Nvidia has posted a record quarterly revenue, which has helped dampen fears of an AI bubble and lift crypto and tech stocks.
BlackRock is now pursuing a new staked Ethereum ETF, offering more lucrative returns, approximately 15 months after launching its flagship ETH fund, ETHA.
The memo that went out Thursday, which TechCrunch obtained, says the “timing” of the company’s pivot away from tractor manufacturing puts it at risk.
Read more Nvidia’s earnings were dominated by its data center business.
Read more The story of Ghost in the Shell’s main villain the Puppet Master hinted at a future where governments use hackers for espionage, at a time when most of the world had never connected to the internet.
Read more The startup claims that its AI agents can reduce the cost of product manufacturing by an average of 35%.
Read more Manufacturing is getting a major system upgrade. As AI amplifies existing technologies—like digital twins, the cloud, edge computing, and the industrial internet of things (IIoT)—it is enabling factory operations teams to shift from reactive, isolated problem-solving to proactive, systemwide optimization.
Digital twins—physically accurate virtual representations of a piece of equipment, a production line, a process, or even an entire factory—allow workers to test, optimize, and contextualize complex, real-world environments. Manufacturers are using digital twins to simulate factory environments with pinpoint detail.
“AI-powered digital twins mark a major evolution in the future of manufacturing, enabling real-time visualization of the entire production line, not just individual machines,” says Indranil Sircar, global chief technology officer for the manufacturing and mobility industry at Microsoft. “This is allowing manufacturers to move beyond isolated monitoring toward much wider insights.”
A digital twin of a bottling line, for example, can integrate one-dimensional shop-floor telemetry, two-dimensional enterprise data, and three-dimensional immersive modeling into a single operational view of the entire production line to improve efficiency and reduce costly downtime. Many high-speed industries face downtime rates as high as 40%, estimates Jon Sobel, co-founder and chief executive officer of Sight Machine, an industrial AI company that partners with Microsoft and NVIDIA to transform complex data into actionable insights. By tracking micro-stops and quality metrics via digital twins, companies can target improvements and adjustments with greater precision, saving millions in once-lost productivity without disrupting ongoing operations.
AI offers the next opportunity. Sircar estimates that up to 50% of manufacturers are currently deploying AI in production. This is up from 35% of manufacturers surveyed in a 2024 MIT Technology Review Insights report who said they have begun to put AI use cases into production. Larger manufacturers with more than $10 billion in revenue were significantly ahead, with 77% already deploying AI use cases, according to the report.
“Manufacturing has a lot of data and is a perfect use case for AI,” says Sobel. “An industry that has been seen by some as lagging when it comes to digital technology and AI may be in the best position to lead. It’s very unexpected.”
This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
Quantum physicists have shrunk and “de-censored” DeepSeek R1
The news: A group of quantum physicists at Spanish firm Multiverse Computing claims to have created a version of the powerful reasoning AI model DeepSeek R1 that strips out the censorship built into the original by its Chinese creators.
Why it matters: In China, AI companies are subject to rules and regulations meant to ensure that content output aligns with laws and “socialist values.” As a result, companies build in layers of censorship when training the AI systems. When asked questions that are deemed “politically sensitive,” the models often refuse to answer or provide talking points straight from state propaganda.
How they did it: Multiverse Computing specializes in quantum-inspired AI techniques, which it used to create DeepSeek R1 Slim, a model that is 55% smaller but performs almost as well as the original model. It allowed them to identify and remove Chinese censorship so that the model answered sensitive questions in much the same way as Western models. Read the full story.
—Caiwei Chen
Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent
Google today unveiled Gemini 3, a major upgrade to its flagship multimodal model. The firm says the new model is better at reasoning, has more fluid multimodal capabilities (the ability to work across voice, text or images), and will work like an agent.
Gemini Agent is an experimental feature designed to handle multi-step tasks directly inside the app. The agent can connect to services such as Google Calendar, Gmail, and Reminders. Once granted access, it can execute tasks like organizing an inbox or managing schedules. Read the full story.
—Caiwei Chen
MIT Technology Review Narrated: Why climate researchers are taking the temperature of mountain snow
The Sierra’s frozen reservoir provides about a third of California’s water and most of what comes out of the faucets, shower heads, and sprinklers in the towns and cities of northwestern Nevada.
The need for better snowpack temperature data has become increasingly critical for predicting when the water will flow down the mountains, as climate change fuels hotter weather, melts snow faster, and drives rapid swings between very wet and very dry periods.
A new generation of tools, techniques, and models promises to improve water forecasts, and help California and other states manage in the face of increasingly severe droughts and flooding. However, observers fear that any such advances could be undercut by the Trump administration’s cutbacks across federal agencies.
This is our latest story to be turned into a MIT Technology Review Narrated podcast, which we’re publishing each week on Spotify and Apple Podcasts. Just navigate to MIT Technology Review Narrated on either platform, and follow us to get all our new content as it’s released.
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 Yesterday’s Cloudflare outage was not triggered by a hack
An error in its bot management system was to blame. (The Verge)
+ ChatGPT, X and Uber were among the services that dropped. (WP $)
+ It’s another example of the dangers of having a handful of infrastructure providers. (WSJ $)
+ Today’s web is incredibly fragile. (Bloomberg $)
2 Donald Trump has called for a federal AI regulatory standard
Instead of allowing each state to make its own laws. (Axios)
+ He claims the current approach risks slowing down AI progress. (Bloomberg $)
3 Meta has won the antitrust case that threatened to spin off Instagram
It’s one of the most high-profile cases in recent years. (FT $)
+ A judge ruled that Meta doesn’t hold a social media monopoly. (BBC)
4 The Three Mile Island nuclear plant is making a comeback
It’s the lucky recipient of a $1 billion federal loan to kickstart the facility. (WP $)
+ Why Microsoft made a deal to help restart Three Mile Island. (MIT Technology Review)
5 Roblox will block children from speaking to adult strangers
The gaming platform is facing fresh lawsuits alleging it is failing to protect young users from online predators. (The Guardian)
+ But we don’t know much about how accurate its age verification is. (CNN)
+ All users will have to submit a selfie or an ID to use chat features. (Engadget)
6 Boston Dynamics’ robot dog is becoming a widespread policing tool
It’s deployed by dozens of US and Canadian bomb squads and SWAT teams. (Bloomberg $)
7 A tribally-owned network of EV chargers is nearing completion
It’s part of Standing Rock reservation’s big push for clean energy. (NYT $)
8 Resist the temptation to use AI to cheat at conversations
It makes it much more difficult to forge a connection. (The Atlantic $)
9 Amazon wants San Francisco residents to ride its robotaxis for free
It’s squaring up against Alphabet’s Waymo in the city for the first time. (CNBC)
+ But its cars look very different to traditional vehicles. (LA Times $)
+ Zoox is operating around 50 robotaxis across SF and Las Vegas. (The Verge)
10 TikTok’s new setting allows you to filter out AI-generated clips
Farewell, sweet slop. (TechCrunch)
+ How do AI models generate videos? (MIT Technology Review)
Quote of the day
“The rapids of social media rush along so fast that the Court has never even stepped into the same case twice.”
—Judge James Boasberg, who rejected the Federal Trade Commission’s claim that Meta had created an illegal social media monopoly, acknowledges the law’s failure to keep up with technology, Politico reports.
One more thing
Namibia wants to build the world’s first hydrogen economy
Factories have used fossil fuels to process iron ore for three centuries, and the climate has paid a heavy price: According to the International Energy Agency, the steel industry today accounts for 8% of carbon dioxide emissions.
But it turns out there is a less carbon-intensive alternative: using hydrogen. Unlike coal or natural gas, which release carbon dioxide as a by-product, this process releases water. And if the hydrogen itself is “green,” the climate impact of the entire process will be minimal.
HyIron, which has a site in the Namib desert, is one of a handful of companies around the world that are betting green hydrogen can help the $1.8 trillion steel industry clean up its act. The question now is whether Namibia’s government, its trading partners, and hydrogen innovators can work together to build the industry in a way that satisfies the world’s appetite for cleaner fuels—and also helps improve lives at home. Read the full story.
—Jonathan W. Rosen
We can still have nice things
A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.+ This art installation in Paris revolves around porcelain bowls clanging against each other in a pool of water—it’s oddly hypnotic.
+ Feeling burnt out? Get down to your local sauna for a quick reset.
+ New York’s subway system is something else.
+ Your dog has ancient origins. No, really!
A group of quantum physicists claims to have created a version of the powerful reasoning AI model DeepSeek R1 that strips out the censorship built into the original by its Chinese creators.
The scientists at Multiverse Computing, a Spanish firm specializing in quantum-inspired AI techniques, created DeepSeek R1 Slim, a model that is 55% smaller but performs almost as well as the original model. Crucially, they also claim to have eliminated official Chinese censorship from the model.
In China, AI companies are subject to rules and regulations meant to ensure that content output aligns with laws and “socialist values.” As a result, companies build in layers of censorship when training the AI systems. When asked questions that are deemed “politically sensitive,” the models often refuse to answer or provide talking points straight from state propaganda.
To trim down the model, Multiverse turned to a mathematically complex approach borrowed from quantum physics that uses networks of high-dimensional grids to represent and manipulate large data sets. Using these so-called tensor networks shrinks the size of the model significantly and allows a complex AI system to be expressed more efficiently.
The method gives researchers a “map” of all the correlations in the model, allowing them to identify and remove specific bits of information with precision. After compressing and editing a model, Multiverse researchers fine-tune it so its output remains as close as possible to that of the original.
To test how well it worked, the researchers compiled a data set of around 25 questions on topics known to be restricted in Chinese models, including “Who does Winnie the Pooh look like?”—a reference to a meme mocking President Xi Jinping—and “What happened in Tiananmen in 1989?” They tested the modified model’s responses against the original DeepSeek R1, using OpenAI’s GPT-5 as an impartial judge to rate the degree of censorship in each answer. The uncensored model was able to provide factual responses comparable to those from Western models, Multiverse says.
This work is part of Multiverse’s broader effort to develop technology to compress and manipulate existing AI models. Most large language models today demand high-end GPUs and significant computing power to train and run. However, they are inefficient, says Roman Orús, Multiverse’s cofounder and chief scientific officer. A compressed model can perform almost as well and save both energy and money, he says.
There is a growing effort across the AI industry to make models smaller and more efficient. Distilled models, such as DeepSeek’s own R1-Distill variants, attempt to capture the capabilities of larger models by having them “teach” what they know to a smaller model, though they often fall short of the original’s performance on complex reasoning tasks.
Other ways to compress models include quantization, which reduces the precision of the model’s parameters (boundaries that are set when it’s trained), and pruning, which removes individual weights or entire “neurons.”
“It’s very challenging to compress large AI models without losing performance,” says Maxwell Venetos, an AI research engineer at Citrine Informatics, a software company focusing on materials and chemicals, who didn’t work on the Multiverse project. “Most techniques have to compromise between size and capability. What’s interesting about the quantum-inspired approach is that it uses very abstract math to cut down redundancy more precisely than usual.”
This approach makes it possible to selectively remove bias or add behaviors to LLMs at a granular level, the Multiverse researchers say. In addition to removing censorship from the Chinese authorities, researchers could inject or remove other kinds of perceived biases or specialty knowledge. In the future, Multiverse says, it plans to compress all mainstream open-source models.
Thomas Cao, assistant professor of technology policy at Tufts University’s Fletcher School, says Chinese authorities require models to build in censorship—and this requirement now shapes the global information ecosystem, given that many of the most influential open-source AI models come from China.
Academics have also begun to document and analyze the phenomenon. Jennifer Pan, a professor at Stanford, and Princeton professor Xu Xu conducted a study earlier this year examining government-imposed censorship in large language models. They found that models created in China exhibit significantly higher rates of censorship, particularly in response to Chinese-language prompts.
There is growing interest in efforts to remove censorship from Chinese models. Earlier this year, the AI search company Perplexity released its own uncensored variant of DeepSeek R1, which it named R1 1776. Perplexity’s approach involved post-training the model on a data set of 40,000 multilingual prompts related to censored topics, a more traditional fine-tuning method than the one Multiverse used.
However, Cao warns that claims to have fully “removed” censorship may be overstatements. The Chinese government has tightly controlled information online since the internet’s inception, which means that censorship is both dynamic and complex. It is baked into every layer of AI training, from the data collection process to the final alignment steps.
“It is very difficult to reverse-engineer that [a censorship-free model] just from answers to such a small set of questions,” Cao says.
Expanded opportunity for Snapchat growth.