from the of-course-they-did dept

It took longer than I thought it would, but the Trump administration’s appeal of the court ruling and injunction that put a pause on RFK Jr.’s remaking of the Advisory Committee on Immunization Practices (ACIP) and vaccine policy has come through.

If you need a reminder on how we got here, here you go. In June of ’25, Kennedy fired every single member of the CDC’s ACIP panel, a group of advisers that recommends vaccine schedules for the country. He then appointed what were eventually 13 new members to ACIP, nearly all of them virulent anti-vaxxers or otherwise aligned with Kennedy’s misinformed views on medicine and science. The American Academy of Pediatrics (AAP) sued earlier this year, arguing that Kennedy had violated the American Procedures Act (APA) by his actions, specifically because he did not follow evidence, proper procedure, or factual science in the appointments. The court agreed, ruling against the administration and issuing a preliminary injunction on HHS for staffing ACIP with the new appointees and nixing any of the recommendations it had made thus far.

And so now the administration has appealed that ruling, though it’s any wonder as to what the administration’s arguments will be for the appeal.

A filing Wednesday evening in the District of Massachusetts indicates the administration is appealing Judge Brian Murphy’s order March 16. Murphy put any decisions made by the Center for Disease Control and Prevention’s vaccine advisory committee on hold, ruling that Kennedy replaced the committee “unlawfully.”

Assistant Attorney General Brett Shumate signed the appeal.

Advertisement

The Justice Department could file a motion for emergency relief to get the court to act on its appeal immediately. That would require the 1st U.S. Circuit Court of Appeals to act quickly in deciding whether to stay, or pause, the March 16 ruling.

Regardless, the court activity on this will likely eventually take months to work out. The AAP isn’t backing down, with its attorney vowing to respond to the appeal and taking the posture that it believes they will prevail. Reading the APA statute, I very much tend to agree, but this is Trump 101 stuff. Never back down, exhaust every legal avenue to get your way, and hope someone along the way fucks this up so you get your way. That the final leg on this journey might be a Supreme Court that often looks like another part of the Executive Branch, rather than an independent arm of government is certainly part of the calculus.

But in the meantime, two things remain true. The Trump administration, which has at times made noises about wanting to rein in Kennedy and his nonsense, is working hard to allow him to continue to make America sicker. And because of Kennedy’s refusal to follow basic protocol and science, the country is without a competent body to advise on vaccination schedules.

Meanwhile, the status of the advisory panel — a key group meant to be composed of vaccine experts independent of government influence — is in limbo.

A meeting that had been scheduled for March at which members were expected to discuss Covid shots has been postponed indefinitely. The committee is supposed to meet again in late June. There is no agenda yet.

Advertisement

Given the current makeup of the panel, that may indeed be the best possible outcome currently. But it’s not a long term plan, nor a long term positive. ACIP existed for a reason. The country needs intelligent, sincere, and sane people advising the country on how to combat infectious diseases with medicine and technology.

We are without that right now, purely because Trump thought putting Kennedy in charge of HHS was anything other than a form of national self-harm.

Filed Under: acip, anti-vaxxers, cdc, doj, donald trump, rfk jr., vaccines

Strange Inventions set out to create a clock that would stand out from the crowd. Water becomes the focal point of the show, with the numerals formed by little bottles filled to various levels. The display consists of four grids arranged side by side. Each grid contains 15 tiny bottles placed neatly in a 3×5 design. Some bottles contain blue-dyed water, while others are empty. The arrangement of full and empty bottles produces a clear image of the current hour and minute.

Each of the 60 bottles in the display is attached to one small membrane pump. When a “pixel” has to light up, its pump must draw water from the basin at the bottom and transfer it directly into the bottle. Water only flows in one direction. Servos are responsible for the opposite end of things. Each number grid contains a frame that the servos can swing forwards on. When it’s time to update the display, the servos tip all of the bottles at once, causing all of the water to gush out and fall back into the basin, where it may be reused. At this stage, the pumps only work for the bottles that need to be changed.

Sale

Amazon Echo Show 5 (newest model), Smart display, Designed for Alexa+, 2x the bass and clearer sound…

• Alexa can show you more – Echo Show 5 includes a 5.5” display so you can see news and weather at a glance, make video calls, view compatible…
• Small size, bigger sound – Stream your favorite music, shows, podcasts, and more from providers like Amazon Music, Spotify, and Prime Video—now…
• Keep your home comfortable – Control compatible smart devices like lights and thermostats, even while you’re away.

The majority of the pixels are always filled, so when the minute passes, the system just skips them and does not perform a full update. However, if a large number of pixels need to be switched, updating the display can take several minutes. The inventor began with a different plan. A single peristaltic pump would force water through tubes and solenoid valves into any bottles that need it. The same pump would then reverse direction, pulling the water back out. A shared loop of tubes would tie everything together.

However, things started going wrong almost immediately. As soon as the water started flowing, air bubbles formed inside the tubes, and pressure varied dramatically from bottle to bottle. Some filled up quickly, while others took a long time. This all came crashing down when it ruined the display’s clean appearance. You can imagine how aggravating it was to keep running into the same problems over and over. It needed to be totally redesigned, so individual pumps for each bottle were utilized rather than a single common pump. The servos were also brought in to manage the bottle-emptying. Water now only goes one way via the pumps and then down into the basin via gravity.

The tubes running through the center of each bottle help prevent snagging when the servos tip the bottles. The servos themselves have a little more oomph thanks to some springs that assist them push through any impediment. A little sanding and oil assist keep things working smoothly, allowing the grids to swing cleanly each time. It took three months between the first prototype and the finished clock. Early iterations taught the team an important lesson about how messy water can be when working with electronics. Finally, the clock simply recycles the same blue water repeatedly, while a tiny brain (microcontroller) ensures that the pumps and servos function smoothly and on time.
[Source]

The iPhone 18 Pro will not be getting a dual-layer OLED display like the one in the iPad Pro, with overheating the main obstacle to introduction.

The iPhone uses OLED for its display tech, and there have been rumors about Apple making a tweak to improve the brightness. A regular leaker with a mixed history insists it isn’t happening soon.

According to a Monday post from Weibo leaker Instant Digital, the iPhone 18 Pro won’t have dual-layer OLED for the display.

The dual-layer OLED refers to what is better known in the Apple catalog as Tandem OLED, which is used by the iPad Pro. In essence, instead of using one OLED panel, the system uses two, with one layered upon the other.

As OLED is a self-illuminating display tech that doesn’t require backlighting, stacking the panels increases the amount of emitted light while minimizing wastage. For consumers, this means a much brighter display.

Dual-layer OLED as a throttling savior

Tandem OLED can also be a benefit for the iPhone by making it more capable outdoors. Instant Digital’s post discusses how the iPhone 17 Pro didn’t do a great job of maintaining brightness outdoors.

Part of the problem is thermal management, as the display can help the iPhone heat up and eventually throttle under hot conditions.

To the leaker, a dual-layer OLED approach would be beneficial since it would give a way to have a much brighter display, while generating less heat. Achieving this would make the iPhone better to use in sunlight, without fear of throttling.

As for the iPhone 18 Pro, Apple is still in the process of determining the suppliers of display panels. Usual suppliers Samsung Display and LG Display are the frontrunners, but BOE isn’t doing well enough to supply for the premium models.

While Weibo leakers are not usually considered to be highly accurate, with Instant Digital among the most prominent, we cannot dismiss the rumor as just being an obvious prediction outright. Others have discussed the idea in the past, with the timetable lining up with this rumor.

Back in August 2025, there were claims that LG Display was working on a tandem OLED technology for use in a future iPhone. At the time, a source said that the display could arrive in 2028, far too late for the iPhone 18 Pro.

There have been some incidents involving a Mac telling a user that the ChatGPT app is malware and moving it to the trash. Overall, ChatGPT isn’t malware, and there’s a very easy fix.

This is not Apple making a judgement on the value of the ChatGPT app on the Mac, it’s macOS doing its job. Since 2022, macOS has included Xprotect, a feature you never usually need to know about, but which safeguards the Mac against malware.

In this case, as reported by users worldwide on social media, Xprotect had concluded that the ChatGPT app contained malware. Therefore the whole app is suspect, therefore macOS moves it to the Trash and won’t launch it.

There is nothing wrong with ChatGPT and it will not have installed malware on users’ Macs. To continue using it, the simplest way is to reinstall ChatGPT directly from the developer.

Apple’s Xprotect warning was legit, though, because of how it decides apps are legitimate, and how OpenAI changed that proof for both ChatGPT and ChatGPT Atlas. Apps are notarized through a certificate that shows they are legitimate, but OpenAI switched to a new certificate because of security concerns.

“We recently identified a security issue involving a third-party developer tool, Axios, that was part of a widely reported, broader industry incident,” wrote OpenAI in a blog post. “Out of an abundance of caution we are taking steps to protect the process that certifies our macOS applications are legitimate OpenAI apps.”

The company stresses that it found no evidence of the apps being altered, or user data accessed. It did also notifications via its apps that users should update urgently by May 8, 2026.

But for those users who did not, their copies of ChatGPT and ChatGPT Atlas are no longer notarized so the Mac will not run them. It’s inconvenient having to redownload the apps, but if OpenAI had not done this, bad actors could conceivably have created what would entirely appear to be legitimate ChatGPT apps.

Then, too, if Xprotect didn’t work the way it does, there would still be no way to prevent such malware from actually being installed on Macs.

In addition to being a very popular first-person shooter game, Counter-Strike 2 is a great demonstration of the finest economic systems of games created so far. In particular, such an economic system is represented by virtual trading of items worth more than $8 billion. Indeed, one should bear in mind that this is not a typo – this figure really represents the cost of items worth $8 billion. It may be added that this amount is larger than GDP in many countries, despite not having any effect on the gameplay. So, how did some in-game items attain such a multi-billion dollar economy and what fuels it? We will explain. 

How a Digital Skin Gets Its Price Tag

Every skin in CS2 has a set of properties that determine its value, and understanding them is the first step to making sense of this economy.

Rarity tier is the most obvious one. Skins are categorized from Consumer Grade (white, the most common) all the way up to Covert (red, the rarest non-knife items) and Contraband (the ultra-rare category with only one item — the M4A4 Howl). Knives and gloves sit in their own Extraordinary tier, which is part of why they command such premium prices.

Then there’s float value — a number between 0.00 and 1.00 that determines a skin’s visual condition. A float of 0.01 means the skin looks virtually brand new (Factory New), while 0.85 means it’s scratched up and Battle-Scarred. Two AK-47 Redlines might look similar at a glance, but a 0.01 float Factory New will sell for significantly more than a 0.15 Minimal Wear.

And finally, there are pattern-based factors. Certain skins like Case Hardened and Fade have pattern indexes that produce unique visual results. A Case Hardened AK-47 with a full blue gem pattern can sell for tens of thousands of dollars, while the same skin with a standard pattern might go for $40. Doppler knives have distinct phases, each with its own pricing tier. Even sticker placements matter — a skin with rare Katowice 2014 stickers in the right positions can multiply the base price several times over.

The Marketplace Ecosystem

Here’s where things get interesting from a tech perspective. Unlike most games where you buy skins from a single in-game store, CS2 has an entire ecosystem of competing marketplaces.

Steam Community Market is Valve’s own platform and the default option for most players. It’s integrated directly into the Steam client, making it convenient, but it comes with a 15% transaction fee and locks your earnings in Steam Wallet — you can’t cash out to real money.

It resulted in the creation of an extensive array of third-party marketplaces, which include websites like Skinport, DMarket, CSFloat, Buff163, and countless other options. These websites allow users to exchange skins for real money, using various means of payment, such as PayPal payments, bank transfers, and cryptocurrency transactions. The fee structures on these websites vary considerably, ranging from zero percent up to 10 percent and beyond. 

This price fragmentation is exactly why analytics and comparison tools have become essential for anyone who takes CS2 trading seriously. Experienced traders routinely check CS2 prices across multiple platforms before making a move, because the price gap between the cheapest listing and the most expensive one for the same skin can easily be 15-30%.

Market Cap Tracking — Like Crypto, But For Skins

One of the more fascinating developments in the CS2 economy has been the adoption of financial tracking concepts borrowed from traditional and crypto markets.

The total CS2 market capitalization — the combined estimated value of every tradeable item in the ecosystem — is tracked in real time, much like how CoinMarketCap tracks cryptocurrency values.At the end of 2025, the peak market capitalization of CS2 was more than $6 billion; however, the market capitalization dropped by roughly 30% in a single move when Valve made an update (to be discussed later).

Such advanced monitoring is essential for the user to see whether the general market is expanding or contracting. If there is an increase in the market cap, then demand and investments are likely increasing; otherwise, a sharp drop may indicate a Valve update, season, or a major event in the global gaming economy.

The data-driven platform collects information from over 20 marketplaces and provides dashboards that contain trend analysis, volumes, and price movements that could have been taken directly from a professional stock trading platform. The economy of CS2 has reached such a degree of development that the very concept of “gaming” becomes irrelevant. 

Trade-Up Contracts: The Economy’s Built-In Upgrade Path

Valve didn’t just build a marketplace — they built game mechanics directly into the economic system. However, the most crucial part is the Trade-Up Contracts where the user gets a skin from the next level collection using ten skins from the current level collection.

Even though this concept seems quite simple, it requires rather complex mathematical calculations. Namely, the output skin’s type is dependent on the input collections’ types, whereas its float is calculated according to the average float of all input skins scaled to the output collection’s range. Thus, an advanced player may affect the probability of getting a certain skin via inputs manipulation.

To explain, if seven skins belong to one collection while three skins are from another, the output skin will most probably originate from the first collection. If the first collection contains a $500 skin at the next tier and the second contains a $30 skin, you can engineer a heavily weighted gamble in your favor.

But here’s the catch — the math only works if you actually run the numbers. The inputs might cost $80 in total, but if the expected value of the output is only $60, you’re making a bad bet regardless of the potential upside. That’s why experienced traders simulate their contracts using a CS2 trade-up calculator before committing any skins. These tools predict every possible outcome with exact probabilities, float projections, and expected profit or loss.

The trade-up system was further shaken in October 2025 when Valve added the ability to trade up Covert skins into knives and gloves — something that was previously impossible. Players could suddenly turn five Covert skins worth roughly $5-10 each into knives that were previously selling for $1,000+. The result? Knife prices crashed overnight, the total market cap dropped by hundreds of millions, and the entire pricing hierarchy had to readjust.

The Tech Infrastructure Behind It All

All that lies beneath all these graphs and calculations is quite a bit of technology. Real-time data feeds, APIs, and aggregators pull pricing information from several different marketplaces simultaneously. 

Automated trading, monitoring services, portfolio management tools, and other such applications are developed by third parties using the marketplace APIs. Some platforms offer their own developer APIs with endpoints for price recommendations, market analytics, and cross-platform price comparison — essentially creating the financial infrastructure layer that the CS2 economy needed to operate at scale.

Steam itself provides API access for inventory data, market listings, and transaction history, which third-party services use to power everything from inventory valuation tools to automated trading systems.

The sophistication has reached a point where the CS2 economy has its own version of Bloomberg terminals — dashboards that track market-wide trends, individual item price histories, trading volumes, liquidity scores, and even volatility metrics. Professional traders monitor these tools the same way a Wall Street analyst watches stock tickers.

Why It Matters Beyond Gaming

The CS2 skin economy isn’t just a curiosity — it’s a case study in how digital ownership, market dynamics, and community-driven value creation work at scale.

This is what some of the main points which can be derived from this are. Firstly, scarcity defines value in all instances. It has been illustrated in the CS2 skins case study, in which it is clear that it does not matter whether items are tangible or useful in order for them to have economic value. 

Second, platform decisions have outsized economic impact. Valve’s single update in October 2025 erased over a billion dollars in virtual item value. No other company has that kind of direct influence over a player-driven economy of this scale.

And third, the line between gaming economies and financial markets is dissolving. When your hobby comes with real-time price tracking, market cap analytics, trade-up calculators, and cross-platform arbitrage opportunities, you’re not just playing a game anymore. You’re participating in a micro-economy that happens to live inside one.

Whether you’re a casual CS2 player who’s never sold a skin or a veteran trader running profit calculations on every drop, the scale and sophistication of what’s been built here is worth paying attention to. An $8 billion economy that runs on cosmetic pixels, community trust, and a few really good APIs — that’s the kind of thing you only find in gaming.

A leaked internal briefing note describes a new admin role on the £330m Federated Data Platform that lets external staff bypass case-by-case data approvals. Patient groups and Labour MPs have called the change dangerous.

NHS England has decided to allow external personnel from contractors, including Palantir, to access identifiable patient data through a new administrative role on its main data platform, the Guardian reported on Sunday, citing an internal briefing note.

The change applies to the National Data Integration Tenant, a controlled environment that NHS England describes as a “haven” for identifiable patient data before that data is pseudonymised and passed into other systems connected to the Federated Data Platform (FDP).

Under existing rules, anyone working on the platform has to apply for approval to access specific datasets, a process known as a Controlled Data Access (CDA) request.

The briefing note seen by the Guardian says NHS England is creating a new “admin” role that grants broader permissions to approved external staff in a single approval, on the basis that applying for individual CDAs had become “too inconvenient”.

Palantir Technologies won the £330m FDP contract in 2023 and is the primary external contractor on the platform. NHS England has said that anyone external requiring access under the new arrangement must hold government security clearance and be approved by an NHS England director or more senior official. The list of contractors with potential access also includes consultancy firms supporting the programme.

The Federated Data Platform is designed to pull operational data from trusts across the NHS into a single environment for planning, waiting-list management and resource allocation.

Identifiable patient information is meant to remain inside the National Data Integration Tenant, with only pseudonymised or aggregate data passed to downstream FDP modules. The new admin role applies to staff working inside the tenant itself.

Patient groups and several Labour MPs criticised the change. Labour MP Rachael Maskell told the Guardian the move was “dangerous” and urged ministers to intervene to halt the broader FDP project. medConfidential, a patient-data-rights group, said the new role represented a material shift in how identifiable data is governed inside NHS England’s largest data programme.

NHS England said the change had been internally approved by its information-governance team and that all access remained subject to existing legal and clinical-safety frameworks.

Palantir declined to comment on the specifics of the access arrangement. The company has previously said it processes NHS data only on instructions from NHS England and does not own or commercialise the underlying data.

The change reignites a long-running political dispute over the FDP contract. Critics have argued from the start of the procurement process that giving a single US-headquartered defence and intelligence contractor central access to the NHS data spine creates concentration risk and undermines public trust.

NHS England and the Department of Health have defended the contract on the basis that it improves operational efficiency and reduces clinical-safety risk by consolidating fragmented data systems.

Wes Streeting, the Health Secretary, has not made a public statement on the latest disclosure. The government’s broader position has been that the FDP is essential to NHS modernisation and that data-access processes are subject to robust controls.

The briefing note suggests the trade-off the new admin role embodies: faster operational onboarding of external staff against tighter case-by-case oversight of which datasets they access.

The Information Commissioner’s Office has not formally commented on whether the new role triggers any additional regulatory review. The Guardian’s reporting did not indicate when the role was scheduled to take effect; NHS England did not provide a date when asked.

Conspiracy theorists, wellness influencers, and grifters have already started promoting wild claims about the hantavirus outbreak that began aboard the MV Hondius, a cruise ship on the Atlantic.

Some conspiracy theorists compared the outbreak to the Covid-19 pandemic, claiming it was another effort to control the global population, while others pushed a false narrative that the Covid-19 vaccine caused hantavirus. Many others promoted ivermectin as a treatment, using the incident as a way to sell emergency medical kits featuring the antiparasitic drug typically used as a horse dewormer.

In more recent days, many of these same people spreading conspiracy theories have promoted the baseless and antisemitic claims that the entire incident is a false flag orchestrated by Israel.

Conspiracy theories flooding social media in response to breaking news are nothing new, but what is notable about those being pushed around the hantavirus outbreak is just how closely they echo the conspiracy theories promoted during the Covid-19 pandemic.

“One of the most striking shifts since the Covid pandemic is how rapidly misinformation narratives now organize themselves around emerging outbreaks,” Katrine Wallace, an epidemiologist at University of Illinois Chicago School of Public Health, tells WIRED.

“Within hours of the first hantavirus headlines, social media accounts were already promoting ivermectin, attributing the outbreak to Covid vaccines, and warning about a hantavirus vaccine that does not exist. The claims themselves were often contradictory, but that contradiction no longer appears to limit their spread.”

Once the hantavirus outbreak started making headlines around the world, conspiracy theorists and grifters jumped into action, spreading dangerously ill-informed claims and, of course, trying to sell people ivermectin.

“Ivermectin should work against it,” Mary Talley Bowden wrote on X. Bowden, a doctor, is a prominent promoter of medical misinformation who has promoted ivermectin as a treatment for Covid-19 and prescribed ivermectin to a Covid-19 patient. Hours after her first post on Hantavirus went viral, she followed up to say that she is selling ivermectin to Texans. Bowden did not respond to a request for comment.

Her post, which has been viewed 4 million times, was shared by former Congresswoman Marjorie Taylor Greene, who added that vitamin D and zinc would help fight the infection. Greene even claimed that not getting the Covid-19 vaccine had somehow allowed her to “develop natural immunity” against hantavirus.

Greene separately claimed, without evidence, that the pharmaceutical company Moderna had purposely manipulated the virus in order to allow them to cash in by developing a hantavirus vaccine. Greene did not respond to a request for comment.

Other prolific health disinformation promoters boosted the ivermectin claims, including Simone Gold, the founder of Covid denial group America’s Frontline Doctors, and Peter McCullough, a disinformation peddler who promoted the “sudden death” conspiracy theory about the Covid-19 vaccine, which falsely claimed that those who received the shot were at risk of dropping dead without any warning.

McCullough is also the chief scientific officer for The Wellness Company, which has been described as “Goop for the GOP.” The company has used the hantavirus outbreak to promote a $325 “Contagion Emergency Kit” which includes both ivermectin and hydroxychloroquine.

All the false claims and posts about ivermectin gained enough traction online that the World Health Organization responded to say that there is no research to suggest ivermectin is an effective treatment for hantavirus.

Conspiracy theorists have, meanwhile, been pushing the baseless idea that a side effect of Covid vaccines includes a hantavirus infection.

Sony is still keeping things wide open when it comes to its next-generation console.

According to the company’s latest earnings call, the PlayStation 6 still has no confirmed release date or price. Sony says it is closely monitoring rising memory costs and wider market conditions before locking anything in.

The update came during Sony’s annual corporate strategy and investor Q&A session. At this session, CEO Hiroki Totoki addressed concerns about how global supply pressures could shape the company’s next console. His message was simple: nothing is decided yet.

As shared by Eurogamer, Totoki explained that the ongoing memory shortage is driving up manufacturing costs. This is increasing the overall bill of materials (BoM) for hardware like the PS6. While Sony has already secured enough components for 2026 and has some pricing agreements in place, the outlook beyond that is far less stable.

Looking further ahead, he warns that memory prices will likely remain high into FY2027. This means Sony is still evaluating how that will impact production costs and retail pricing. That uncertainty is now directly feeding into PS6 planning.

“We have not yet decided at what timing we will launch the new console, or at what prices,” Totoki said, adding that Sony is running multiple scenarios to figure out its next move. That includes, notably, the possibility of adjusting its business model. However, the company didn’t expand on what that could actually mean for players.

For now, it’s all positioning rather than specifics. Sony isn’t confirming features, pricing tiers, or even a launch window. The company is just acknowledging that external pressures, particularly AI-driven demand for memory, are reshaping its hardware strategy.

The comments echo wider industry concerns, with companies across the tech space warning about supply bottlenecks affecting future products. Within PlayStation itself, former executives and hardware leads have also hinted that the next generation may need a different approach compared to the PS5 era.

As it stands, the PS6 remains firmly in the planning phase. And while speculation around features and timelines continues to build, Sony’s message is clear: the final shape of its next console will depend as much on the global market as it does on its own roadmap.

Waymo has issued a software update to its fleet of nearly 4,000 vehicles to help them avoid flooded roads as part of a recall announced by the National Highway Traffic Safety Administration (NHTSA) on Tuesday.

But the company hasn’t fully solved the problem of how its vehicles behave in these conditions. In documents released by NHTSA, the federal safety regulator says Waymo is still “developing the final remedy for this recall.”

The issue appears to be that Waymo’s robotaxis were slowing, but not stopping, when encountering flooded roads that they could not traverse, according to NHTSA. Robotaxis using both Waymo’s fifth- and sixth-generation autonomous vehicle systems are affected.

The regulator said the recall applies to 3,791 vehicles — giving us a more up-to-date understanding of just how many vehicles Waymo has on the roads in around a dozen U.S. cities.

Waymo has now issued multiple recalls for its self-driving cars. The company’s first recall came in February 2024 after it discovered two robotaxis in Phoenix had separately crashed into the same towed vehicle. Since then, Waymo has issued recalls to fix low-speed crashes with parking gates and telephone poles, as well as to address illegal driving in the vicinity of school buses.

Waymo decided to issue the recall in late April after its robotaxis struggled to navigate flooding in central Texas; in one incident, an empty robotaxi was swept away in San Antonio. The company has also paused operations in the city.

The initial update sent to its fleet places “restrictions at times and in locations where there is an elevated risk of encountering a flooded, higher-speed roadway,” according to NHTSA.

“We have identified an area of improvement regarding untraversable flooded lanes specific to higher-speed roadways, and have made the decision to file a voluntary software recall with NHTSA related to this scenario,” Waymo said in a statement. “We are working to implement additional software safeguards and have put mitigations in place, including refining our extreme weather operations during periods of intense rain, limiting access to areas where flash flooding might occur.”