TheChek News Flow

The company’s products have had breakdowns and safety issues, documents show

The Navy was attempting to launch and recover more than 30 drone boats from a combat ship off the coast of California in May when more than a dozen of the uncrewed vessels failed to carry out their missions. The boats had rejected their inputs and automatically idled as a fail-safe, making them “dead” in the water.

The botched experiment quickly became a potential hazard to other vessels in the exercise. Military personnel scrambled overnight to clean up the mess, towing the boats to shore until 9 a.m. the next day.

The drone boats were relying on autonomy software called Lattice, made by California-based Anduril Industries. The Navy said the exercise was handled safely, but the incident alarmed Navy personnel, who said in a routine follow-up report that company representatives had misguided the military. In comments that were unusual for such a report, which was viewed by The Wall Street Journal, four sailors warned of “continuous operational security violations, safety violations, and contracting performer misguidances (Anduril Industries).” If the software configuration wasn’t immediately corrected and vetted, they wrote, there would be “extreme risk to force and potential for loss of life.”

Since its founding in 2017, Anduril Industries has become one of the hottest companies in a crowded field of defense-tech startups, promising to deliver hardware and software that will usher in a new era of autonomous warfare and equip the U.S. military with the speed that only a startup can offer. The privately held company was valued at more than $30 billion in its last funding round and has scored an impressive number of military contracts to build prototypes of everything from unmanned jet fighters to mixed-reality headsets to battlefield-management systems.

The company’s founder, tech billionaire Palmer Luckey, has made his convictions clear. “We spend our own money building defense products that work, rather than asking taxpayers to foot the bill. The result is that we move much faster and at lower cost than most traditional [large defense companies],” Luckey said in a TED Talk published in April. “Unlike traditional contractors, we build, test and deploy our products in months, not years.”

The startup’s fast-moving approach comes with its share of setbacks—during closed military exercises, at private drone ranges and even on the battlefield in Ukraine.

In California, a mechanical issue damaged the engine in Anduril’s unmanned jet fighter Fury in a ground test over the summer ahead of a critical first flight for the Air Force. In August, a test involving its Anvil counterdrone system caused a 22-acre fire in Oregon. And in the exercises with unmanned boats over the summer off the coast of California, Anduril’s Lattice software struggled to command and control vessels.

Anduril’s only real battlefield experience—in Ukraine—has been marred by problems as well, including vulnerability to enemy jamming, according to former employees and others familiar with the systems in Ukraine. Some front-line soldiers of Ukraine’s SBU security service, for instance, found that their Altius loitering drones crashed and failed to hit their targets. The drones were so problematic that they stopped using them in 2024 and haven’t fielded them since, according to people familiar with the matter.

“The challenge will be, can they deliver? They have minimum viable products in a bunch of different areas,” said Bryan Clark, a former Navy strategist who is now at the Hudson Institute.

After the Journal inquired about the incidents, the company said that its methods can lead to testing failures. “We recognize that our highly iterative model of technology development—moving fast, testing constantly, failing often, refining our work, and doing it all over again—can make the job of our critics easier,” the company said in a statement. “That is a risk we accept. We do fail … a lot.”

Representatives from Anduril and those who defend the company say that it is encountering the same sorts of issues that occur in any weapons development program and that the company’s roster of engineers is making impressive strides. They say that none of the incidents suggest fundamental problems with its products and that the point of testing is to challenge the software, find the bugs and fix them.

The company said it has maintained a “near continuous” presence in Ukraine to update its software and weapons, and that its drones have proven effective against a large number of Russian assets.

Software setbacks

Anduril’s software platform, called Lattice, aims to connect various weapons systems to enable a single servicemember to control a range of drones. On its website, the company says Lattice can orchestrate “machine-to-machine tasks at scales and speeds beyond human capacity.”

The platform “lets us deploy millions of weapons without risking millions of lives. It also allows us to make updates to those weapons at the speed of code,” Luckey said in April.

But in some Navy exercises, Lattice has fallen far short of servicemembers’ expectations, according to documents, defense officials and people familiar with the matter. In some cases, operators have had to manually send commands to boats or control them remotely with a device due to the software’s shortfalls, the people said.

During the May exercise with drone boats in California, unmanned boats made by BlackSea Technologies were relying on Lattice when they began to idle in the water. The boats were rejecting commands and were unable to reliably maneuver away from other traffic, prompting a safety stand-down, according to people familiar with the exercise.

Anduril said the failure wasn’t Lattice’s fault, but rather a bug in the software on the boats, made by BlackSea Technologies. Anduril said it identified the root cause, fixed the problem and returned to the exercise a few days later to successfully complete autonomous mission plans.

BlackSea Technologies referred questions about the exercise to the Navy since its boat, the Global Autonomous Reconnaissance Craft, is an official Navy program. BlackSea also said that its boat is designed to work with a range of software and that it has successfully operated with a half-dozen other software stacks.

A Navy spokesperson didn’t comment on the companies involved, but said the exercise had multiple mitigation measures in place and didn’t create risk to force or potential for loss of life. He said the boats went “dead in the water” as an automatic fail-safe, which prevented them from causing damage or injury.

Three people familiar with the exercise said the problem was Anduril’s because it was the company’s responsibility to implement its software correctly. The authors of the preliminary report, who said Anduril had misguided the military, couldn’t be reached to comment.

Jet fighter lags

Anduril’s most high-profile, and potentially most lucrative, military project has also faced challenges.

When Anduril won a multimillion-dollar Air Force contract last year to develop and test a prototype of an unmanned jet fighter, known as a “Collaborative Combat Aircraft,” it signaled the company’s biggest shot yet at building a major weapon system for the Pentagon. The company, which has never manufactured weapons at a large scale, is building a plant in Ohio to produce the unmanned jet.

Air Force leaders initially set expectations for the aircraft to fly before the end of summer. But during a test in August, a mechanical issue caused a nail to be sucked into the aircraft’s intake, damaging the engine, according to people familiar with the matter.

Anduril didn’t publicly disclose the engine issue, and when asked about the timing of the flight during a call with reporters in September, Luckey said the delay was due to the rigors of Air Force ground testing because the Air Force had only one Anduril plane to work with at the time.

“If it was up to my engineers, we’d push the throttle and shoot into the air months ago,” Luckey said during the call.

Executives also told reporters that the company was taking time to get the software right, which they said would allow them to “leapfrog” the test plan by flying semiautonomously.

By the time the drone jet Fury took to the skies over Southern California on Oct. 31, the test flight came two months after the first flight of Anduril’s main competitor in the Air Force program, General Atomics.

Anduril said the nail that caused the engine damage was the result of a temporarily installed test instrument and had nothing to do with the structural design of the aircraft. The Air Force said Anduril and General Atomics were ahead of the program’s schedule, which required the companies to conduct a test flight of their aircraft by the end of the year.

“Both Anduril and General Atomics are in the very early developmental stages of what promises great opportunity, but there’s a long way to go to realize that opportunity,” said David Deptula, a retired Air Force lieutenant general who is now dean of the Mitchell Institute for Aerospace Studies, an aerospace think tank.

Trial by fire

During an August drone intercept test in Oregon, Anduril’s Anvil counterdrone system crashed and caused a 22-acre fire near the Pendleton Airport, according to an incident report obtained by the Journal through a Freedom of Information Act request.

The report said Anduril had tried to put out the flames with its own vehicle, but three trucks from the local fire department had to be brought in to extinguish the fire. Anduril said it has since developed a mitigation plan for impact and intercept testing at the range to mitigate or avoid any fires in the future.

Photographs show a scorched and mangled drone, and satellite images reveal the extent of the damage to the terrain.

Major defense companies that have been testing for decades typically understand their boundaries and have adequate mitigation measures on site, analysts say. “Anduril is less prepared institutionally to do this, so they are finding their way around,” said Jonathan Wong, a senior policy researcher at Rand, referring to the fire.

The drone range declined to comment. Anduril said that the test was conducted in accordance with all range safety procedures. It said the fire was a possible known outcome and not a system failure.

“We test five days a week nearly 52 weeks out of the year at several of our test sites across the US,” the company said. “We expect things like this to occur once in a while due to the volume of tests.”

It has never been easier to buy stuff. But dealing with product and service problems has never felt so difficult, consumers say.

Seventy-seven percent of customers reported a product or service problem in the past year, a new high compared with 74% in 2023, according to the National Customer Rage Survey.

Sixty-eight percent of consumers found complaining to companies required high or very high effort, up from 65% in 2023.

Consumer perceptions of customer experience have declined for four consecutive years, with an average score of 68.3 out of 100 in June, according to Forrester research.

Guests were sent scrambling on Nov. 9 when the hotel-and-apartment rental company Sonder abruptly shut down during their stays. Their keycards suddenly stopped working, locking some customers out with their belongings still inside. And when they sought help from Sonder workers and loyalty-program representatives at Marriott, which had been offering Sonder rooms under a licensing deal, they discovered that they were sometimes breaking the news.

It was just another day for U.S. consumers.

Seventy-seven percent of customers reported experiencing a product or service problem in the previous 12 months, according to the latest National Customer Rage Survey, conducted in February.

That was a new high, surpassing 74% in 2023, when the study was last conducted, and 66% during the height of the pandemic in 2020. Only 32% told researchers they had experienced a problem in 1976, when a similar version of the study was first conducted.

Other studies have reached similar conclusions.

The Customer Experience and Communications Consumer Insights survey showed seventy-one percent of U.S. and Canadian respondents think most companies need to improve their customer experience, a record high, according to the seventh annual study conducted in August and published in November by fintech company Broadridge Financial Solutions.

And the research and advisory firm Forrester in June found that U.S. and Canadian consumer perceptions of the customer experience have dropped for a fourth consecutive year, with brands’ average score reaching a record low of 68.3 out of 100. The index reflects consumers’ attitudes across six metrics, including how easy a brand is to deal with and how interacting with the brand feels.

People are partly upset over their feeling that it is much easier to buy products and services than it is to get help when there is a problem, according to Scott Broetzmann, the president and chief executive of Customer Care Measurement & Consulting, which conducts the National Customer Rage Survey with the W.P. Carey School of Business at Arizona State University.

Sixty-eight percent said their recent experiences with complaining to companies required high or very high amounts of effort, up from 65% in 2023, the study found. Respondents’ top two frustrations were long messages to endure before speaking with a representative and trying to figure out how to contact a company.

“I can order dental floss and it’ll be at my doorstep in 30 minutes, but when it comes to problem-handling, it’s still an effortful, frustrating, emotionally stressful” experience, Broetzmann said.

That dynamic extends to the subscription models proliferating everywhere. The Federal Trade Commission under the Biden administration tried to force companies to make it as easy to cancel memberships as it is to sign up. But a federal appeals court nullified the so-called click-to-cancel rule, ruling that the FTC didn’t follow proper procedures during the rule-making process.

Retailers from Saks Fifth Avenue to Abercrombie & Fitch have also been fighting a rising tide of returns with measures like return fees and shorter return windows, annoying some customers.

Many companies now are turning to artificial intelligence to field complaints, steering customers to online chatbots before they can reach human staff. Companies say the technology helps solve simple problems faster and lets representatives spend more time working on more complex issues. But most National Rage Survey respondents gave AI chatbots ambiguous or modestly unfavorable ratings as tools for complaints.

It also found that consumers who identified as upper class are nearly twice as satisfied with resolution outcomes compared with those in the middle, working, and lower classes. More companies in recent years have taken to stratifying their customer service, providing those who pay for higher-tier memberships with special phone lines and direct access to customer support.

Some companies do offer service or an experience that seems to resonate with consumers in a good way. Chewy, a pet supplies retailer, topped Forrester’s Customer Experience Index for a fourth year in a row, with a score of 80.3 out of 100.

But the overall trend is downward. Seven percent of the 469 brands Forrester studied improved their score by a statistically significant amount from 2024 in the U.S. and Canada, compared with 25% whose scores fell.

“Like the proverbial frog that doesn’t feel the water becoming increasingly hotter, many North American brands are inching into more treacherous positions with their customers’ loyalty,” Forrester Principal Analyst Pete Jacques wrote in the report.

While you’ve been sweating the details over Thanksgiving, famed investor Michael Burry – the one portrayed by Christian Bale played in “The Big Short” – has been waging an increasingly aggressive war against Nvidia.

It’s a battle worth watching because Burry might actually win it. What makes this different from every other warning about an AI bubble is that Burry now has the audience and the freedom from regulatory constraints to potentially become the catalyst for the very collapse he’s predicting. He’s betting against the AI boom, but he’s also proactively trying to convince his growing number of followers that the emperor – Nvidia – has no clothes.

What everyone is now wondering is whether Burry can create enough doubt to truly hobble Nvidia and, by association, the other main characters in this story, including OpenAI.

Burry has really thrown himself into the effort in recent weeks. He’s been slinging mud at Nvidia; he also traded nasty comments with Palantir CEO Alex Karp after regulatory filings revealed Burry held bearish put options on both companies – a bet worth over $1 billion that they’d crash. (Karp went on CNBC and called Burry’s strategy “batshit crazy,” to which Burry responded by mocking Karp for not understanding how to read an SEC filing.) The spat encapsulates the market’s central divide: is AI going to transform everything and thus worth every billion invested, or are we now in mania territory that’s destined to end badly?

Burry’s allegations are specific and damning. He says Nvidia’s stock-based compensation has cost shareholders $112.5 billion, essentially “reducing owner’s earnings by 50%.” He has suggested that AI companies are cooking their books by slow-walking depreciation on equipment that’s losing value fast. (Burry believes that Nvidia customers are overstating the useful lives of Nvidia’s GPUs in order to justify runaway capital expenditures.) As for all that customer demand, Burry has basically proposed it’s a mirage because AI customers are “funded by their dealers” in a circular financing scheme.

Enough people have begun citing Burry that Nvidia, despite all its muscle and might and blowout earnings report last week, felt compelled to respond recently. In a seven-page memo sent to Wall Street analysts last weekend by Nvidia’s investor relations team – a development first reported by Barron’s – the company fired back, saying that Burry’s math is wrong, including because he “incorrectly included RSU taxes” (the real buyback figure is $91 billion, not $112.5 billion, the memo says). Nvidia’s employee compensation is also “consistent with peers.” And Nvidia is definitely, absolutely, not Enron, thank you very much.

Burry’s response, in a nutshell: I didn’t compare Nvidia to Enron. I’m comparing Nvidia to Cisco circa the late 1990s, when it overbuilt infrastructure that nobody actually needed at the time and its stock cratered 75% when everyone realized as much.

This could all look like a tempest in a teapot by Thanksgiving next year. Or not.

Nvidia’s stock has gone up twelvefold since early 2023. The company’s market cap at this moment is $4.5 trillion. Its ascent to becoming the world’s most valuable company is faster than anything the market has seen previously.

But Burry has a track record that’s complicated. He called the housing crisis, which brought him great acclaim. But since 2008, he has been predicting various apocalypses pretty much constantly, earning him the label “permabear” from critics, while people who listen to him with a kind of cult-like devotion have missed some of the greatest bull runs in market history. Burry smartly bought GameStop early, for example, but he then sold his shares before the meme stock explosion. He shorted Tesla and lost a fortune. After his smart housing crisis call, frustrated investors actually fled his fund because of extended underperformance.

Earlier this month, Burry deregistered his investment firm, Scion Asset Management, with the SEC. He said it was because of “regulatory and compliance restrictions that effectively muzzled my ability to communicate,” explaining that he was frustrated, watching people misinterpret his tweets on X.

Last weekend, he launched a Substack called “Cassandra Unchained” that he’s now using to prosecute his case against the entire AI industrial complex. The descriptor for the newsletter, a yearly subscription to which costs $400, is that it is now Burry’s “sole focus as he gives you a front row seat to his analytical efforts and projections for stocks, markets, and bubbles, often with an eye to history and its remarkably timeless patterns.”

People are definitely listening. The newsletter launched less than a week ago, and it already has 90,000 subscribers. Which brings us again to the truly unsettling question hanging over all of this: Is Burry the canary in the coal mine, warning of a collapse that’s inevitable, or could his fame, his track record, his now unrestricted voice, and a fast-growing audience trigger the very implosion he’s predicting?

History suggests this isn’t so crazy. Jim Chanos, the famous short seller, didn’t create Enron’s accounting fraud, but his high-profile criticisms in 2000 and 2001 gave other investors permission to question the company and accelerated its unraveling. Prominent hedge fund manager David Einhorn’s detailed takedown of Lehman Brothers’ accounting tricks at a 2008 conference made other investors more skeptical and may have hastened the loss of confidence that led to collapse. In both cases, the underlying problems were real, but a credible critic with a platform created a crisis of confidence that became self-fulfilling.

If enough investors believe Burry about AI overbuilding, they will sell. The selling will validate his bearish thesis. More investors will sell. Burry doesn’t need to be right about every detail – he just needs to be persuasive enough to trigger the stampede. Looking at Nvidia’s November performance, it’s easy to conclude Burry’s warnings are taking hold; seeing its shares’ performance over the entire year, it’s less obvious that’s the case.

Much clearer is that Nvidia has everything to lose, including an almost mind-blowingly massive market cap and its position as the most indispensable company of the AI age. Meanwhile, Burry has nothing to lose but his reputation and a new megaphone that he’ll presumably be using at full volume for the foreseeable future.

The investment bank has been hired to sell up to 51.2% of the share capital of the parent company of the Sandro, Maje, Claudie Pierlot and Fursac fashion chains

PARIS – Three key shareholders of French fashion group SMCP have hired investment bank Lazard Frères to sell up to 51.2 percent of its share capital.

The parent company of the Sandro, Maje, Claudie Pierlot and Fursac fashion chains said it was informed on Thursday that the sale process has been initiated.

“SMCP welcomes this potential sale, which would enable the group to stabilize its shareholder structure and focus on pursuing its development strategy,” the company said in a brief press release. “This sale process is expected to last several months.”

The shares up for sale include:

• the 28 percent held by GLAS, trustee for the bonds issued in 2018 by European TopSoho S.à.r.l.
• the 15.5 percent stake now held and managed by ETS, previously at the center of a “missing shares” saga
• a 7.7 percent stake held by ETS, which is being managed by Alastair Beveridge and Daniel Imison from Alix Partners LLP, appointed as receivers by GLAS

“If the stake acquired in this context were to exceed 30 percent of the company’s share capital, the purchaser of this stake – acting alone or in concert – may be required to initiate a public tender offer for all SMCP shares,” the release said. “At this stage, however, there is no certainty that the process will be successful, and the final decision on the sale remains with the holders of the aforementioned shareholdings.”

The release noted that the entity ETS is subject to bankruptcy proceedings in Luxembourg and represented by a curator.

Last August, SMCP regained control of the 15.5 percent stake as a result of a ruling by the Singapore High Court, following a lengthy legal battle.

The stake has been at the center of a years-long shareholder dispute dating back to the financial collapse of Chinese conglomerate Shandong Ruyi, which had acquired a majority stake in SMCP in 2016 through European TopSoho ahead of the group’s 2017 initial public offering.

On Monday, the President signed an executive order laying out his vision for a “Manhattan Project” for AI. The plan calls for the national labs—the Department of Energy-funded research centers—to work together with AI companies to train AI models based on government science data. For instance, it tasks the Secretary of Energy with assessing the computing resources that “industry partners” could provide to the effort.

That collaboration, along with others in the order, would build on existing relationships between AI companies and their counterparts in universities and national labs.

For example, earlier this year, system administrators from xAI sought out advice on how to set up data centers from Dr. Dan Stanzione, who runs the Texas Advanced Computing Cluster at the University of Texas, Austin. They talked about “high speed file systems to support AI workloads,” Stanzione said, meaning how multiple servers can efficiently read the same files at the same time.

According to Stanzione, those xAI employees worked on the Memphis data center that the company established last September and used to train Grok 4, a model that caught xAI up to the frontier of AI companies, at least at the time.

“Everybody who runs big systems, we all talk, because we all have the same problems,” he said. TACC’s former employees work at Amazon, Nvidia, Google and Microsoft, so they tend to discuss best practices with TACC.

The executive order also calls for internships to train students to perform AI research for science. That provision also builds on existing government programs to equip the next generation of workers with AI experience. For example, TACC has been providing computing capacity to the National AI Research Resource, a government pilot program to support AI researchers and small companies.

“We need a lot of AI literate workers,” said Stanzione, who was on the task force that helped set up the program. Unless you get a job inside one of the leading AI companies, “where do you learn those skills?” he asked. “How do you get good at large language models?” Student researchers use TACC for AI projects such as fine-tuning language models to excel at chemistry, he said.

NAIRR has not yet received permanent funding, but Trump’s AI action plan gave a shoutout to the project, as did the House Bipartisan AI Task Force in its report last year. A bill to make NAIRR permanent has been introduced in the House.

Expanding the pipeline of AI researchers is surely a boon to AI startups, but a more immediate benefit to startups would be getting access to the scientific AI models that result from the mission. That could free those startups from training models themselves, saving them a fortune on Nvidia chips.

Academic data centers have managed to get a hold of a surprising number of those chips. TACC, for instance, will install about 4,000 of Nvidia’s latest generation of AI chips, the Blackwell, in its next facility. TACC negotiated a price for the chips in 2020, before ChatGPT, and “Nvidia has stuck by the deal, to their credit,” Stanzione said.

The total price of the computing capacity for the new data center, including the Blackwells and thousands of Vera chips, along with the networking equipment and cooling for them, came out to $150 million, Stanzione said. With current Nvidia prices, he estimated the Blackwells alone would likely cost a couple hundred million dollars.

Lessons from the Bank of England’s QE programme

Letter writing, accountability and a postmortem of QE

A recent blog post by David Aikman, director of the National Institute of Economic and Social Research, asks whether the current system for holding the Bank of England’s accountable is good enough. Letters explaining inflation misses have lost their bite. And the Bank falls behind international best practice in conducting periodic independent reviews, something Dr Aikman calls “formal postmortems”.

The 2023 Bernanke Review was a rare example of an independent, well-resourced, critique. It could — and should — be the first of many. Because the Bank’s quantitative easing programme clearly demands such treatment. What might a postmortem teach us? From our dual perspectives, eight lessons jump out.

With certain gilts showing losses of 70 per cent or more to the Bank, probably the most important lesson is not to take more interest rate risk than you need. Yes, the gilt curve is naturally longer dated. But buying longer dated gilts is a choice. Certain other central banks such as the Reserve Bank of Australia confined their QE to particular maturities, partly because of that risk.

Explicitly not taking into account financial risk is not a good look for an organisation charged with prudential risk management.

A good reason to avoid long-dated bonds comes from the unusually short-term nature of the household borrowings in the UK (think 2-5 year fixed rate mortgages rather than 30 year in the US).

In the UK, you will get more QE bang-per-buck from the front end.

Targeting QE in this way implies taking a greater interest in the shape of the yield curve. The BoE should not be troubled by this.

The Federal Reserve provides an example. During Operation Twist it tackled long dated yields. Later, during the pandemic, it focused disproportionality on short dated securities.

By contrast, the BoE’s “equal focus along the curve” meant continuing to buy ultra long dated gilts in 2021 / 2022 — after that part of the curve had inverted. When long dated gilts yielded less than other parts of the curve, a nimbler approach could have refocused QE, both to improve policy effectiveness and lower financial risk.

(4)   There are things to buy other than nominal government bonds

The BoE QE programme was notable for being more concentrated into conventional government bonds with fewer other assets. In contrast to the Fed, it held no inflation protected bonds. And its credit programme was far smaller, with only 2% of its balance sheet in corporate bonds, and less flexible. So the BoE largely passed up the opportunity to help bring down spreads or to hedge an element of the inflation risk. Again, these choices could be reviewed.

If active QT is such a good idea, perhaps ponder why few central banks are doing it. Consider slowing the aggregate pace (active plus passive) to match the peer group.

Investors had long warned the Bank about the impact from a gilt sales programme. Back in September 2022, before active QT had begun, a BoE survey showed investors estimated the BoE’s plans to sell gilts had already moved long-dated gilt yields up by 15 bps, and would go up a further 10 bps (ie totalling 25 bps) as those plans were implemented. 

That was the last time the Bank formally surveyed investors for their opinion.

A year later, the Bank was claiming a smaller impact of “less than 10 basis points”. It did not acknowledge that September 2022 investor feedback in their formal write-up and only quoted from earlier surveys taken at a time when the full scale of QT was unknown.

A review could look at the degree to which the Bank should incorporate investor perspectives.

When high profile research, led by an ex-MPC member, estimates that QT is moving gilt yields by up to 70bps, 3-4x more than the BoE’s own view, avoid claiming this is “broadly consistent” with the Bank’s own analysis.

Don’t risk another rebuke from the House of Lords for a biased assessment. Don’t lean too heavily on US-focused research — where no active QT is taking place. Instead, look around for UK focused research. For example, our own separate academic and investor-based assessments puts the BoE QT impact on yields at around 35-40bps.

If the BoE view is incomplete, reconsider and adjust course.

The QE announced in November 2020 locked the Bank into bond purchases until December 2021. Yet by May 2021, CPI inflation was above target. By August 2021 CPI was 3.2 per cent. Still the QE programme rolled on. Predictability is all well and good, but it can stray into inflexibility.

More recently, in April, the BoE commendably cancelled some gilt sales due to market pressure. When the postponed auctions resumed in July, prevailing yield levels were almost identical. Again, this could be seen as dogma, rather than market savvy.

So, avoid the “high bar to change course” language. Do not overcommit. Give yourself optionality to review as you go. Today, an active sales programme totalling some £127bn deserves to be looked at in detail more than once a year, in the September MPC meetings.

QE was not plain sailing, especially for the BoE. In the UK mark-to-market losses are over 5 per cent of GDP — around 4x greater than comparable costs in the US. Today, active QT leaves the Bank accused of placing unnecessary strain on Britain’s government debt. These issues are not addressed in the Bank’s latest attempt to scrutinise QE. There is little in the way of lessons learned or admissions of mistakes.

No QE/QT programme will be perfect. And likewise, some feedback will resonate more than others. But let us try to learn from the past, not gloss over it.