Anthropic on Monday released Code Review, a multi-agent code review system built into Claude Code that dispatches teams of AI agents to scrutinize every pull request for bugs that human reviewers routinely miss. The feature, now available in research preview for Team and Enterprise customers, arrives on what may be the most consequential day in the company’s history: Anthropic simultaneously filed lawsuits against the Trump administration over a Pentagon blacklisting, while Microsoft announced a new partnership embedding Claude into its Microsoft 365 Copilot platform.

The convergence of a major product launch, a federal legal battle, and a landmark distribution deal with the world’s largest software company captures the extraordinary tension defining Anthropic’s current moment. The San Francisco-based AI lab is simultaneously trying to grow a developer tools business approaching $2.5 billion in annualized revenue, defend itself against an unprecedented government designation as a national security threat, and expand its commercial footprint through the very cloud platforms now navigating the fallout.

Code Review is Anthropic’s most aggressive bet yet that engineering organizations will pay significantly more — $15 to $25 per review — for AI-assisted code quality assurance that prioritizes thoroughness over speed. It also signals a broader strategic pivot: the company isn’t just building models, it’s building opinionated developer workflows around them.

How a team of AI agents reviews your pull requests

Code Review works differently from the lightweight code review tools most developers are accustomed to. When a developer opens a pull request, the system dispatches multiple AI agents that operate in parallel. These agents independently search for bugs, then cross-verify each other’s findings to filter out false positives, and finally rank the remaining issues by severity. The output appears as a single overview comment on the PR along with inline annotations for specific bugs.

Anthropic designed the system to scale dynamically with the complexity of the change. Large or intricate pull requests receive more agents and deeper analysis; trivial changes get a lighter pass. The company says the average review takes approximately 20 minutes — far slower than the near-instant feedback of tools like GitHub Copilot’s built-in review, but deliberately so.

“We built Code Review based on customer and internal feedback,” an Anthropic spokesperson told VentureBeat. “In our testing, we’ve found it provides high-value feedback and has helped catch bugs that we may have missed otherwise. Developers and engineering teams use a range of tools, and we build for that reality. The goal is to give teams a capable option at every stage of the development process.”

The system emerged from Anthropic’s own engineering practices, where the company says code output per engineer has grown 200% over the past year. That surge in AI-assisted code generation created a review bottleneck that the company says it now hears about from customers on a weekly basis. Before Code Review, only 16% of Anthropic’s internal PRs received substantive review comments. That figure has jumped to 54%.

Crucially, Code Review does not approve pull requests. That decision remains with human reviewers. Instead, the system functions as a force multiplier, surfacing issues so that human reviewers can focus on architectural decisions and higher-order concerns rather than line-by-line bug hunting.

Why Anthropic thinks $20 per review is a bargain

The pricing will draw immediate scrutiny. At $15 to $25 per review, billed on token usage and scaling with PR size, Code Review is substantially more expensive than alternatives. GitHub Copilot offers code review natively as part of its existing subscription, and startups like CodeRabbit operate at significantly lower price points. Anthropic’s more basic code review GitHub Action — which remains open source — is itself a lighter-weight and cheaper option.

Anthropic frames the cost not as a productivity expense but as an insurance product. “For teams shipping to production, the cost of a shipped bug dwarfs $20/review,” the company’s spokesperson told VentureBeat. “A single production incident — a rollback, a hotfix, an on-call page — can cost more in engineer hours than a month of Code Review. Code Review is an insurance product for code quality, not a productivity tool for churning through PRs faster.”

That framing is deliberate and revealing. Rather than competing on speed or price — the dimensions where lightweight tools have an advantage — Anthropic is positioning Code Review as a depth-first tool aimed at engineering leaders who manage production risk. The implicit argument is that the real cost comparison isn’t Code Review versus CodeRabbit, but Code Review versus the fully loaded cost of a production outage, including engineer time, customer impact, and reputational damage.

Whether that argument holds up will depend on the data. Anthropic has not yet published external benchmarks comparing Code Review’s bug-detection rates against competitors, and the spokesperson did not provide specific figures on bugs caught per dollar or developer hours saved when asked directly. For engineering leaders evaluating the tool, that gap in publicly available comparative data may slow adoption, even if the theoretical ROI case is compelling.

What the internal numbers reveal — and what they don’t

Anthropic’s internal usage data provides an early window into the system’s performance characteristics. On large pull requests exceeding 1,000 lines changed, 84% receive findings, averaging 7.5 issues per review. On small PRs under 50 lines, that drops to 31% with an average of 0.5 issues. The company reports that less than 1% of findings are marked incorrect by engineers.

That sub-1% figure is the kind of stat that demands careful unpacking. When asked how “marked incorrect” is defined, the Anthropic spokesperson explained that it means “an engineer actively resolving the comment without fixing it. We’ll continue to monitor feedback and engagement while Code Review is in research preview.”

The methodology matters. This is an opt-in disagreement metric — an engineer has to take the affirmative step of dismissing a finding. In practice, developers under time pressure may simply ignore irrelevant findings rather than actively marking them as wrong, which would cause false positives to go uncounted. Anthropic acknowledged the limitation implicitly by noting the system is in research preview and that it will continue monitoring engagement data. The company has not yet conducted or published a controlled evaluation comparing agent findings against a ground-truth baseline established by expert human reviewers.

The anecdotal evidence is nonetheless striking. Anthropic described a case where a one-line change to a production service — the kind of diff that typically receives a cursory approval — was flagged as critical by Code Review because it would have broken authentication for the service. In another example involving TrueNAS’s open-source middleware, Code Review surfaced a pre-existing bug in adjacent code during a ZFS encryption refactor: a type mismatch that was silently wiping the encryption key cache on every sync. These are precisely the categories of bugs — latent issues in touched-but-unchanged code, and subtle behavioral changes hiding in small diffs — that human reviewers are statistically most likely to miss.

A Pentagon lawsuit casts a long shadow over enterprise AI

The Code Review launch does not exist in a vacuum. On the same day, Anthropic filed two lawsuits — one in the U.S. District Court for the Northern District of California and another in the D.C. Circuit Court of Appeals — challenging the Trump administration’s decision to label the company a supply chain risk to national security, a designation historically reserved for foreign adversaries.

The legal confrontation stems from a breakdown in contract negotiations between Anthropic and the Pentagon. As CNN reported, the Defense Department wanted unrestricted access to Claude for “all lawful purposes,” while Anthropic insisted on two redlines: that its AI would not be used for fully autonomous weapons or mass domestic surveillance. When talks collapsed by a Pentagon-set deadline on February 27, President Trump directed all federal agencies to cease using Anthropic’s technology, and Defense Secretary Pete Hegseth formally designated the company a supply chain risk.

According to CNBC, the complaint alleges that these actions are “unprecedented and unlawful” and are “harming Anthropic irreparably,” with the company stating that contracts are already being cancelled and “hundreds of millions of dollars” in near-term revenue are in jeopardy.

“Seeking judicial review does not change our longstanding commitment to harnessing AI to protect our national security,” the Anthropic spokesperson told VentureBeat, “but this is a necessary step to protect our business, our customers, and our partners. We will continue to pursue every path toward resolution, including dialogue with the government.”

For enterprise buyers evaluating Code Review and other Claude-based tools, the lawsuit introduces a novel category of vendor risk. The supply chain risk designation doesn’t just affect Anthropic’s government contracts — as CNBC reported, it requires defense contractors to certify they don’t use Claude in their Pentagon-related work. That creates a chilling effect that could extend well beyond the defense sector, even as the company’s commercial momentum accelerates.

Microsoft, Google, and Amazon draw a line around Claude’s commercial availability

The market’s response to the Pentagon crisis has been notably bifurcated. While the government moved to isolate Anthropic, the company’s three largest cloud distribution partners moved in the opposite direction.

Microsoft on Monday announced it is integrating Claude into Microsoft 365 Copilot through a new product called Copilot Cowork, developed in close collaboration with Anthropic. As Yahoo Finance reported, the service enables enterprise users to perform tasks like building presentations, pulling data into Excel spreadsheets, and coordinating meetings — the kind of agentic productivity capabilities that sent shares of SaaS companies like Salesforce, ServiceNow, and Intuit tumbling when Anthropic first debuted its Cowork product on January 30.

The timing is not coincidental. As TechCrunch reported last week, Microsoft, Google, and Amazon Web Services all confirmed that Claude remains available to their customers for non-defense workloads. Microsoft’s legal team specifically concluded that “Anthropic products, including Claude, can remain available to our customers — other than the Department of War — through platforms such as M365, GitHub, and Microsoft’s AI Foundry.”

That three of the world’s most powerful technology companies publicly reaffirmed their commitment to distributing Anthropic’s models — on the same day the company sued the federal government — tells enterprise customers something important about the market’s assessment of both Claude’s technical value and the legal durability of the supply chain risk designation.

Data security and what enterprise buyers need to know next

For organizations considering Code Review, the data handling question looms especially large. The system necessarily ingests proprietary source code to perform its analysis. Anthropic’s spokesperson addressed this directly: “Anthropic does not train models on our customers’ data. This is part of why customers in highly regulated industries, from Novo Nordisk to Intuit, trust us to deploy AI safely and effectively.”

The spokesperson did not detail specific retention policies or compliance certifications when asked, though the company’s reference to pharmaceutical and financial services clients suggests it has undergone the kind of security review those industries require.

Administrators get several controls for managing costs and scope, including monthly organization-wide spending caps, repository-level enablement, and an analytics dashboard tracking PRs reviewed, acceptance rates, and total costs. Once enabled, reviews run automatically on new pull requests with no per-developer configuration required.

The revenue figure Anthropic confirmed — a $2.5 billion run rate as of February 12 for Claude Code — underscores just how quickly developer tooling has become a material revenue line for the company. The spokesperson pointed to Anthropic’s recent Series G fundraise for additional context but did not break out what share of total company revenue Claude Code now represents.

Code Review is available now in research preview for Claude Code Team and Enterprise plans. Whether it can justify its premium in a market already crowded with cheaper alternatives will depend on whether Anthropic can convert anecdotal bug catches and internal usage stats into the kind of rigorous, externally validated evidence that engineering leaders with production budgets require — all while navigating a legal and political environment unlike anything the AI industry has previously faced.

The Netherlands’ military intelligence service and domestic intelligence agency have issued a join warning claiming that Russian hackers have launched “a large-scale global cyber campaign to gain access to Signal and WhatsApp accounts belonging to dignitaries, military personnel and civil servants.” According to the Dutch alert, hackers are imitating support chatbots to trick key targets into revealing their PINs for those communication platforms, which allows the bad actors to access incoming messages.

Last year in the US, the Pentagon advised members not to use Signal after the platform was subjected to similar phishing scams by Russian hackers. (Although the same US military leaders proved capable of creating their own security breaches without foreign interference just days prior.)

Having another national government raise concerns about Signal and WhatsApp phishing scams offers yet another reminder to never provide security details or click links without a check on who is really asking for your info.

Innovation rarely happens in isolation. Usually, the systems that engineers design are shaped by global teams whose members’ knowledge and ideas move across borders as easily as data.

That is especially true in my field of robotics and automation—where hardware, software, and human workflows function together. Progress depends not only on technical skill but also on how engineers frame problems and evaluate trade-offs. My career has shown me how cross-cultural experiences can shape the framing.

Working across different cultures has influenced how I approach collaboration, design decisions, and risk. I am an IEEE member and a mechanical engineer at Re:Build Fikst, in Wilmington, Mass., but I grew up in India and began my engineering education there.

Experiencing both work environments has reinforced the idea that diversity in science, technology, engineering, and mathematics fields is not only about representation; it is a technical advantage that affects how systems are designed and deployed.

Gaining experience across cultures

I began my training as an undergraduate student in electrical and electronics engineering at Amity University, in Noida. While studying, I developed a strong foundation in problem-framing and disciplined adaptability.

Working on a project requires identifying what the system needs to demonstrate and determining how best to validate that behavior within defined parameters. Rather than starting from idealized assumptions, Amity students were encouraged to focus on essential system behavior and prioritize the variables that most influenced the technology’s performance.

The approach reinforced first-principles thinking—starting from fundamental physical or system-level behavior rather than defaulting to established solutions—and encouraged the efficient use of available resources.

At the same time, I learned that efficiency has limits. In complex or safety-critical systems, insufficient validation can introduce hidden risks and reduce reliability. Understanding when simplicity accelerates progress and when additional rigor is necessary became an important part of my development as an engineer.

After getting my undergraduate degree, I moved to the United States in 2021 to pursue a master’s degree in robotics and autonomous systems at Arizona State University in Tempe. I encountered a new engineering culture in the United States.

In the U.S. research and development sector, especially in robotics and automation, rigor is nonnegotiable. Systems are designed to perform reliably across many cycles, users, and conditions. Documentation, validation, safety reviews, and reproducibility are integral to the process.

Those expectations do not constrain creativity; they allow systems to scale, endure, and be trusted.

Moving between the two different engineering cultures required me to adjust. I had to balance my instinct for efficiency with a more formal structure. In the United States, design decisions demand more justification. Collaboration means aligning with scientists, software engineers, and technicians. Each discipline brings different priorities and definitions of success to the team.

Over time, I realized that the value of both experiences was not in choosing one over the other but in learning when to apply each.

The balance is particularly critical in robotics and automation. Resourcefulness without rigor can fail at scale. A prototype that works in a controlled lab setting, for example, might break down when exposed to different users, operating conditions, or extended duty cycles.

At the same time, rigor without adaptability can slow innovation, such as when excessive documentation or overengineering delays early-stage testing and iteration.

Engineers who navigate multiple educational and professional systems often develop an intuition for managing the tension between the different experiences, building solutions that are robust and practical and that fit real-world workflows rather than idealized ones.

Much of my work today involves integrating automated systems into environments where technical performance must align with how people will use them. For example, a robotic work cell (a system that performs a specific task) might function flawlessly in isolation but require redesign once operators need clearer access for loading materials, troubleshooting faults, or performing routine maintenance. Similarly, an automated testing system must account not only for ideal operating conditions but also for how users respond to error messages, interruptions, and unexpected outputs.

In practice, that means thinking beyond individual components to consider how systems will be operated, maintained, and restored to service after faults or interruptions.

My cross-cultural background shapes how I evaluate design trade-offs and collaboration across disciplines.

How diverse teams can help improve tech design

Engineers trained in different cultures can bring distinct approaches to the same problem. Some might emphasize rapid iteration while others prioritize verification and robustness. When perspectives collide, teams ask better questions earlier. They challenge defaults, find edge cases, and design technologies that are more resilient to real-world variability.

Diversity of thought is certainly important in robotics and automation, where systems sit at the intersection of machines and people. Designing effective automation requires understanding how users interact with technology, how errors propagate, and how different environments influence the technology. Engineers with cross-cultural experience often bring heightened awareness of the variability, leading to better design decisions and more collaborative teams.

Engineers from outside of the United States play a critical role in the country’s research and development ecosystem, especially in interdisciplinary fields. Many of us act as bridges, connecting problem-solving approaches, expectations, and design philosophies shaped in different parts of the world. We translate not just language but also engineering intent, helping teams move from theories to practical deployment.

As robotics and automation continue to evolve, the challenges ahead—including scaling experimentation, improving reproducibility, and integrating intelligent systems into real-world environments—will require engineers who are comfortable working across boundaries. Navigating boundaries, which could be geographic, disciplinary, or cultural, is increasingly part of the job.

The engineering ecosystems in India and the United States are complex, mature, and evolving. My journey in both has taught me that being a strong engineer is not about adopting a single mindset. It’s about knowing how to adapt.

In an interconnected, multinational world, innovation belongs to engineers who can navigate the differences and turn them into strengths.

Apple went budget with the MacBook Neo. Now reportedly the company is preparing to go high end, and high price.

Apple is gearing up to launch a MacBook “Ultra” in the fall, outfitted with the first OLED display in MacBook history, according to a report from Bloomberg’s Mark Gurman.

Gurman says the new laptop will have a touchscreen and new M6 chips. Last month, Bloomberg reported that Apple could be launching its first touchscreen MacBook — with a Dynamic Island. The Dynamic Island is a feature on more recent iPhone models — iPhone 14 and later — that’s shaped like a long horizontal pill atop the screen and shows alerts, notifications, timers and music.

It if happens, the Ultra would be at the opposite end of the cost spectrum from the Neo, which Apple launched earlier this month for $599 as a rival to Chromebooks and Windows laptops in the same price range. Gurman says the MacBook Ultra could cost 20 percent more than the new MacBook Pro (M5 series), which Apple lists at $1,699.

If you want to read more about the new Neo, CNET has been all over it: Here’s why students might love it, the colors we liked and didn’t like and why the Neo is a really is a game-changer.

Gurman suggested that Apple might keep selling the M5 series MacBook Pro even after the Ultra launches. That would give the company a wider range of MacBooks at various costs — the Neo ($599), the Air (starting at $1,099), the Pro ($1,699) and the Ultra. 

Tech analyst Paolo Pescatore said adding the premium MacBook Ultra would “signal a clear shift in strategy” for Apple.

“If this turns out to be the case, then Apple appears to be stretching the Mac further upmarket,” Pescatore told CNET. “The opportunity is to drive higher spending and keep premium users firmly within Apple’s ecosystem. The challenge will be avoiding confusion, especially if the lines between MacBook Pro and iPad Pro become even more blurred.”

Gurman also said that Apple’s first foldable iPhone, rumored to be launched later this year, might be called the “iPhone Ultra.”

Apple’s long-rumoured HomePad smart display could include a magnetic wall-mounting system similar to MagSafe, according to a new leak.

The feature would allow the device to snap securely onto a wall-mounted fixture. As a result, it may be easier to position the display in different areas of the home.

The detail comes from prototype collector and leaker Kosutami, who claims to have seen a version of the device in person. In a post on X, they said one HomePad prototype includes a MagSafe-style snap-to-wall mechanism. Additionally, it has built-in doorbell integration, suggesting the device could double as a central hub for smart home alerts and video feeds.

from the can’t-trust-this-doj dept

When companies sued to block Trump’s IEEPA tariffs last year, one of the key arguments they made was obvious: if these tariffs turn out to be illegal, we’ll never get our money back. We need an injunction now. The government had an equally confident response: relax, if the tariffs are struck down, we’ll just issue refunds. No big deal. No injunction needed.

Multiple courts bought it. And now, with the Supreme Court having ruled the tariffs unlawful and a judge ordering the refunds, CBP is telling the court that it actually can’t comply with the order. The promises that defeated all those injunctions? Turns out nobody bothered to check whether they were actually true.

Once again, courts trusted what the government told them. Once again, it turns out they were wrong.

Let’s rewind to see how we got here.

Back in April 2025, when importers like V.O.S. Selections were seeking a preliminary injunction to stop the tariffs from being collected, the Department of Justice told the Court of International Trade there was simply no need for such drastic relief. In its brief opposing the injunction, the DOJ was explicit:

And, even if future entries are liquidated, defendants do not intend to oppose the Court’s authority to order reliquidation of entries of merchandise subject to the challenged tariffs if the tariffs are found in a final and unappealable decision to have been unlawfully collected. Such reliquidation would result in a refund of all duties determined to be unlawfully assessed, with interest.

No injunction needed! Refunds would flow. With interest, even. The government repeated this refund promise in case after case after case. In the Learning Resources stay motion, the government told the D.C. district court that there was no risk at all that the government wouldn’t repay:

For any plaintiff who is an importer, even if a stay is entered and defendants do not prevail on appeal, plaintiffs will assuredly receive payment on their refund with interest. “[T]here is virtually no risk” to any importer that they “would not be made whole” should they prevail on appeal. See Sunpreme v. United States, 2017 WL 65421, at *5 (Ct. Int’l Trade Jan. 5, 2017). The most “harm” that could incur would be a delay in collecting on deposits. This harm is, by definition, not irreparable.

In the Axle case, same thing.

In any event, were Axle to ultimately prevail, it could receive a refund of duties paid that would otherwise be eligible for duty-free treatment under the de minimis exemption on any unliquidated entries. 28 U.S.C. §§ 2643-44. To the extent any future entries are liquidated, the Court may order reliquidation of entries subject to the challenged de minimis exemption if the duties paid by Axle are, in a final and unappealable decision, found to have been unlawfully collected. Such reliquidation would result in a refund of all duties determined to be unlawfully assessed, with interest.

In the Princess Awesome joint stipulation, the government formally agreed that there was nothing to fear about getting repaid:

Defendants stipulate that they will not oppose the Court’s authority to order reliquidation of entries of merchandise subject to the challenged IEEPA duties and that they will refund any IEEPA duties found to have been unlawfully collected, after a final and unappealable decision has been issued finding the duties to have been unlawfully collected

And the courts relied on these representations. In December 2025, when AGS Company Automotive Solutions sought a preliminary injunction to stay the liquidation of its entries, the three-judge panel denied the motion specifically because of the government’s refund promises:

For the reasons stated above, we conclude that the Government has taken the “unequivocal position” that “liquidation will not affect the availability of refunds after a final decision” in V.O.S. Gov’t Resp. at 2–3. The Government would be judicially estopped from “assum[ing] a contrary position” in the future.

Note the court’s foresight here. The panel explicitly invoked judicial estoppel—basically saying “okay, now that you’ve said this to a court, you’re bound by it going forward.” You get the sense that the court had a sense of where all this was going.

Then the Supreme Court ruled 6-3 in Learning Resources, Inc. v. Trump that the IEEPA tariffs were unlawful. Judge Eaton at the Court of International Trade—designated as the sole judge to handle IEEPA refund cases—last week ordered CBP basically pay back everyone who paid an IEEPA tariff. Everyone. Not just those who sued.

In court, when the DOJ pushed back a bit, Eaton was blunt:

“Customs knows how to do this,” Eaton said during a court hearing on Wednesday. “They do it every day. They liquidate entries and make refunds.”

Enter the declaration of Brandon Lord, CBP’s Executive Director of the Trade Programs Directorate, filed the day after Judge Eaton’s order. He points out that, actually, there are a TON of tariffs to repay.

As of March 4, 2026, over 330,000 importers have made a total of over 53 million entries in which they have deposited or paid duties imposed pursuant to the International Emergency Economic Powers Act (“IEEPA”), 50 U.S.C. § 1701 et seq. (the “subject entries”). As of March 4, 2026, the total amount of IEEPA duties and estimated duty deposits collected pursuant to IEEPA is approximately $166 billion. Approximately 20.1 million entries remain unliquidated as of March 4, 2026.

And, apparently, it turns out that CBP is not at all prepared to repay what it owes:

In light of the Court’s March 5, 2026 amended order, CBP is now facing an unprecedented volume of refunds. Its existing administrative procedures and technology are not well suited to a task of this scale and will require manual work that will prevent personnel from fully carrying out the agency’s trade enforcement mission. Personnel would be redirected from responsibilities that serve to mitigate imminent threats to national security and economic security.

Lord’s declaration lays out a big list of technical and logistical obstacles. CBP’s Automated Commercial Environment (ACE) system can apparently only batch-process 10,000 entry summary lines at a time, and there are over 1.6 billion entry summary lines that need updating. Importers frequently lumped their IEEPA duties together with other duties on the same line, meaning CBP personnel would have to manually untangle the amounts. Processing each individual refund takes about 5 minutes, which across 53 million entries works out to over 4.4 million hours.

There’s also a mess involving different entry types and automatic liquidation timelines—Lord’s declaration goes into a bunch of technical details about “formal” vs. “informal” entries, claiming that 4 million entries will automatically process next week and “CBP does not have a process to prevent” it. Even if the legal details are deep in the weeds, the message is clear: even with the Supreme Court ruling in hand, CBP claims parts of this train are still moving and they can’t stop it.

CBP says it can build new ACE functionality in 45 days that would streamline the process. The proposed system actually sounds reasonable. Which makes it worse: if you spent the better part of a year telling every court that would listen that refunds were totally manageable, that there was “virtually no risk” importers wouldn’t be made whole, that “such reliquidation would result in a refund of all duties determined to be unlawfully assessed, with interest”—then maybe, just maybe, you should have spent some of that year building the system to actually do it? Send over a DOGE bro or two to vibe code up a solution?

The Supreme Court case wasn’t a surprise. The government was a party to it. They knew the ruling was coming. They knew that if they lost, refunds would be necessary on a massive scale. And even just based on how the oral arguments went, they should have known how this would turn out.

Instead, CBP appears to have done absolutely nothing to prepare. The government used the promise of easy refunds as a sword to defeat injunction after injunction, convincing courts that importers would suffer no irreparable harm because the money could always be returned. Having successfully avoided those injunctions—allowing the tariffs to keep being collected for months on end, swelling that $166 billion pot—the government now tells the court that returning the money is an operational nightmare that requires new technology it hasn’t built yet.

This is exactly the scenario the AGS panel warned about. And if the government tries to argue that it can’t provide refunds—rather than that it just needs more time—it will run headlong into the judicial estoppel doctrine that the court already set up like a tripwire. As the AGS panel put it, quoting the Supreme Court: “where a party assumes a certain position in a legal proceeding, and succeeds in maintaining that position, he may not thereafter, simply because his interests have changed, assume a contrary position.”

Every month the government successfully avoided an injunction was another month it kept collecting tariffs. That $166 billion didn’t accumulate by accident. The government had every incentive to promise easy refunds and zero incentive to actually prepare for them. The longer importers waited for relief, the bigger the pot grew.

And now, with the Supreme Court having ruled those tariffs illegal, and with courts having explicitly warned that the government would be judicially estopped from changing its position, CBP says it needs 45 days to build new software before it can start writing checks.

“Customs knows how to do this,” Judge Eaton said. “They do it every day.”

Maybe. But apparently nobody in the entire federal government thought to ask whether CBP could actually deliver on the promises DOJ was making to court after court after court. Either that, or they just didn’t care what the answer was.

Filed Under: brandon lord, cbp, court of international trade, ieepa, refunds, tariffs, trump admin