← Tech Brew Ride Home
Tech Brew Ride Home
China (AI) Rising
00:21:22
Chinese AI models grabbed over 30% of US token use as Beijing weighed curbing access. Anthropic expanded Cowork to mobile and web, Meta launched its first AI image generator on Instagram and WhatsApp, and Xbox's $80B Game Pass bet failed.
Loading summary
Transcript9 lines
- [00:01]
A
Are all batteries the same? That's like asking if all soccer players are the same. Take Messi, the most decorated player ever. Is there any other player who has achieved that? No, just him. Now take Duracell. Is there any other battery with powerboost ingredients inside? No, just Duracell. Remember, goats only trust goats because they're built different and Messi only trusts Duracell.
- [00:35]
B
Welcome to the Tech Brew Ride Home for Wednesday, July 8, 2026 I'm Brian McCullough. Today, Chinese AI models grabbed over 30% of US token use as Beijing weighs curbing access. Anthropic expanded cowork to mobile and the web. Meta launched its first AI image generator on Instagram and WhatsApp and Xbox's $80 billion Game Pass bet has officially failed. Here's what you missed today in the world of tech. Every day shareholders meet to discuss important matters about the companies you invest in. Now you can make your voice heard, too. Vanguard Investor Choice makes it easy to set your proxy voting preferences for your eligible Vanguard and index funds. Whether you hold a Vanguard fund directly or through another brokerage firm, all it takes is a few clicks to select your proxy voting preferences and be heard on important shareholder topics like executive pay and Director elections. Visit vanguard.com investor choice to learn more. It's your shares. It's your voice. It's easy. Vanguard investors own shares of our index funds, and those funds own shares in the companies they invest in. Vanguard Marketing Corporation Distributor I'm going to continue banging on this drum because I think it's important. In the background of the race to IPO between OpenAI and Anthropic, there continues to be this which, if you think about it, may be contributing to the sense of urgency to get to the open market. According to OpenRouter, Chinese AI models have drawn more than 30% of token usage by US companies each week since February 8, peaking at 46%, which would be up from 11% over the previous 12 months. So it's more than tripled. Quoting CNBC. Chinese built AI models are gaining traction among US companies as they narrow the performance gap with leading American rivals while remaining significantly cheaper to use. Recent models releases from Chinese companies including Deepseek and Zai are seen by many as highly competitive compared to leading frontier systems from the likes of Anthropic and OpenAI. Those advances in capability come as token prices for the most advanced models rise at many US AI labs, leaving companies grappling with unexpectedly high costs associated with using the tech. The share of tokens used by US companies on Chinese AI models via OpenRouter, a platform that enables developers to access a range of AI models, has sat above 30% each week since February 8, with that figure rising as high as 46%. The average across the previous 12 months was just 11%, falling to 4.5% in the first half of 2025. Chinese AI models are particularly attractive to American companies now as AI costs skyrocket, kyle Chan, fellow at the John L. Thornton China center at think tank Brookings, told CNBC. Where previous U.S. companies were prioritizing AI adoption regardless of model, now they're getting more cost conscious. End quot In June, AI Startup Lindi moved 100% of its traffic from Anthropic's Claude models to Deepseek, a Chinese company which burst onto the scene with a bombshell Release in early 2025 and launched a new model in April. We did it and you could see the cost curve go down like crash to the ground, CEO Flo Crevello told cnbc. He said the decision will save Linde millions of dollars within months. Deepseek saw its share of gateway tokens climb between May and June on Vercel, a platform that allows developers to dep, deploy and run apps and websites. Zai's GLM 5.2 was released to great fanfare in June and saw the fastest adoption of any model tracked by Vercel in 2026, Harpreet Arora, head of agentic infrastructure at Vercel, told CNBC. In its first full week after launch, daily token volume grew about 27x and the number of customers using it grew about 80x. Price is doing the work here, aurora said. When a task doesn't need the best model, teams are beginning to route it to the cheapest one that's good enough. And the recent wave of models coming out of CH China is winning that trade. Open source Chinese models can be 60 to 90% cheaper than the leading anthropic and OpenAI models, Justin Somerville, who works on data and analytics at OpenRouter, told CNBC. We're seeing companies increasingly motivated to turn to cheaper AI stacks. They can control and adapt themselves. And given the state of open source and open weight models, that often mean leveraging Chinese options, yassin Jerntay, head of machine learning at Hugging Face, told cnbc. There's a real risk that users get stuck having to choose between performance formant but expensive US proprietary models whose price and accessibility can quickly fluctuate, or using Chinese models as the only feasible alternative whenever they want to control costs or their own AI stack. Yes, but what if that option is taken away? Quoting Reuters Chinese authorities have held meetings with top tech firms over the past month about potentially restricting overseas access to China's most advanced AI models, including those yet to be released, three people familiar with the discussion said. The talks follow a number of steps by Beijing to keep homegrown AI within the country and underscore how China, like the us is now treating cutting edge artificial intelligence as a critical national asset that needs controls. Companies present at the talks include tech giants Alibaba and ByteDance, as well as startup Zai, said the people, who were not authorized to speak to media and declined to be identified. At the meetings, led by China's Ministry of Commerce, participants discussed putting limits on the most advanced AI models, both closed source and more open versions, according to two of the sources. Officials talked about making any leak or theft of proprietary AI technology an offense under China's stringent national security law, one of the sources said. The officials also raised the possibility of implementing new measures to restrict who can fund domestic AI startups, the source added. The scope of the potential restrictions is still being discussed, two sources said, adding that they may only apply to future models, though it was not immediately clear when or even if they would come into force. End quote. I've heard of Santa Sam, but Claudesmas in July Anthropic has extended Claude Fable 5 access to all paid plans through July 12th. Access to that model was set to shift to token based usage yesterday. I've been hearing all week that we're probably going to get GPT 5.6 sometime this week, probably on Thursday, so maybe this front runs that just a bit. Also, there is this quoting ZDNet Anthropic announced today it's bringing the agentic helper for Claude users called Claude Cowork to the web and to mobile devices. After analyzing 1.2 million cowork sessions, Anthropic is also sharing some unexpected usage details. Let's start with the Cowork upgrade. According to the company, a user can start a task at a desk, check progress from a phone, and pick up finished output from any browser. It says you no longer need your laptop open to run Cowork. I normally run Cowork on my massive Mac studio with 128 gigabytes of RAM and a 38 inch monitor. But it's the same idea. The machine could be off and the task would still run. Close the laptop and head to your meeting. Claude keeps going, said Anthropic. In fact, scheduled tasks can work entirely online as long as the task itself is is online. So Cowork can sift through email threads, check transcripts, and examine news all while you sleep. The limitation is that you have to use connected apps. Claude code doesn't have access to your browser experience. There's a mobile version of the app for Android and iOS, and this serves two purposes. First, you can check status and get notifications on your phone. If Claude Cowork reaches a point where it needs permission or clarification, it'll push a notification to your phone. Drafts of messages won't be sent until you review and approve them. Along with the new feature announcement, Anthropic offered up some very interesting statistics derived from its session logs. Anthropic said it sampled 1.2 million anonymized and aggregated Claude cowork sessions from May 11 to May 31 across more than 600,000 organizations. With all the hullabaloo about Claude code, you might have expected that Cowork use would be dominated by coding related tasks. But Anthropic said that more than 90% of the sampled sessions were unrelated to software development. A full 50% of cowork usage can be attributed to business process content creation. Business process and operations was the largest category at 33.4%, followed by content creation and copywriting at 16.4%. Our data suggests that people are using Claude Cowork to assemble and structure the information they can use to act on their expertise, anthropic said. So here's the thought that's been banging around in my head ever since I learned about this announcement and started writing this piece. If I have Cowork and it runs autonomously in the cloud and on schedule, do I even need a dedicated server with OpenClaw anymore? Can't most of what I want done with OpenClaw be done with Claude and be safer and more reliable at the same time? OpenClaud doesn't incur usage limits with on device AI. It doesn't have a fee. But if I'm already paying for Claude Max, how far can I take it? Stay tuned. I'm sure I'll be back with some hands on tests. End quote. Think about the last time you walked
- [10:05]
C
into a big meeting and wished your talking points could just stay with you as you spoke. Don't worry about taking copious notes. Instead, there's even G2, their productivity smart glasses designed with teleprompting conversation support, real time translation and more. And for the developers, listening. Even G2 just launched Terminal, a new feature that keeps your AI coding agent in view. If you use Claude code or codec, you can see when your agent is running, waiting for approval or needs direction. Then approve or guide it directly from your glasses. To learn more about even G2 go to evenrealities.com use promo code techbrew to get 10% off evenring1 and or even clip when you add them to your even G2 order. Ever spent hours of your workday tracking down information only to find that it wasn't documented at all? It's an age old corporate conundrum. Critical knowledge isn't available or accessible to the folks who need it. That's the very problem our sponsor Scribe was built to fix. Scribe is a workflow AI platform that automatically turns your workflows into clear cut documentation. All you do is turn on the Scribe browser extension or desktop app and go through your processes as you would normally.
- [11:19]
B
Scribe builds a guide as you go,
- [11:21]
C
capturing every click, step and screenshot automatically.
- [11:24]
B
So what would have taken hours gets
- [11:26]
C
done in under a minute. To see what Scribe could look like for your org, head to Scribe How Ridehome and mention Ride home for your first month of Scribe capture free on select plans. That's S C R I B E How Ridehome
- [11:45]
B
Meta has launched Muse image in Meta, AI, Instagram and WhatsApp and is previewing Muse Video, the first media generation models from its Superintelligence labs. Quoting the times, the new AI image model, Muse Image, is available on Instagram and WhatsApp. The company said users can generate AI photos similar to those they would normally post, including vacation selfies from a beach or a reel from a photo booth. Muse Image acts as the creative partner that knows your world, making it easy to turn your ideas into high quality visuals, the company said in a blog post it added, one tap restores an old family photo, lets you see yourself with trending hairstyles, or reimagines you as a claymation character. The company also previewed an AI video generator, Muse Video. It will become available to the public in the coming months in the Meta AI app, the standalone platform for Meta's AI chatbot, MuseImage is the first AI image generator created under the company's new AI division, Meta Superintelligence Labs. After Meta fell behind in the AI race last spring, Mark Zuckerberg, the chief executive, spent billions of dollars revamping the division in an attempt to compete with Google, Anthropic, OpenAI and others in the last year. Meta has slowly narrowed the gap. But Muse Spark, its top chatbot model released in April still lags rivals MuseImage will replace technology from Midjourney, an AI startup that Meta previously worked with to generate AI images in the Meta AI app, Users can tag friends in the app to create personalized birthday cards, group trip memes, or playful edits based on their friends Instagram accounts, the company said. In the coming weeks, advertisers will be able to use the new AI model to make ads, end quote. But there's also this little nugget. Do you have a public Instagram account? Well, without asking your permission, other users could now pull you into their AI creations, quoting Wired as part of this update, public Instagram profiles are now automatically opted into being fodder for generative AI remixes. All someone has to do is tag your account's profile in a prompt if it's public, and they can use Meta AI to generate an image using your likeness. Meta positions this feature as a cheeky way to personalize generations with images of real people. Whether you want to design a custom event invitation, mock up a collaborative creative concept, or generate a personalized graphic, tagging a username lets Meta AI use public photos to build a visual that's ready to post, reads one of Meta's announcement blogs about the new AI tool. If you want to avoid these AI generations of your Instagram posts without switching your account to private, you'll have to dig into the app's settings. Open the Instagram app, tap your profile, and then tap the three lines in the top right corner of the screen, then scroll down to the Sharing and Reuse tab. Here is where you should see a section labeled Allow people to use your content on Instagram and with AI features on Meta. With a toggle for posts and one for reels, Instagram's Help center site includes more details about how this feature will impact users, saying that people may be able to create content with your Instagram content using AI features @meta. If you leave your account public and on the default settings. A previously archived version of this page from 2025 does not include similar AI focused language. While switching your account to private or adjusting your settings will prevent additional images from being generated, already existing AI images made with your content will not be deleted. It's also notable that Instagram users may be remixing your images using AI without your knowledge, you will not be notified about content created using AI features at Meta reads the company's help page. Of the numerous aspects of this rollout that make me uncomfortable, not even getting a notification when someone does this seems like the biggest red flag of all even though it's overwhelming just how often companies require you to opt out of AI training rather than opting in. Just like how Google Search now stores media uploads like from reverse image searches to train its AI as the default option for consumers, this new toggle is worth switching. Finally, today I'll admit to not being as plugged into the gaming world as maybe I should have to be able to grok this happening. But the reason why Microsoft is basically doing a full turn it off, take out the cartridge and blow on it to its Xbox business is interesting to me. Quoting Bloomberg Microsoft's Xbox spent nearly $80 billion in the last decade on deals that would give it popular video game titles like Call of Duty and Skyri, betting that gamers would flock to its Netflix like subscription service offering hundreds of options for endless play. What started as an attempt to boost Xbox's last place position among console makers after Sony's PlayStation and Nintendo Switch ended in a glut of overspending, leaving Xbox bloated with studios, staff and too many games people didn't want to play. When Xbox launched its Game Pass subscription service in 2017, executives set a goal of reaching 77 million subscribers by fiscal year 2026, which ended last month, according to a document published during a lawsuit contesting Microsoft's 2023 purchase of Activision Blizzard. Today, the platform has just 30 million subscribers, according to a person familiar with the matter, 4 million fewer than it did when the company last publicly shared data in 2024. Employees had been growing concerned that the number of Game Pass subscribers had peaked, according to two former employees. Later, millions of customers cancelled their subscriptions after Game Pass announced last October it was raising prices by 50%. Executives at Xbox hoped the $20 monthly Game Pass ultimate subscription, which rose to $30 last October before Xbox reverted to $23 in April, would draw users into its content ecosystem and eventually boost sales. Some employees warned that putting in demand high budget games on Game Pass the day they released could devalue them and cannibalize sales of individual higher priced copies. It was a controversial decision internally, according to five former employees When Game Pass launched, it was design activate back catalog games without a clear path to potential audiences, not essentially offer hit games at a fraction of the price. Call of Duty, for example, cost $70 before it landed on Game Pass, where for less than half the price players could access that game and hundreds of others. In 2024, Xbox gave up more than $300 million in sales of Call of Duty on Xbox consoles and PCs, a person familiar with the matter told Bloomberg last year. Instead, PlayStation accounted for 82% of Call of Duty Black Ops 6 sales that year under Game Pass's problem was a central flaw, according to the people. While consumers might be willing to pay a monthly fee to Netflix to stream thousands of TV shows and movies, most gamers don't take the same approach, preferring to stick with a handful of favorite games they play on repeat. Research from Cercana shows that the majority of us gamers buy two games at most per year, and a third of the market doesn't even purchase one. So so what's interesting to me is I'm positing that maybe it's all the same across media. You know how all Hollywood can produce is iterations on old IP and new original content is increasingly rare at the multiplex. We know that the vast majority of streams on Spotify is back catalog stuff. So what if games are the same way? I'll maybe buy two new games a year, but I'm just as happy to go back and play Red Dead 2 again as I am to try the new Grand Theft Auto. Does all of media work this way? There's this burst of creativity when a medium is new and then we just spend the rest of the cycle going back to the classic media. This is getting to you a little early because we have to fly back to the States all day today. Literally all day. I'm up again at 6am and I likely won't sleep again for at least 20 hours. So talk to you tomorrow, which for me will likely feel like next week.
- [19:50]
D
Hi, Ryan Reynolds here for Mint Mobile. Are you looking for a beach read this summer? May I suggest your big wireless build? It's got suspense, mystery, a slightly flat emotional arc, and a shocking twist where you realize you've been overpaying the entire time. Fortunately though, mine Mint Story is better. Every plan $15 a month, even unlimited. That's it. Happy ending. Zero tears. Give it a try at mintmobile. Com. Switch upfront payment of $45 for three months, $90 for six months or $180 for 12 month plan required $15 per month equivalent taxes and fees extra initial plan term only greater than 50 gigabytes may slow when network is busy. See terms.