate mortgages, and business loans. However, unchanged rates still leave financing expensive through year-end.Related: Bank of America CEO warns inflation will back Fed into a corner

More than 200 individuals had a total of more than $85 billion in tax-sheltered retirement accounts