India has 100 million weekly active ChatGPT users, making the country one of OpenAI’s largest markets globally, CEO Sam Altman said ahead of a government-hosted AI summit.

On Sunday, Altman outlined ChatGPT’s growing adoption in India in an article published in the Indian English daily Times of India, as OpenAI prepares to formally participate in the five-day India AI Impact Summit in New Delhi, beginning Monday. Altman is attending the event alongside senior executives from several of the world’s leading AI companies.

The growth comes as OpenAI, like other leading AI firms, looks to India’s young population and its more than a billion internet users to fuel global expansion. The ChatGPT maker opened a New Delhi office in August 2025 after months of groundwork in the country, and has adjusted its approach for India’s price-sensitive market, including rolling out a sub-$5 ChatGPT Go tier that was later made free for a year for Indian users.

In the article, Altman said India is ChatGPT’s second-largest user base after the United States, highlighting the South Asian nation’s growing weight in OpenAI’s global strategy. The disclosure comes as ChatGPT’s overall usage has surged worldwide, with the platform reaching 800 million weekly active users as of October 2025 and reported to be approaching 900 million.

Altman also highlighted the role of students in driving adoption, saying India has the largest number of student users of ChatGPT globally.

Indian students have become a key growth segment for leading AI companies more broadly, as rivals race to embed their tools in classrooms and learning workflows. Google has similarly targeted the market, offering Indian students a free one-year subscription to its AI Pro plan in September 2025. Separately, India accounts for the highest global usage of Gemini for learning, Chris Phillips, Google’s vice president and general manager for education, said last month.

“With its focus on access, practical Al literacy, and the infrastructure that supports widespread adoption, India is well positioned to broaden who benefits from the technology and to help shape how democratic AI is adopted at scale,” Altman wrote.

ChatGPT’s rapid growth also highlights a broader challenge for AI companies in India: translating widespread adoption into sustained economic impact. Indian government initiatives such as the IndiaAI Mission — a national program aimed at expanding computing capacity, supporting startups and accelerating AI adoption in public services — seek to address those gaps. However, the country’s price-sensitive market and infrastructure constraints have made monetization and large-scale deployment more complex than in developed economies.

“Given India’s size, it also risks forfeiting a vital opportunity to advance democratic AI in emerging markets around the world,” Altman wrote, warning that uneven access and adoption could concentrate AI’s economic gains in too few hands.

Altman also signaled that OpenAI plans to deepen its engagement with the Indian government, writing that the company would soon announce new partnerships aimed at expanding access to AI across the country. He did not provide details, but said the focus would be on widening reach and enabling more people to put AI tools to practical use.

The India AI Impact Summit is expected to draw a wide cross-section of global technology and political leaders, including Anthropic CEO Dario Amodei, Sundar Pichai of Google, and senior Indian business figures such as Mukesh Ambani and Nandan Nilekani. Political leaders including Emmanuel Macron, Sheikh Khaled bin Mohamed bin Zayed Al Nahyan, and Luiz Inácio Lula da Silva are also expected to attend, spotlighting India’s ambition to position itself as a central player in global AI debates.

For global AI firms, including OpenAI, the summit underscores how India’s vast user base is translating into growing influence over how the technology evolves.

OpenAI did not respond to a request for comment.

• Google has made a small but useful tweak to Android Auto
• Driving status can now be ascertained via Bluetooth
• This reduces the chances of mistakenly muting notifications

If you’re a regular Android Auto user, you may have noticed that there are a rather complicated set of rules around notifications being automatically blocked while you’re driving – but it seems Google has now sorted out one of the most frustrating issues.

As spotted by Android Authority, the option to have Do Not Disturb automatically enabled in Android Auto when motion is detected seems to have been tweaked. This setting was known for accidentally muting alerts on the phones of passengers as well as drivers.

Apple’s Mac Mini has been sitting quietly on desks all across the world, pushing through daily business and creative chores with the simplicity of a small powerhouse. In spirit, if not in design, it is right next to an equally little, yet quite different, machine known popularly as the Mini Mac, or Wondermac in some circles. This little device began life as a Maclock, a desk clock that looks exactly like the original 1984 Macintosh, complete with the instantly identifiable beige shell, a teeny-tiny screen bezel, and even a false floppy disk port, but what occurs next adds a lot of complexity to what was once a simple gimmick.

YouTuber This Does Not Compute disassembled the Maclock and then rebuilt it from the ground up, using a Raspberry Pi Zero 2 W, a chip small enough to fit discreetly inside the clock’s shell but powerful enough to tackle the work at hand. He combined it with a 2.8-inch Waveshare LCD panel, which is small enough to fit comfortably behind the curved plastic lens that attempts to simulate an old-school CRT monitor. The resolution is a respectable 640×480, which complements the old atmosphere perfectly. For power, he used a customized USB-C connection on the back, which was tuned to produce a clean 5 volts after some fiddling with the original charging circuit to bypass it. Then there’s a little heatsink that quietly keeps the Pi from overheating without creating any fan noise.

Sale

OIJ Maclock Retro Pixel Alarm Clock, 1984 Computer-Inspired USB Desk Clock with Pixel Display, Multiple…

• 【Retro 1984 PC Design】 This miniature ivory-colored ABS/PMMA alarm clock revives the charm of the early computer era. A perfect nostalgic piece…
• 【Multi-Mode Pixel Display】 The high-contrast LCD simultaneously displays time, calendar, date, and room temperature. A hidden pixel smiley…
• 【Adjustable Brightness for Comfort】 Gentle, warm backlighting in the morning and dim lighting at night protect your sleep. Simple bottom dial…

The front bezel required some custom work, which he solved by designing and 3D printing a black bracket to put the Waveshare display in place, cleaning up a bit of plastic in the process to ensure everything fit together smoothly. The original clock’s screws end up holding everything together. As an optional extra, you can sandwich the original’s old LCD layer between the lens and the Waveshare screen to maintain the real border look, albeit this will cost you a sliver of useful screen space. When all is said and done, the finished piece appears to be so beautifully put together that it could have come directly from the factory.

The real magic comes from the software, Raspberry Pi OS, notably the 32-bit version, which is stored on a fancy 32GB microSD card. After setting up the essentials like Wi-Fi and SSH, he installed the display drivers and adjusted the output to match the Mac case’s portrait orientation. The big event, however, is Mini vMac, a lightweight emulator that replicates ancient Macintosh computers in one fell swoop. Unfortunately for the project, newer versions were too hefty for the Pi Zero 2 W, so he reverted to an older release that he compiled from source. The laptop then boots directly into System 7 after loading a Mac ROM file and some system disk images, ready to run some vintage apps or simply sit there looking nice with one of those mesmerizing After Dark screen savers.

Wondermac operates as a headless machine, displaying its replicated desktop on that little screen, essentially serving as a constant desk ornament. It uses very little power and is extremely silent. Overall, it cost him roughly $100, depending on where you can obtain the Maclock and parts. He’s retained a repository of all the software phases on GitHub, from getting started to mucking around with compilation notes, though he admits the project isn’t seeing much active activity right now.

On today’s episode of You Asked: Can we give it a rest with pumping up the TV brightness? Why is Samsung adding Nvidia G-Sync support but not Dolby Vision? And what type of TV will work best in a room with wraparound windows?

Can we stop pumping up TV brightness?

@allclear4115 asks: I’m depressed about TVs nowadays.. Not because they’re too bright, but because they can’t be tamed at all. Rest in peace, dark room!

First off, I definitely understand the brightness fatigue. Everything seems to be “X times brighter than the previous generation” or “brightest ever!” The claimed Nits, the measurement for brightness, is all over the box and the marketing. It’s just everywhere you look, and it can be exhausting.

But I think one of the reasons we see brightness being pushed all over the place is that it’s one of the ways manufacturers can show the most improvement year after year.

Take OLED TVs, for example. The large panels we know now have been around for more than a decade, but they certainly weren’t always this bright. Even as few as four or five or six years ago, they weren’t hitting the levels they are now. And as we’ve talked about a lot, that overall brightness mainly helps with contrast against pure blacks of OLED and peak highlights having the sparkle that makes a viewing experience really special.

With a lot of Mini LED TVs, that increased brightness means being able to overpower ambient light in bright rooms and even sunlight. So brightness does bring some notable quality-of-life improvements.

But that doesn’t mean they aren’t tamed. In fact, some TVs have specific picture profiles for making a pleasurable viewing experience in dark rooms. Hisense, for example, has Theater Day and Theater Night modes. You can set one to have the brightness juiced up during the day to overpower ambient light. And at night, when you don’t want the TV lighting up the entire room, you can turn it down quite a bit.

It’s a similar story for a lot of TVs we talk about on this channel. While they can get very bright and reviewers love to talk about that, they also come with lots of settings that can be adjusted and tailored for what suits you best.

So again, I get where you’re coming from, and I’m sensing a little bit of sarcasm there too. I just don’t think TVs are too bright unless you want them to be. And personally, I’d rather have some room to spare on the bright end than finding my TV too dull when the sun’s out.

Why G-Sync but not Dolby Vision?

Parker Tyler asks: So they’ll pay the licensing for G-Sync, but they won’t for Dolby Vision?

The frustration from Samsung not supporting Dolby Vision continues, and I’m kinda torn on it. I’ve seen the improvements that Dolby Vision 2 brings, but I also have a Samsung Mini LED TV at home that I watch every day without Dolby Vision and have zero complaints.

And I’d imagine a lot of Samsung OLED owners feel the same way. Those QD-OLED panels on the high-end models bring a more colorful punch than most of the competition, and HDR 10 and 10 Plus are nothing to scoff at.

I’m just guessing that Samsung doesn’t see Dolby Vision as enough of an “upgrade” to put it in their TVs. And they sell tons of units every year. So a good chunk of the market also agrees.

For gaming, on the other hand, I would assume having Nvidia G-Sync gives more to that crowd than Dolby Vision does to the general audience. I’m also assuming they’ve done the research and see that adding that feature would sell more of their OLED TVs to gamers.

They’ve also doubled down with HDR 10 Plus Advanced, a new format that boasts great picture control through enhanced metadata. Sound familiar? Safe to say Dolby Vision isn’t coming to Samsung anytime soon. But at least they’re servicing the rest of the market with gaming upgrades.

What’s the brightest TV for a room with wraparound windows?

@YeCannyDaeThat asks: I have a really bright room with wraparound windows. Which is the brightest type of TV?

The brightest type of TV that you’ll find most available is Mini LED. And in that scenario with wraparound windows, especially if there’s a window behind the TV as well, you’ll definitely want to look at some of the brightest available.

TCL and Hisense make some really great options. Think TCL QM8K or the Hisense U8QG. You can also find these in some large sizes to really help the TV stand out among the light around it. And since those are 2025 models, you can probably get them for a really good price right now.

Sony’s Bravia 9 is another great pick for a bright room. It also does a great job holding its contrast at high luminance levels.

I haven’t seen it with my own eyes, but TCL also has a QM9K that looks insanely bright, rated to hit up to 6,500 nits with 6,000 dimming zones. It’s overkill in most scenarios, but maybe not in yours?

I also shouldn’t slight OLED options here, since there are a couple that I think would be suitable for bright rooms. Samsung’s S95F and its anti-glare screen do great battling reflections. LG’s G5 has also stepped up its game in terms of brightness compared to previous models.

You’ll have to see which fits your budget, but like I said before, I think if you can get one of those bright TVs in a bigger size, it’ll really help it overcome the light in your room.

Without a doubt, one of the hottest new startup accelerators in tech right now is Andreessen Horowitz’s Speedrun program. Launched in 2023, the accelerator has an acceptance rate of less than 1%. In a January blog post, the program said that over 19,000 startups pitched and fewer than 0.4% were accepted into the latest cohort. 

The program used to focus on gaming startups, then expanded into entertainment and media, and is now a “horizontal program,” Joshua Lu, the program’s general manager and a partner at a16z, told TechCrunch. Today, founders of any type of startup can apply, and the program runs for about 12 weeks in San Francisco. It once had a program in Los Angeles, but Lu said the focus will be on SF from now on. 

There are two cohorts a year, and around 50 to 70 startups are accepted into each. The program invests up to $1 million into each company, though the downside is that it’s a bit pricey. It typically invests $500,000 up front in exchange for 10% of the startup’s company via a SAFE note, and another $500,000 if the next round is raised within 18 months, at whatever terms agreed to by the other investors.

In comparison, Y Combinator typically takes a fixed 7% of the company for $125,000, with another $375,000 “invested on an uncapped MFN safe.”

Speedrun said its program is more “equity expensive” because of what it offers founders. It provides them with access to a16z’s advisory and business networks that assist with tasks like go-to-market, brand development, media strategy, and talent sourcing. Plus it offers the startups perks like $5 million in credits to vendors such as AWS, OpenAI, Nvidia, and Deel.

Given the high interest, and low acceptance rate, TechCrunch spoke to Lu for some tips on how startups can best stand out. The latest cohort began in January and will end in April with a Demo Day. Applications for the next cohort open in April, though it looks at off-season applications year-round, Lu said. 

Focus on the founding team  

Speedrun focuses on early-stage startups. Because of this, they really examine who is on the founding team and whether their skills complement each other, Lu said.

“That doesn’t mean one has to be technical and one has to be commercial and one has to be marketing,” Lu said. It means that “we prefer not to see any glaring holes in capabilities or interests. We want the founding team to be self-aware and for that to be part of the hiring plan.”  

They also like to see teams that have worked together before or have a shared history. 

“There are lots of things that a founding team has to navigate in their startup journey and having a bit of pattern recognition, being able to work with each other, knowing how to disagree and how to come out the other side of a disagreement, those are all things people on founding teams with shared histories have an easier time with, on average,” he continued. 

Even though AI has lowered the barriers to building software, it’s still incredibly helpful for a founding team to be technical, Lu said. At the same time, because AI has made it much faster to build and validate hypotheses and get a product out there, Lu said the Speedrun team likes to see when a startup already has a little bit of market validation or traction for their product.

“Speedrun as a program is really great at helping teams pour gasoline on a very small spark or fire,” he said. “We look for teams that have endeavored to build and try to show us that there’s a little spark we can fan the flames on.” 

Limit the market “theory”

Lu said one common mistake founders often make in the application process is spending too much energy talking about the market theory or why there is a defined problem and why their solution is the right one. “All of that may be true,” he said.

At the same time, he added, even the biggest, most successful tech companies faced unexpected blockades when they were young, sometimes even pivoting completely. What a company thinks it’s going to build at the beginning isn’t necessarily what will make it successful at the end.

“What we really want to hear about is why this founding team is really good together,” he continued, “why they’re a great founding team, the best possible founding team to solve this particular problem.” And then on top of that, any validation on the idea itself. 

It’s okay to use AI for the application, but…

Lu said the program encourages every founder to use AI to “clean up” their application. He said there is now no excuse for grammar errors or misspellings given the rising sophistication of AI tools. He also said AI can help founders sort out their thoughts, making them clearer, more concise, and more coherent. 

But if AI did all the work in explaining the startup, that may backfire. If a founder makes it to the next round, it will be a live video-call interview. “At that point, their live narration explanation skills are going to be put to the test,” he said. So founders should be prepared to talk cogently about their startup without the help of AI.

Only about 10% of founders make it to the video-call stage. There are typically two to three investors on the judging panel at a time.

After the live interview, the team typically conducts a few more screening calls with the founders, and then a final decision on the cohort is made.

Be greedy to network

There are, of course, other accelerator programs for startups to choose from. Lu said Speedrun itself was inspired by some of these other programs. 

Still, he said, this accelerator prides itself on giving founders access to a large, specialized operating team. In fact, he said the best teams that get the most out of the program are the ones most “greedy about getting exposure to the amazing people and programs” Speedrun has to offer. 

Lu listed off just a few points: a16z has around 600 people, and 10% of that staff is on the investment team, he said; everyone else is an operator who supports the companies the firm works with. As a result, founders in Speedrun will have access to experts who can help with marketing, banking, finance, management, and many other functions. So it helps to know who the startup wants to connect with and why. 

“We tell founders that come through the program, what you get out of Speedrun is what you put into it,” he said. “We think founders who want to take advantage of world experts in many different domains early in their startup journey would be really smart to choose us.”

Advice from a founder in the program

Founder Mohamed Mohamed, who is in the recent cohort, just announced a $5 million raise for his proptech startup Smart Bricks led by a16z’s Speedrun. He was attracted to the program because he said it stood out as one of the few “explicitly designed for co-founders working on frontier AI applications,” and he picked it because he wanted a program that would allow him to “stress-test an ambitious technical vision.”

Mohamed said he treated the application like an internal strategy memo rather than a pitch. “Instead of polishing buzzwords, we focused on clarity — the real problem, why it’s structurally hard, and why our team is unusually well-positioned to solve it,” he said. “We were explicit about what was working, what wasn’t, and where we needed help. I think that honesty and clear articulation of why this problem matters” is what helped the company in the application process.

He called the whole process “rigorous but refreshingly thoughtful,” and said it was designed to understand how founders think, not just what they have built so far. “The conversations went deep into product architecture, data strategy, and long-term ambition. It felt closer to a partner-level discussion than a typical accelerator interview, which was a strong signal for us,” he said.

His overall advice is to be “intellectually honest and precise.” For example, he said in his application he avoided “over-optimizing” for the sake of hyping up his company. “If you’re vague, derivative, or overly defensive about your idea, it shows quickly. Don’t try to sound bigger than you are; clarity about where you actually are is far more compelling than inflated narratives,” he said.

In the end, “Speedrun isn’t looking for perfect companies; they’re looking for founders who can reason clearly about complex problems and build with conviction,” he said. “Articulate the hard parts of what you’re doing and why they’re worth tackling. Depth beats polish every time.”

Correction, story originally misstated YC’s investment for its 7%. It has been corrected.

When trying to find a vegan meal delivery service, several factors should be considered to find your perfect fit. 

Prepared meals versus meal kits: Prepared meals come ready to heat in the microwave or in a pan on the stovetop. Meal kits, on the other hand, provide all the ingredients and instructions you’ll need to prepare the meals yourself. Though prepared meals are easier and take less time, meal kits may be good for those who like cooking or want to learn to cook.

Fresh versus frozen: On this list, we only have one meal delivery service that offers fresh meals you keep in the fridge for eating or heating: Thistle. All other prepared meal delivery services offer preprepared meals that are frozen. 

Advertisement

Dietary restrictions: If you have certain dietary restrictions, you’ll want to ensure that the service you choose caters to all of them. Some services are strictly vegan, while others are vegetarian. If you’re gluten-free, counting calories or carbs or want high-protein meals, those are also factors you should consider. 

Cost: Out of the services tested for this list, costs range from $6 to $18 per serving. Before choosing a meal delivery service, you will want to determine your budget and what you’re willing to pay for time saved and convenience. Also, are you willing to pay more for meals that are already prepared versus meal kits you make yourself?

Variety: Depending on your diet, only a certain number of meals may be offered per week. Each meal delivery service should allow you to view its menus a few weeks in advance, so you can take a look at how many meals you’ll have to choose from. If you plan to rely on this service for most of your meals, you’ll want to ensure it caters to your specific diet with a wider variety of dishes.

Number of servings: If you’re looking for a meal delivery service for a couple or family, check to see how many servings are offered. Meal kits are likely to offer a greater number of servings (around four to six), while prepared meals are better suited toward individuals.