OpenAI is disbanding its “AGI Readiness” team, which advised the company on OpenAI’s own capacity to handle increasingly powerful AI and the world’s readiness to manage that technology, according to the head of the team.
On Wednesday, Miles Brundage, senior advisor for AGI Readiness, announced his departure from the company via a Substack post. He wrote that his primary reasons were that the opportunity cost had become too high and he thought his research would be more impactful externally, that he wanted to be less biased and that he had accomplished what he set out to at OpenAI.
Brundage also wrote that, as far as how OpenAI and the world is doing on AGI readiness, “Neither OpenAI nor any other frontier lab is ready, and the world is also not ready.” Brundage plans to start his own nonprofit, or join an existing one, to focus on AI policy research and advocacy. He added that “AI is unlikely to be as safe and beneficial as possible without a concerted effort to make it so.”
Former AGI Readiness team members will be reassigned to other teams, according to the post.
“We fully support Miles’ decision to pursue his policy research outside industry and are deeply grateful for his contributions,” an OpenAI spokesperson told CNBC. “His plan to go all-in on independent research on AI policy gives him the opportunity to have an impact on a wider scale, and we are excited to learn from his work and follow its impact. We’re confident that in his new role, Miles will continue to raise the bar for the quality of policymaking in industry and government.”
In May, OpenAI decided to disband its Superalignment team, which focused on the long-term risks of AI, just one year after it announced the group, a person familiar with the situation confirmed to CNBC at the time.
News of the AGI Readiness team’s disbandment follows the OpenAI board’s potential plans to restructure the firm to a for-profit business, and after three executives — CTO Mira Murati, research chief Bob McGrew and research VP Barret Zoph — announced their departure on the same day last month.
Earlier in October, OpenAI closed its buzzy funding round at a valuation of $157 billion, including the $6.6 billion the company raised from an extensive roster of investment firms and big tech companies. It also received a $4 billion revolving line of credit, bringing its total liquidity to more than $10 billion. The company expects about $5 billion in losses on $3.7 billion in revenue this year, CNBC confirmed with a source familiar last month.
And in September, OpenAI announced that its Safety and Security Committee, which the company introduced in May as it dealt with controversy over security processes, would become an independent board oversight committee. It recently wrapped up its 90-day review evaluating OpenAI’s processes and safeguards and then made recommendations to the board, with the findings also released in a public blog post.
News of the executive departures and board changes also follows a summer of mounting safety concerns and controversies surrounding OpenAI, which along with Google, Microsoft, Meta and other companies is at the helm of a generative AI arms race — a market that is predicted to top $1 trillion in revenue within a decade — as companies in seemingly every industry rush to add AI-powered chatbots and agents to avoid being left behind by competitors.
In July, OpenAI reassigned Aleksander Madry, one of OpenAI’s top safety executives, to a job focused on AI reasoning instead, sources familiar with the situation confirmed to CNBC at the time.
Madry was OpenAI’s head of preparedness, a team that was “tasked with tracking, evaluating, forecasting, and helping protect against catastrophic risks related to frontier AI models,” according to a bio for Madry on a Princeton University AI initiative website. Madry will still work on core AI safety work in his new role, OpenAI told CNBC at the time.
The decision to reassign Madry came around the same time that Democratic senators sent a letter to OpenAI CEO Sam Altman concerning “questions about how OpenAI is addressing emerging safety concerns.”
The letter, which was viewed by CNBC, also stated, “We seek additional information from OpenAI about the steps that the company is taking to meet its public commitments on safety, how the company is internally evaluating its progress on those commitments, and on the company’s identification and mitigation of cybersecurity threats.”
Microsoft gave up its observer seat on OpenAI’s board in July, stating in a letter viewed by CNBC that it can now step aside because it’s satisfied with the construction of the startup’s board, which had been revamped since the uprising that led to the brief ouster of Altman and threatened Microsoft’s massive investment in the company.
But in June, a group of current and former OpenAI employees published an open letter describing concerns about the artificial intelligence industry’s rapid advancement despite a lack of oversight and an absence of whistleblower protections for those who wish to speak up.
“AI companies have strong financial incentives to avoid effective oversight, and we do not believe bespoke structures of corporate governance are sufficient to change this,” the employees wrote at the time.
Days after the letter was published, a source familiar to the mater confirmed to CNBC that the Federal Trade Commission and the Department of Justice were set to open antitrust investigations into OpenAI, Microsoft and Nvidia, focusing on the companies’ conduct.
FTC Chair Lina Khan has described her agency’s action as a “market inquiry into the investments and partnerships being formed between AI developers and major cloud service providers.”
The current and former employees wrote in the June letter that AI companies have “substantial non-public information” about what their technology can do, the extent of the safety measures they’ve put in place and the risk levels that technology has for different types of harm.
“We also understand the serious risks posed by these technologies,” they wrote, adding the companies “currently have only weak obligations to share some of this information with governments, and none with civil society. We do not think they can all be relied upon to share it voluntarily.”
OpenAI’s Superalignment team, announced last year and disbanded in May, had focused on “scientific and technical breakthroughs to steer and control AI systems much smarter than us.” At the time, OpenAI said it would commit 20% of its computing power to the initiative over four years.
The team was disbanded after its leaders, OpenAI co-founder Ilya Sutskever and Jan Leike, announced their departures from the startup in May. Leike wrote in a post on X that OpenAI’s “safety culture and processes have taken a backseat to shiny products.”
Altman said at the time on X he was sad to see Leike leave and that OpenAI had more work to do. Soon afterward, co-founder Greg Brockman posted a statement attributed to Brockman and the CEO on X, asserting the company has “raised awareness of the risks and opportunities of AGI so that the world can better prepare for it.”
“I joined because I thought OpenAI would be the best place in the world to do this research,” Leike wrote on X at the time. “However, I have been disagreeing with OpenAI leadership about the company’s core priorities for quite some time, until we finally reached a breaking point.”
Leike wrote that he believes much more of the company’s bandwidth should be focused on security, monitoring, preparedness, safety and societal impact.
“These problems are quite hard to get right, and I am concerned we aren’t on a trajectory to get there,” he wrote at the time. “Over the past few months my team has been sailing against the wind. Sometimes we were struggling for [computing resources] and it was getting harder and harder to get this crucial research done.”
Leike added that OpenAI must become a “safety-first AGI company.”
“Building smarter-than-human machines is an inherently dangerous endeavor,” he wrote on X. “OpenAI is shouldering an enormous responsibility on behalf of all of humanity. But over the past years, safety culture and processes have taken a backseat to shiny products.”
Three years since the arrival of OpenAI‘s ChatGPT, more devices featuring generative AI technology have hit the market in time for the 2025 holiday shopping season, with many offering deals for Black Friday.
Shoppers can pick from more advanced smart glasses, smart speakers with genAI and a pendant AI friend that acts as a confidant.
These latest gizmos come from megacaps like Amazon, Alphabet and Meta and smaller players like Friend and Plaud.
Despite the arrival of this new wave of products, reviews for many of the devices are mixed, and nothing has separated itself as a clear leader of the pack.
That’s in part because much of the spending on artificial intelligence has been focused on other things.
Since ChatGPT was released in late 2022, the bulk of the tech industry has reoriented itself to prioritize building out large language models in a race to reach artificial general intelligence, or AI with the capabilities that are on par with, or surpass, humans.
Thus far, much of the development in Silicon Valley has focused on AI apps, including chatbots like Anthropic’sClaude, image generators like Google’s Nano Banana or feeds for AI-generated short-form videos like OpenAI’s Sora. All things people can access on their existing smartphones without a spiffy new gadget.
But the world of AI hardware is growing fast.
If you’re in the market for the latest AI devices, here’s what’s available to snag this holiday season.
Daniel Rausch, vice president of Alexa and Echo, announces the Echo Studio and Echo Dot Max during an Amazon event showcasing new products in New York City, U.S., September 30, 2025.
Kylie Cooper | Reuters
Alexa+ Echo speakers
Amazon wants to make sure its Alexa voice assistant and Echo smart speakers don’t get left behind in the era of genAI.
The company unveiled Alexa+ in February, promising a smarter, more conversational and personalized version of its 11-year-old digital assistant. In September, it followed up with a new set of Echo speakers and displays, which are the first devices to come with Alexa+ out of the box.
The lineup includes a $100 Echo Dot Max, $180 Echo Show, $220 Echo Studio and $220 Echo Show 11.
The Echo Dot Max is an entry-level, all-purpose smart speaker, while the Echo Studio is larger, pricier and offers better sound quality. The main difference between Amazon’s smart displays, the Echo Show 8 and Echo Show 11, is the touchscreen size.
All of the devices have improved sensors, speakers and microphones.
Amazon is offering 11% off the cost of the Echo Show 11 and 10% off the Echo Dot Max as part of its Black Friday promotions.
With the upgrades, Amazon is aiming to have users engage more often with the devices than their predecessors. Consumers frequently complained that Alexa had grown outdated while the Echo devices offered little utility beyond setting timers, spouting weather forecasts, playing music and controlling smart home accessories, like turning lights on and off.
Amazon’s recent Alexa ad tries to paint a different picture.
Comedian Pete Davidson strolls through his kitchen when an Alexa-equipped Echo Show announces, unprompted, that the “Coffee’s on, and your Uber is on its way.” Davidson then casually banters back and forth with Alexa about his preferred nickname.
The interaction is meant to showcase a few of Alexa+’s biggest selling points — users don’t have to repeat a so-called “wake word” after every command, allowing the conversation to flow more naturally.
The devices can also now connect to external services to take actions on users’ behalf. As of now, Alexa+ can book an Uber or OpenTable reservation, generate a song via Suno, plan a trip through Fodor’s, schedule a repairman visit and purchase concert tickets through Ticketmaster. Amazon has said it expects to add more capabilities soon.
Alexa+ isn’t yet available to the general public. Consumers have to wait to receive Early Access or purchase a new Echo model to use it.
Amazon is offering Alexa+ for free to users with Early Access, but at some point, the company will begin charging non-Prime members $19.99 a month for the service.
The company is also making moves in wearables.
Amazon in July announced plans to acquire AI company Bee for an undisclosed amount, indicating that it could have more hardware infused with the technology in the works. Bee is known for its $50 wristband that uses AI and microphones to listen to and analyze conversations, then provide to-do lists, summaries and reminders for everyday tasks.
— Annie Palmer
A person holds Google Pixel 10, Pixel 10 Pro and Pixel 10 Pro Fold mobile phones during the ‘Made by Google’ event, organised to introduce the latest additions to Google’s Pixel portfolio of devices, in Brooklyn, New York, U.S., August 20, 2025.
Brendan McDermid | Reuters
Google’s AI-powered Pixel 10 series
Although the Gemini-powered Google Home Speaker won’t roll out until the spring, Alphabet did deliver some generative AI tech this year.
Launched in August, the Pixel 10 smartphones thoroughly integrate Google’s AI into several features, such as live translation, text-based photo editing and the built-in Gemini assistant.
The baseline Pixel 10 starts at $799, while the Pro lineup includes the $999 Pixel 10 Pro, the $1,199 Pixel Pro XL and the $1,799 Pixel 10 Pro Fold. The Pro line offers a higher quality camera and display, as well as additional video features.
Among the AI products is “Magic Cue,” which connects data across different apps to surface relevant information and suggest helpful actions. For example, if a user receives a message asking about a dinner reservation’s location, Magic Cue can find the answer from the calendar app.
For snapping pictures, Google provides an AI “Camera Coach,” which scans the scene of a photo and offers recommendations about framing, lighting and other techniques to improve the image.
The Pixel 10 Pro phones come with a one-year subscription to Google’s “AI Pro” plan, which typically costs $19 per month and offers multiple AI tools, including writing assistant NotebookLM and video generator Veo 3.
All the Pixel 10 models are currently on sale for $200 to $300 off until Dec. 6, except for the Pixel 10 Pro Fold, which has a $300 markdown until Dec. 2, the company said.
— Jaures Yip
The Meta Ray-Ban Display AI glasses at Meta headquarters in Menlo Park, California, US, on Tuesday, Sept. 16, 2025.
David Paul Morris | Bloomberg | Getty Images
Meta’s AI-infused Ray-Ban smart glasses
Meta’s partnership with eyewear giant EssilorLuxottica, originally inked in 2019, has spawned a surprise hit in the Ray-Ban Meta smart glasses that both companies are keen to boast about.
With the Meta AI digital assistant, users can command the camera-equipped glasses to take photos, play tunes and to answer questions about nearby landmarks.
In September, the two companies debuted the latest version of the glasses, dubbed Ray-Ban Meta (Gen 2).
The new model has double the battery life of its predecessor and an improved camera. It costs $379, which is $80 than the prior version.
Meta and Luxottica this year also launched two smart glasses aimed at athletes under the Oakley brand.
The $399 Oakley Meta HSTN glasses are pitched toward casual athletes who want to take photos while playing sports like golf, while the $499 Oakley Meta Vanguard smart glasses are geared toward the action-sports crowd, like skiers.
The Vanguard glasses feature a flashier wraparound design and two buttons on the frames’ underside that lets helmet-wearing athletes easily take photos and videos and perform other actions.
For those willing to spend big money and test new technology, Meta and Luxottica also rolled out the $799 Meta Ray-Ban Display glasses in September.
They are the first glasses Meta sells to the public that include a display, albeit a small one, in just one of the lenses. The display is intended to show users small bits of information, like navigation directions. The glasses also include a wristband that utilizes neural technology so users can command the device with gestures like rotating one’s fingers to adjust volume.
Buying the $799 glasses, though, is not easy.
Meta requires that people sign-up for in-person demos at stores like Best Buy and LensCrafters before buying the product, and the company warns that “availability varies by store, so you may not be able to purchase a pair immediately after your demo.”
Early reviews for the display glasses have been mixed.
Some reviewers have praised the device’s color display, camera and innovative wristband. Still, others have criticized its high price and have said its lack of apps limit functionality.
Meta is currently offering a few Black Friday and Cyber Monday deals for some of its various AI-powered smart glasses that will last until Dec. 1.
People can save 20% on all versions of the Ray-Ban Meta (Gen 1) at Best Buy, Target, Amazon and also at Meta’s website and the Ray-Ban website and stores. Meta is also offering 20% off the cost of prescription lenses for people who buy the Ray-Ban Meta (Gen 2) and Oakley Meta HSTN glasses from its website.
— Jonathan Vanian
Friend AI Pendant
Source: Friend
The AI friend you wear around as a pendant
Most AI chatbots want to make the user more productive. The makers of this smart pendant want AI to be your friend.
Users wear Friend, as the product is aptly called, around their necks while the $129 device listens to the conversations happening around it.
Friend’s chatbot is powered by Google Gemini, and it offers commentary on the user’s conversation and life. Those comments appear as notifications through the device’s corresponding smartphone app.
For example, when one reviewer played a new Taylor Swift song for her AI friend, the device commented through a notification that it didn’t “think it’s bad at all” and “pretty typical for pop.”
The device is at the center of the societal debate about the rise of AI.
Friend plastered a subway station in New York this fall with ads that suggested that the pendant was better than a real friend, promising that it “will never bail on our dinner plans.”
The posters were immediately defaced with messages like “AI wouldn’t care if you lived or died.”
Those wanting to experience what it’s like to wear around an AI friend should place orders swiftly.
The company’s website currently says units will be shipping “Winter 2025/26,” but Friend founder Avi Schiffmann told CNBC that devices ordered early enough will ship before Christmas.
— Kif Leswing
Plaud Note
Source: Plaud
Plaud, the AI recorder
The Plaud Note looks more like a credit card than a voice recorder, but it’s an ideal purchase for any note taker who wants to capture meetings, lectures or any dictation.
With over 30 hours of recording time and battery that last 60 days on standby, the slim device can produce transcriptions in 112 languages. The transcriptions include tags for each speaker on the audio.
The recorder’s companion app is powered by OpenAI’s GPT-5, Anthropic’s Claude Sonnet 4 and Google’s Gemini 2.5 Pro. The app uses those AI models to generate detailed summaries and notes. Users can select from over 3,000 summary templates, such as phone Q&As or seminar notes.
The Plaud App’s basic plan offers 300 minutes of transcription per month, though users can upgrade to a pro plan for 1,200 minutes for $8.33 per month or a more expensive unlimited plan for $19.99 per month.
The recorder can easily be attached to phones with MagSafe magnets, meaning all Apple smartphones since the iPhone 12 series, or phone cases with similar magnets.
The company also offers the Plaud NotePin, a smaller, pill-shaped version of the recorder that can be worn as a magnetic pin, clip, wristband or necklace.
Typically priced at $159, both devices are currently on sale for 20% off during Black Friday and Cyber Monday, with another 15% markdown set for Christmas, the company said.
With year-end approaching, it’s a good time to make sure your tax house is in order. It’s especially important for crypto investors, given a new IRS brokerage reporting requirement covering transactions after Jan. 1, 2025.
The IRS generally treats crypto like property, similar to stocks or real estate, so selling crypto can trigger a capital gain or loss. And while crypto investors should have been keeping good records all along, the new reporting requirement gives them an even more compelling reason. That’s because brokerages now have to send what’s known as a Form 1099-DA. For tax year 2025, they’re required to report gross proceeds for each digital asset sale the broker processes. In 2026 and beyond, it’s mandatory for brokers to report gross proceeds and cost basis information for covered securities.
Because brokers haven’t had to issue 1099s for selling or exchanging crypto in the past, it was easier for people to act as tax cheats, said Ric Edelman, financial advisor, author and founder of the Digital Assets Council of Financial Professionals. “Many people mistakenly believe that there’s no reporting obligation,” Edelman said.
As crypto investors do their tax planning for a year which saw bitcoin rise to new heights, but more recently endure a huge selloff that has shaved over $40,000 off its record price, it’s important to understand the new, stricter recordkeeping requirements.
Let’s say you bought ethereum for $1,500 and paid a $50 transaction fee, your cost basis would be $1,550, according to an example provided by Coinbase. “Essentially, your gain or loss is the difference between the gross proceeds and the cost basis. If you sold that 1 ETH for $2,000, your taxable gain would be $450 ($2,000 – $1,550).”
Get your crypto recordkeeping in order now
Brokers are required to report the cost basis information for tax year 2026, and if you haven’t been keeping good records thus far, you’re going to have to start. “It’s a taxpayer’s responsibility to track and substantiate whatever cost basis they’re providing,” said Daniel Hauffe, senior manager for tax policy and advocacy at The American Institute of Certified Public Accountants.
For many crypto investors, this will be complicated, especially if they transferred their tokens to a broker after holding them elsewhere and haven’t kept careful records. In that case, the broker won’t have the amount you purchased the crypto for; the broker would only know the price when you transferred it, Hauffe said.
Ideally, taxpayers should try to iron out these issues now, before brokers are required to report the basis, and that may require speaking to a qualified tax professional.
Crypto investors who have been keeping track of their holdings haphazardly in the past should also consider hiring a tax crypto recordkeeping provider. There are a number of these services, including ProfitStance, Taxbit, TokenTax and ZenLedger.
Edelman said it’s best to use a recordkeeping provider because of the complexities involved. “If you try to do this manually, it is complicated and you’re likely to make errors,” he said.
Crypto staking, and staking ETFs, to be a major tax focus
While the IRS issued core guidance about the tax treatment of cryptocurrency more than a decade ago, the market has changed significantly since then, underscoring the need for updated guidance in several areas.
In 2024, the IRS, in Notice 2024-57, said it was continuing to study different types of crypto transactions to determine appropriate taxation. This has left many taxpayers in limbo and scratching their heads on how to report certain types of transactions. While the IRS has said it won’t impose penalties for limited types of transactions while the regulations are being ironed out, taxpayers still have to keep careful records so they can appropriately account for them.
One area in which cryptocurrency investors are awaiting direction is staking transactions. Guidance on this and other types of more complicated crypto transactions are expected next year, Edelman said. Some advocates say taxes should only be applicable at the time these rewards are spent, sold, or otherwise disposed of. Thus far, however, the IRS has said that these rewards should be taxed as income upon receipt, Hauffe said.
Additional guidance in staking specifically could be especially important now that the IRS has confirmed exchange-traded funds issuers can provide staking rewards, said Zach Pandl, head of research at Grayscale, a digital asset-focused investment platform. The availability of cryptocurrency within ETFs has widened the playing field for ordinary investors to gain some exposure to the asset class, and the latest guidance suggests more investors will face tax consequences from staking rewards. “Staking rewards are increasingly common for investors because they’ve now been activated in ETFs,” Pandl said.
Bitcoin’s big drop could be a tax-loss advantage
For some crypto investors, there may be an opportunity in the next month or so for tax-loss harvesting, which involves selling investments at a loss and using those losses to offset gains in other investments, Pandl said.
Bitcoin’s struggles since its record highs in October could present an opportunity for investors to benefit from a tax perspective, depending on when they bought the crypto. Some investors could also benefit from tax-gain harvesting, a strategy that involves selling the investment when you think it’ll have the least impact on your taxes.
“This is the time to be thinking about that and planning for it,” said Stuart Alderoty, president of the National Cryptocurrency Association, a non-profit focused on crypto education. “You can harvest gains and you can harvest losses as well,” he said.
Many accountants don’t understand digital assets
Taxation depends largely on a person’s tax bracket and whether they are short-term or long-term gains. For example, if you’ve held the crypto for more than a year, profits are subject to long-term capital gains rates of 0%, 15% or 20%. If the crypto was held for less than a year, ordinary tax rates between 10% to 37% apply.
Due to the complexity and unique nature of crypto, determining taxation is complicated by other factors, especially since IRS rules about crypto are in flux. As one example, it is important to make sure to report the crypto transaction on the right form. For example, if you sold, exchanged or otherwise disposed of a digital asset you held as a capital asset, use Form 8949. If you were paid as an employee or independent contractor with digital assets, report the digital asset income on Form 1040, U.S. Individual Income Tax Return.
On top of that, many crypto owners are confused about the federal income tax question pertaining to digital assets. On the first page, near the top, they’re asked to identify whether at any time during the tax year, they either received (as a reward, award or payment for property or services) or sold, exchanged or otherwise disposed of a digital asset.
Many people think “received” means buy, but it doesn’t, Edelman said. Rather, the IRS says it refers to digital assets received for payment for property or services provided, a reward or award, mining, staking and similar activities or an airdrop as it relates to a hard fork.
For these and other issues regarding crypto taxation, make sure you’re talking to a tax advisor who is knowledgeable about crypto. “Most accountants are not because they haven’t had any training in this area,” Edelman said.
This week, volatility took hold of the AI trade as bubble fears continued to rise and Nvidia‘s blowout earnings failed to steady the market.
“Unless you’re the most optimistic person on the planet … you know you’re in a bubble, right?” Dan Niles, founder of Niles Investment Management, told CNBC’s Deirdre Bosa. “There is no question you’re in a bubble.”
Industry insiders raise AI bubble alarms
Industry insiders are also beginning to raise the alarm, with Alphabet CEO Sundar Pichai warning of an overrun.
“Given the potential of this technology, the excitement is very rational. It’s also true when we go through these investment cycles, there are moments we overshoot collectively as an industry,” Pichai told the BBC. “I think it’s both rational and there are elements of irrationality through moments like this.”
At a recent internal all-hands meeting, Pichai reiterated a point he’s made previously about the risks of Google not investing aggressively enough, CNBC reported Friday.
“I think it’s always difficult during these moments because the risk of underinvesting is pretty high,” said Pichai, pointing to Google’s cloud business, which just recorded 34% annual revenue growth to more than $15 billion in the quarter. Its backlog reached $155 billion.
“I actually think for how extraordinary the cloud numbers were, those numbers would have been much better if we had more compute,” he said.
Google’s AI momentum
Meanwhile, Google on Thursday surpassed Microsoft in market cap for the first time, as the search giant was lifted by renewed AI momentum. The search company launched Gemini 3 on Tuesday, which shot to the top of AI model rankings. Google also rolled out an updated version of its viral AI image generator Nano Banana on Thursday.
“I’ve never had more fun than right now,” Josh Woodward, vice president of Google Labs and Gemini, told CNBC in an interview. “I think it’s partly the pace, it’s partly the abilities these models give to people who can imagine new use cases and products. It’s unparalleled.”
Nvidia’s China threat
Nvidia’s earnings on Wednesday failed to restore confidence in the tech trade, despite the company posting a beat-and-raise quarter. Instead, the chipmaker added to fears of escalating geopolitical risk with China. Nvidia’s finance chief Colette Kress told analysts that “sizable purchase orders never materialized in the quarter due to geopolitical issues and the increasingly competitive market in China.”
Aaron Ginn, co-founder and CEO of the graphics processing unit management company Hydra Host, said the West’s attitude toward Chinese AI is the biggest threat to Nvidia’s dominance.
“We just have to accept that we fell behind the eight ball in the fact that China is a manufacturing powerhouse,” he said. “We have the ability to beat back that trade balance to where we are now leaders.”