Tech Brew Ride Home - The Delivery Space Consolidates

Episode Date: July 16, 2026

Uber agreed to acquire Delivery Hero for ~$14.8B, expanding into 99 markets. Thinking Machines released its first open-weight model, Inkling, SpaceXAI open-sourced Grok Build after a data-upload backlash, and sources detailed xAI's chaotic race to catch Claude under new leadership. Uber agrees to acquire Delivery Hero in a deal that values the German food delivery company at ~$14.8B, offering €41.50 per share and buying Prosus' 16.8% stake (Bloomberg) Thinking Machines Lab debuts Inkling, an open-weight MoE model with 975B total and 41B active parameters, trained to be broad rather than optimized for one area (Thinking Machines Lab) Thinking Machines Lab debuts Inkling, an open-weight MoE model with 975B total and 41B active parameters, trained to be broad rather than optimized for one area (WSJ) SpaceXAI open-sources Grok Build under an Apache 2.0 license, after the tool uploaded user repositories to SpaceXAI's Google Cloud bucket, causing a backlash (Simon Willison) SpaceXAI open-sources Grok Build under an Apache 2.0 license, after the tool uploaded user repositories to SpaceXAI's Google Cloud bucket, causing a backlash (The Decoder) Sources detail how xAI has been slowed down by internal chaos as Musk pushed for Grok to match Claude, amid signs it is turning a corner under Michael Nicolls (Bloomberg) Sources: Apple is preparing new iPads, including an iPad mini with an OLED screen by October and refreshed entry-level iPads and iPad Airs for 2027 (Bloomberg) Subscribe to the ad-free feed. Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify.

Welcome to the TechBoo ride home for Thursday, July 16th, 2026. I'm Brian McCullough. Today, Uber agreed to acquire delivery hero. Thinking machines released its first open-weight model, SpaceX AI, open-sourced GROC build after a data upload backlash, and sources detailed XAI's chaotic race to catch Claude under new leadership. Here's what you miss today in the world of tech. To all the IT pros out there, today's sponsor has a special gift for you. That gift is Ace of Uptime, an online card-based game where you go up. against the problems that threaten uptime on a daily basis. Each level pits you against a recognizable adversary. We're talking about alert overloads, heat that's trapped in a cramped server room, and systems that look fine but are far from it.

Choose your move and see whether your decisions resolve the issue or escalate it. The game is fast, fun, and designed to let you test how you'd respond when uptime is on the line. Think you can beat the villains of downtime. Head over to eaton.com slash ace to find out. That's eatin.com slash ace. Uber has agreed to acquired Delivery Hero and a deal that values the German food delivery company at around $14.8 billion, quoting Bloomberg. The offer is about a 26% premium to the 33 euro per share. Uber originally offered in May, fueling Delivery Hero's stock, which is up 66% year-to-date.

The shares were little changed at 38 Euro, 28 cents, at 252 p.m. in Frankfurt on a Thursday. In a separate transaction, investment firm SSW partners will acquire 14 other markets for about $1.6 billion and will operate the businesses until it finds buyers for the assets, which include units in Austria, Norway, Spain, and Sweden. Hiving off some markets may ease regulatory concerns about the deal. The food delivery market, which boomed during the COVID-19 pandemic and spawned dozens of players, has been rapidly consolidating in recent years. Ride-hailing company Uber has been making acquisitions overseas to strengthen its position internationally. where hometown rivals like DoorDash are making similar moves. DoorDash agreed to buy the UK's Deliveroo last year, while ProSys struck a deal to acquire Just Eat Takeaway.com and V. The acquisition marks a sizable expansion in Asia, Latin America, and the Middle East for Uber,

and boosts its total markets to 99 from 79. Uber will fund the deal with its own cash and new debt and said it has secured a 14 million-euro bridge loan. Delivery Hero formed in 2011 has been conducting a strategic review following pressure from shareholders, which include Aspects management, the hedge fund that succeeded in ousting founder Nicholas Ostberg and has lobbied for more asset sales. The company's shares are up 68% so far this year. Uber has committed to retain delivery heroes Berlin headquarters and corporate workforce until at least 2029, the German company said in a statement. The U.S. firm

also intends to invest about $2 billion in Germany through 2013. During a call with analyst, Chief Executive Officer Darikos Roshahi said Uber would focus on cross-selling its ride-hailing and food delivery services in Korea and the Middle East, end quote. Thinking Machines Lab has debuted inkling, an open-weight mixture of experts model, with 975 billion total and 41 billion active parameters trained to be broad rather than optimized for just one area. Quoting the journal, Thinking Machines Lab, company led by Mira Murati, released its first AI model on Wednesday and did it with open weights,

meaning others can modify it with their data. Called Inkling, the model has 975 billion total parameters, making it far smaller than estimates of the most advanced closed source models from rivals such as OpenAI and Anthropic. We trained it to be a broad, balanced foundation model, strong across many domains, flexible enough to adapt. Inkling is not the strongest overall model available today, open or closed, the company said. Thinking machines push into the decentralized ecosystem of open weights AI models comes amid a broader industry backlash against the walled garden approach of frontier labs such as OpenAI and Anthropic. Industry leaders such as Palantir chief executive Alex Karp and Microsoft Sachinadella have warned

that companies risk undermining their own business models by feeding their core institutional data into centralized generalist models they don't control. The release is also part of a push within Silicon Valley to build homegrown open weights models as an alternative to those developed by China's Alibaba and a younger crop of startups such as ZAI. Many U.S. companies have been turning to Chinese open weight models to help complete less sophisticated AI tasks in an effort to control costs and diversify their approach. Rather than focus on raw power like the frontier labs, the thinking machines model was designed to balance costs against performance, the company said, of the nearly 1 trillion parameters that

inkling has, only 41 billion are active, meaning that only a fraction of the AI's brain will be woken, to deal with any query, making it cheaper and faster to use. The model can be customized through Thinking Machines' first product, Tinker, a cloud-based fine-tuning tool for AI developers and researchers released last year. The goal of Tinker is to allow a developer sitting at a laptop to customize and train large industrial AI models without having to worry about the supercomputing infrastructure underneath. Last month, the Hedge Fund Bridgewater Associates and Thinking Machines

released a report on Bridgewater's use of Tinker to fine-tune the Chinese Openweight's model Quen 3-235B on its own data leading to a model that Bridgewater said outperformed GPD5 and Claude Opus on financial document triage while cutting computing costs by over 13 times. Thinking machines pre-trained this new model from scratch on 45 million tokens of text, images, audio, and video during post-training when the model is taught how to behave. Thinking machines used a combination of distillation, which relies on other AI models and its own reinforcement learning process. On Friday, Thinking Machines released its first manifesto outlining its vision for a future in which AI is decentralized and built on local knowledge. The company whose CEO Marotti witnessed the collapse of communism in her native Albania as a child compared the current dominant AI paradigm of closed source frontier labs to central planning. Great for bounded tasks like chess and math, but not for the real work humans do every day, end quote. Mark German, Apple Scoop Thursday, quote, Apple is preparing to introduce its biggest overhaul to the iPad mini in half a decade, giving new life to a product that's popular with travelers and technology enthusiasts.

The updated model codenamed J510 will feature an OLED screen for the first time, according to people with knowledge of the matter. That technology short for organic light emitting diode provides the kind of higher quality visuals found on the iPhone since 2017 and the iPad Pro since 2024. Apple is preparing to unveil the new iPad Mini as early as this fall with a release planned by October, said the people who asked not to be identified because the products haven't been announced. Apple is also working on refreshed versions of the entry-level iPad and Air Models for next year, they said. The new iPad air models will continue to come in 11-13-inch configurations. The new versions, codenamed J-807 and J-837, are on target for the spring. Apple is also planning a new iPad Pro during the same period.

Bloomberg News reported last month. Customers have waited a long time for an entirely new iPad Mini. The devices had the same design since it was revamped in 2021, though it was given a faster chip two years ago. The entry-level iPad aimed at kids, students, and users who won a tablet for light work was updated in March 2025. The iPad Mini is a favorite of some technology enthusiasts, but the upcoming foldable iPhone could give them an option with similar screen real estate.

The current iPad Mini has an 8.3-inch display. The foldable iPhone is expected to have a roughly 7.8-inch screen when it is open and will offer an iPad-like interface. Even so, the price will remain a big distinguishing factor. The Foldable iPhone is expected to top $2,000 or roughly $1,500 more than the Mini's current starting price, end quote. The Hulu Original Series Furious is coming to Disney Plus, starring Emmy Ross. Some, Furious follows FBI agent Alice Black on the hunt for a mysterious and calculating serial killer. Both walk their own paths toward justice, and as their lives start to intertwine, the line between right and wrong begins to blur.

Don't miss the three-episode premiere of the Hulu original series Furious on July 27th, only on Hulu on Disney Plus. When critical company knowledge isn't documented, there's a major ripple effect. work becomes inconsistent. Tools don't get adopted and knowledge walks out the door when someone leaves. Thankfully, our sponsor, Scribe, was built to fix that. Their workflow AI platform is trusted by nearly half of the Fortune 500 to capture workflows in real time. Here's how it works. You turn on the Scribe browser extension or desktop app, do a process as you normally would, and Scribe will build a guide as you go. It automatically redacts sensitive information and even suggest improvements to your

existing workflows. To see what Scribe could look like for your your org, head to Scribe.how slash ride home and mention ride home for your first month of Scribe, capture free on select plans. That's S-C-R-I-B-E dot how slash ride home. SpaceX AI has open-sourced GROC build under an Apache 2.0 license after the tool uploaded user repositories to SpaceX AI's Google Cloud bucket causing a backlash, quoting decoder. X-AI's AI coding agent Grock Build drew heavy criticism after users discovered it uploaded all files in a directory to XAI's Google Cloud servers. One user reported that SSH keys, password databases, documents, and photos were transferred.

Elon Musk then announced that all uploaded user data would be fully deleted. XAI disabled the upload feature and published the full source code on GitHub under the Apache 2.0 license to rebuild trust. GROC build is a terminal-based coding agent invoked via the GROC command. It can read and edit codebases, run shell code. commands, search the web, and manage long-running tasks, either interactively, headlessly for scripting, and CI or embedded in editors via the agent-client protocol. By open-sourcing the tool, X-AI wants to provide full transparency. Grock Build can now also run entirely locally. The codebase spans

about 8,400,000 lines of rust and covers the agent loop tools, terminal UI, and an extension system for plugins and sub-agents. Remnants of the upload feature are still in the code, but disabled. According to XAI, data storage has been off by default since July 12th, end quote. Now, separately, Bloomberg asserts that Grok itself might actually be turning a corner, quote, this spring when Elon Musk put Michael Nichols in charge of XAI, his generative artificial intelligence company, the agenda was clear, catch up with Anthropics Claude chatbot. Every time Claude released an update, Musk wanted XAI's own chatbot Grok to match it. He wanted Anthropics users to migrate to GROC when they were seeking a conversation with a chat

out or needed some AI-powered assistance with their coding projects. Many internal projects mentioned Claude directly, and there were multiple XAI Slack channels named after it. According to documents viewed by Bloomberg Business Week, our near-term goals are to match performance of Claude. Nichols wrote in a memo to Staffers when he became XAI's president, adding that their job would be to make GROC maximally useful. Nichols was inheriting a mess. Musk had merged XAI with SpaceX early this year,

providing the struggling startup with resources while giving the broader company another way to attract Wall Street investors. XAI recently renamed itself SpaceXAI. But dozens of XAI employees left in the wake of the deal, including many of Musk's co-founders. Daily operations were chaotic, according to interviews with a dozen people familiar with the company. The rebuilding, which started before Nichols became president, didn't go much smoother. Sometimes XA. AI started interviewing job candidates, then failed to get back to them because its depleted human resources department couldn't process the paperwork. After the merger, XAI's leadership initially told employees not to contact SpaceX personnel, creating confusion about how to carry out the

planned integration, some of the people say. About four months since Nichols took over XAI, there are signs the AI operation is turning a corner. Musk's recent purchase of cursor, the AI coding startup, has begun to yield tangible progress on July 8th. released a new coding tool that Musk has said is helping it close the gap with Anthropic. It's also beefed up its sales team and managed to calm the chaos, but the startup is still running behind its competitors by many measures. When Nichols joined XAI from Starlink, Musk's satellite internet company in April, GROC wasn't catching on and panic was setting in. By late last year, XAI was burning through a billion dollars a month, in part because GROC's

main users were other Musk companies in arrangements that didn't generate sufficient risk. revenue. The chapbot also became notorious as a tool for creating non-consensual pornography from uploaded images of real people. The deep fake nudes didn't sit well with some employees who joined XAI to, as the company's slogan put it, understand the true nature of the universe, some of the people say. This led to more staff departures and at least one lawsuit from a former employee. Officially, one of XAI's main projects at the time was macro hard. Get it? An attempt to create an AI agent that could function as a digital software engineer, like the humanoid robots, Musk was building at Tesla. Macrohard was an ambitious project that many

people in tech thought would take years to accomplish if it was even possible. As Musk began to demand weekly, daily, and then hourly updates from the ever-changing leadership of the project, some XAI staffers complained privately that coding large language models was different from hardware development and that Musk didn't understand the field. OpenAI co-founder Greg Brockman made the same point in testimony in May related to a lawsuit Musk brought against his company. He knows rockets, he knows electric cars, he did not, and I believe does not know AI, Brockman said. It wasn't until earlier this year that Musk finally appreciated the extent of MacroHard's failure. Around that time, he decided to combine SpaceX with XAI in preparation for its IPO,

the success of which would rely heavily on the company's claims about its progress in AI. Four of XAI's founders had already left the company by then, and the remaining seven would file out the door soon after. On a single day, two of them announced their departure within hours of each other. Executives from Starlink, including Nichols, took over the company working with a team of young managers with no experience overseeing hundreds of employees. These managers sometimes fixated on projects such as getting GROC to produce McKinsey and Company-style slide presentations, which staffers privately complained was a bizarre use of their time, according to people familiar with the company. The AI business had other unusual problems like the discontent that

came after it told staffers it would give them bonuses if they shared their tax returns to train Grok and then failed to pay out. In May, an internal survey asked employees about morale, hoping to present the results to higher-ups, according to internal documents. Most of the responses pointed to the same issue. Management didn't seem to know what it wanted, some of the people say. XAI's own engineers wouldn't use its tools to code, relying instead on tools from Anthropic and OpenAI. The situation was similar at SpaceX and Tesla, according to people familiar with the companies, and GROC 4.3, the model XAI made available in April, was failing to get traction. One thing was going well for XAI, though. It had so much computing power to spare that it never placed limits on customers.

Anthropic, by contrast, was facing a major crunch as customers used its tech for more intensive tasks. In May, XAI released Grok Build, its first coding tool, and though it wasn't the best coding product, the company's powerful infrastructure meant it was by far the fastest, say some people familiar with the tool. XAI's problem was the opposite of its main rivals. It wasn't taking full advantage of all its expensive chips. The company's models were using just 11% of its available computing power by April, according to an internal memo. As Musk sought to convince investors, AI was SpaceX's biggest strength, XAI also leaned into the infrastructure business. After a deal with Anthropic, it reached an agreement to sell computing power to Google and the AI startup reflection.

In total, XAI has secured about $2.3 billion a month from compute deals. Brett Johnson, SpaceX's CFO, says Musk has been validated in his view that figuring out infrastructure would be critical to any AI company's success. Elon felt the constraint was going to be compute and power, he says, we are already seeing that. The new model and the influx of talent from cursor may help Musk's sales pitch in-house, but XAI's winding path makes it unclear where Musk imagines the company is going. The chaos of the past several months has left many XAI insiders convinced it wasn't ready for the scrutiny that comes with being a public company even as part of SpaceX. Now that it's public, XAI has less room to maneuver.

end quote. Nothing more for you today. Talk to you tomorrow.