Late last year, Google briefly took the crown for most powerful AI model in the world with the launch of Gemini 3 Pro — only to be surpassed within weeks by OpenAI and Anthropic releasing new models, s is common in the fiercely competitive AI race.
Now Google is back to retake the throne with an updated version of that flagship model: Gemini 3.1 Pro, positioned as a smarter baseline for tasks where a simple response is insufficient—targeting science, research, and engineering workflows that demand deep planning and synthesis.
The most significant advancement in Gemini 3.1 Pro lies in its performance on rigorous logic benchmarks. Most notably, the model achieved a verified score of 77.1% on ARC-AGI-2.
Advertisement
This specific benchmark is designed to evaluate a model’s ability to solve entirely new logic patterns it has not encountered during training.
This result represents more than double the reasoning performance of the previous Gemini 3 Pro model.
Google Gemini 3.1 Pro benchmark chart. Credit: Google
Beyond abstract logic, internal benchmarks indicate that 3.1 Pro is highly competitive across specialized domains:
Advertisement
Scientific Knowledge: It scored 94.3% on GPQA Diamond.
Coding: It reached an Elo of 2887 on LiveCodeBench Pro and scored 80.6% on SWE-Bench Verified.
Multimodal Understanding: It achieved 92.6% on MMMLU.
These technical gains are not just incremental; they represent a refinement in how the model handles “thinking” tokens and long-horizon tasks, providing a more reliable foundation for developers building autonomous agents.
Improved vibe coding and 3D synthesis
Google is demonstrating the model’s utility through “intelligence applied”—shifting the focus from chat interfaces to functional outputs.
One of the most prominent features is the model’s ability to generate “vibe-coded” animated SVGs directly from text prompts. Because these are code-based rather than pixel-based, they remain scalable and maintain tiny file sizes compared to traditional video, boasting far more detailed, presentable and professional visuals for websites and presentations and other enterprise applications.
Other showcased applications include:
Advertisement
Complex System Synthesis: The model successfully configured a public telemetry stream to build a live aerospace dashboard visualizing the International Space Station’s orbit.
Interactive Design: In one demo, 3.1 Pro coded a complex 3D starling murmuration that users can manipulate via hand-tracking, accompanied by a generative audio score.
Creative Coding: The model translated the atmospheric themes of Emily Brontë’s Wuthering Heights into a functional, modern web design, demonstrating an ability to reason through tone and style rather than just literal text.
Business impact and community reactions
Enterprise partners have already begun integrating the preview version of 3.1 Pro, reporting noticeable improvements in reliability and efficiency.
Vladislav Tankov, Director of AI at JetBrains, noted a 15% quality improvement over previous versions, stating the model is “stronger, faster… and more efficient, requiring fewer output tokens”. Other industry reactions include:
Databricks: CTO Hanlin Tang reported that the model achieved “best-in-class results” on OfficeQA, a benchmark for grounded reasoning across tabular and unstructured data.
Cartwheel: Co-founder Andrew Carr highlighted the model’s “substantially improved understanding of 3D transformations,” noting it resolved long-standing rotation order bugs in 3D animation pipelines.
Hostinger Horizons: Head of Product Dainius Kavoliunas observed that the model understands the “vibe” behind a prompt, translating intent into style-accurate code for non-developers.
Pricing, licensing, and availability
For developers, the most striking aspect of the 3.1 Pro release is the “reasoning-to-dollar” ratio. When Gemini 3 Pro launched, it was positioned in the mid-high price range at $2.00 per million input tokens for standard prompts. Gemini 3.1 Pro maintains this exact pricing structure, effectively offering a massive performance upgrade at no additional cost to API users.
Input Price: $2.00 per 1M tokens for prompts up to 200k; $4.00 per 1M tokens for prompts over 200k.
Output Price: $12.00 per 1M tokens for prompts up to 200k; $18.00 per 1M tokens for prompts over 200k.
Context Caching: Billed at $0.20 to $0.40 per 1M tokens depending on prompt size, plus a storage fee of $4.50 per 1M tokens per hour.
Search Grounding: 5,000 prompts per month are free, followed by a charge of $14 per 1,000 search queries.
For consumers, the model is rolling out in the Gemini app and NotebookLM with higher limits for Google AI Pro and Ultra subscribers.
Advertisement
Licensing implications
As a proprietary model offered through Vertex Studio inGoogle Cloud and the Gemini API, 3.1 Pro follows a standard commercial SaaS (Software as a Service) model rather than an open-source license.
For enterprise users, this provides “grounded reasoning” within the security perimeter of Vertex AI, allowing businesses to operate on their own data with confidence.
The “Preview” status allows Google to refine the model’s safety and performance before general availability, a common practice in high-stakes AI deployment.
By doubling down on core reasoning and specialized benchmarks like ARC-AGI-2, Google is signaling that the next phase of the AI race will be won by models that can think through a problem, not just predict the next word.
Considering that the Nintendo DS already has its own remake of Super Mario 64, one might be tempted to think that porting the original Nintendo 64 version would be a snap. Why you’d want to do this is left as an exercise to the reader, but whether due to nostalgia or out of sheer spite, the question of how easy this would be remains. Correspondingly, [Tobi] figured that he’d give it a shake, with interesting results.
Of note that is someone else already ported SM64 to the DSi, which is a later version of the DS with more processing power, more RAM and other changes. The reason why the 16 MB of RAM of the DSi is required, is because it needs to load the entire game into RAM, rather than do on-demand reads from the cartridge. This is why the N64 made do with just 4 MB of RAM, which is as much RAM as the ND has. Ergo it can be made to work.
The key here is NitroFS, which allows you to implement a similar kind of segmented loading as the N64 uses. Using this the [Hydr8gon] DSi port could be taken as the basis and crammed into NitroFS, enabling the game to mostly run smoothly on the original DS.
Advertisement
There are still some ongoing issues before the project will be released, mostly related to sound support and general stability. If you have a flash cartridge for the DS this means that soon you too should be able to play the original SM64 on real hardware as though it’s a quaint portable N64.
The seemingly accidental exposure of the identities of DHS personnel who crafted Immigration and Customs Enforcement’s mega detention center plan lands amid widespread public pushback against the expansion of ICE detention centers and the department’s brutal immigration enforcement tactics.
Metadata in the document, which concerns ICE’s “Detention Reengineering Initiative” (DRI), lists as its author Jonathan Florentino, the director of ICE’s Newark, New Jersey, Field Office of Enforcement and Removal Operations.
In a note embedded on top of an FAQ question, “What is the average length of stay for the aliens?” Tim Kaiser, the deputy chief of staff for US Citizenship and Immigration Services, asked David Venturella, a former GEO Group executive whom The Washington Post described as an adviser overseeing an ICE division that manages detention center contracts, to “Please confirm” that the average stay for the new mega detention centers would be 60 days.
Advertisement
Venturella replied in a note that remained visible on the published document, “Ideally, I’d like to see a 30-day average for the Mega Center but 60 is fine.”
DHS did not respond to a request for comment about what the three men’s role in the DRI project is, nor did it answer questions about whether Florentino had access to a PDF processor subscription that might have enabled him to scrub metadata and comments from the PDF before sending it to the New Hampshire governor. (The so-called Department of Government Efficiency spent last year slashing the number of software licenses across the federal government.)
The document itself says that ICE intends to update a new detention model by the end of September of this year. ICE says it will create “an efficient detention network by reducing the total number of contracted detention facilities in use while increasing total bed capacity, enhancing custody management, and streamlining removal operations.”
“ICE’s surge hiring effort has resulted in the addition of 12,000 new law enforcement officers,” the DHS document says. “For ICE to sustain the anticipated increase in enforcement operations and arrests in 2026, an increase in detention capacity will be a necessary downstream requirement.”
Advertisement
ICE plans on having two types of facilities: regional processing centers that will hold between 1,000 to 1,500 detainees for an average stay of three to seven days, and the mega detention facilities, which will hold an average of 7,000 to 10,000 people for an average of 60 days. It’s been referred to as a “hub and spoke model,” where the smaller facilities will feed into the mega ones.
“ICE plans to activate all facilities by November 30, 2026, ensuring the timely expansion of detention capacity,” the document says.
Beyond detention centers, ICE plans to buy or lease offices and other facilities in more than 150 locations, in nearly every state in the US, according to documents first reported by WIRED.
The errant comment in the PDF sent to New Hampshire’s governor is not the only issue the set of documents apparently had; according to the New Hampshire Bulletin, a previous version of an accompanying document, an economic impact analysis of a processing site in Merrimack, New Hampshire, referenced “the Oklahoma economy” in the opening lines. The errant document remains on the governor’s website, as of publication.
Advertisement
Across the country, ICE’s mega detention center projects have sparked controversy. ICE’s purchase of a warehouse in Surprise, Arizona, spurred hundreds to attend a city council meeting on the topic, according to KJZZ in Phoenix. In Social Circle, Georgia, city officials have pushed back against DHS’s proposal to build a mega center there, because officials say the city’s water and sewage treatment infrastructure would not be able to handle the influx of people.
Hackers are actively exploiting the CVE-2026-1731 vulnerability in the BeyondTrust Remote Support product, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) warns.
The security issue affects BeyondTrust’s Remote Support 25.3.1 or earlier and Privileged Remote Access 24.3.4 or earlier, and can be exploited for remote code execution.
CISA added it to the Known Exploited Vulnerabilities (KEV) catalog on February 13 and gave federal agencies just three days to apply the patch or stop using the product.
BeyondTrust initially disclosed CVE-2026-1731 on February 6. The security advisory classified it as a pre-authentication remote code execution vulnerability caused by an OS command injection weakness, exploitable via specially crafted client requests sent to vulnerable endpoints.
Proof-of-concept (PoC) exploits for CVE-2026-1731 became available shortly after, and in-the-wild exploitation started almost immediately.
Advertisement
On February 13, BeyondTrust updated the bulletin to say that exploitation had been detected on January 31, making CVE-2026-1731 a zero-day vulnerability for at least a week.
BeyondTrust states that the report from researcher Harsh Jaiswal and the Hacktron AI team confirmed the anomalous activity that they detected on a single Remote Support appliance at the time.
CISA has now activated the ‘Known To Be Used in Ransomware Campaigns?’ indicator in the KEV catalog.
For customers of the cloud-based application (SaaS), the vendor states the patch was applied automatically on February 2, so no manual intervention is needed.
Advertisement
Customers of the self-hosted instances need to either enable automatic updates and verify that the patch was applied via the ‘/appliance’ interface or manually install it.
For Remote Support, the recommendation is to install version 25.3.2. Privileged Remote Access users should switch to version 25.1.1 or newer.
Those still at RS v21.3 and PRA v22.1 are recommended to upgrade to a newer version before applying the patch.
Modern IT infrastructure moves faster than manual workflows can handle.
In this new Tines guide, learn how your team can reduce hidden manual delays, improve reliability through automated response, and build and scale intelligent workflows on top of tools you already use.
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
So, humanoid robots are nearing peak human performance. I would point out, though, that this is likely very far from peak robot performance, which has yet to be effectively exploited, because it requires more than just copying humans.
“The Street Dance of China” Turning lightness into gravity, and rhythm into impact.This is a head-on collision between metal and beats. This Chinese New Year, watch PNDbotics Adam bring the heat with a difference.
NASA’s Perseverance rover can now precisely determine its own location on Mars without waiting for human help from Earth. This is possible thanks to a new technology called Mars Global Localization. This technology rapidly compares panoramic images from the rover’s navigation cameras with onboard orbital terrain maps. It’s done with an algorithm that runs on the rover’s Helicopter Base Station processor, which was originally used to communicate with the Ingenuity Mars Helicopter. In a few minutes, the algorithm can pinpoint Perseverance’s position to within about 10 inches (25 centimeters). The technology will help the rover drive farther autonomously and keep exploring.
Corvus One for Cold Chain is engineered to live and operate in freezer environments permanently, down to -20°F, while maintaining full flight and barcode scanning performance.
I am sure there is an excellent reason for putting a cold storage facility in the Mojave desert.
The video documents the current progress made in the picking rate of the Shiva robot when picking strawberries. It first shows the previous status, then the further development, and finally the field test.
Data powers an organization’s digital transformation, and ST Engineering MRAS is leveraging Spot to get a full view of critical equipment and facility. Working autonomously, Spot collects information about machine health – and now, thanks to an integration of the Leica BLK ARC for reality capture, detailed and accurate point cloud data for their digital twin.
ASTORINO is a modern 6-axis robot based on 3D printing technology. Programmable in AS-language, it facilitates the preparation of classes with ready-made teaching materials, is easy both to use and to repair, and gives the opportunity to learn and make mistakes without fear of breaking it.
What does it mean to build a humanoid robot in seven months, and the next one in just five? This documentary takes you behind the scenes at Humanoid, a UK-based AI and robotics company building reliable, safe, and helpful humanoid robots. You’ll hear directly from our engineering, hardware, product, and other teams as they share their perspectives on the journey of turning physical AI into reality.
This IROS 2025 keynote is from Tim Chung who is now at Microsoft, on “Catalyzing the Future of Human, Robot, and AI Agent Teams in the Physical World.”
Advertisement
The convergence of technologies—from foundation AI models to diverse sensors and actuators to ubiquitous connectivity—is transforming the nature of interactions in the physical and digital world. People have accelerated their collaborative connections and productivity through digital and immersive technologies, no longer limited by geography or language or access. Humans have also leveraged and interacted with AI in many different forms, with the advent of hyperscale AI models (i.e., large language models) forever changing (and at an ever-astonishing pace) the nature of human-AI teams, realized in this era of the AI “copilot.” Similarly, robotics and automation technologies now afford greater opportunities to work with and/or near humans, allowing for increasingly collaborative physical robots to dramatically impact real-world activities. It is the compounding effect of enabling all three capabilities, each complementary to one another in valuable ways, and we envision the triad formed by human-robot-AI teams as revolutionizing the future of society, the economy, and of technology.
This GRASP SFI talk is by Chris Paxton at Agility Robotics, on “How Close Are We To Generalist Humanoid Robots?”
Advertisement
With billions of dollars of funding pouring into robotics, general-purpose humanoid robots seem closer than ever. And certainly it feels like the pace of robotics is faster than ever, with multiple companies beginning large-scale deployments of humanoid robots. In this talk, I’ll go over the challenges still facing scaling robot learning, looking at insights from a year of discussions with researchers all over the world.
This week’s CMU RI Seminar is from Jitendra Malik at UC Berkeley, on “Robot Learning, With Inspiration From Child Development.”
For intelligent robots to become ubiquitous, we need to “solve” locomotion, navigation and manipulation at sufficient reliability in widely varying environments. In locomotion, we now have demonstrations of humanoid walking in a variety of challenging environments. In navigation, we pursued the task of “Go to Any Thing” – a robot, on entering a newly rented Airbnb, should be able to find objects such as TV sets or potted plants. RL in simulation and sim-to-real have been workhorse technologies for us, assisted by a few technical innovations. I will sketch promising directions for future work.
Cable TV providers have spent the past decade losing tens of millions of households to streaming services, but companies like Charter Communications are now slowing that exodus by bundling the very apps that once threatened to replace them.
Charter added 44,000 net video subscribers in the fourth quarter of 2025, its first growth in that count since 2020, after integrating Disney+, Hulu, and ESPN+ directly into Spectrum cable packages — a deal that grew out of a contentious 2023 contract dispute with Disney. Comcast and Optimum still lost subscribers in the quarter, though both saw those losses narrow.
Charter’s Q4 numbers also got a lift from a 15-day Disney channel blackout on YouTube TV during football season, which drove more than 14,000 subscribers to Spectrum. Charter has been discounting aggressively — video revenue fell 10% year over year despite the subscriber gains. Cox Communications launched its first streaming-inclusive cable bundles last month, and Dish Network has yet to integrate streaming apps into its packages at all.
Advantest Corporation disclosed that its corporate network has been targeted in a ransomware attack that may have affected customer or employee data.
Preliminary investigation results revealed that an intruder gained access to certain parts of the company’s network on February 15.
Tokyo-based Advantest is a global leader in testing equipment for semiconductors, measuring instruments, digital consumer products, and wireless communications equipment.
The company employs 7,600 people, has an annual revenue of more than $5 billion, and a market capitalization of $120 billion.
On February 15, the firm detected unusual activity in its IT environment, prompting a response in accordance with incident response protocols, including the isolation of affected systems.
Advertisement
As part of its response, the company contracted third-party cybersecurity specialists to help isolate the threat and investigate its impact.
“Preliminary findings appear to indicate that an unauthorized third party may have gained access to portions of the company’s network and deployed ransomware,” Advantest states.
“If our investigation determines that customer or employee data was affected, we will notify impacted persons directly and provide guidance on protective measures.”
Currently, no data theft has been confirmed, but Advantest noted that this may change as more information emerges from the ongoing investigation.
Advertisement
Should customers or staff be determined to be impacted, Advantest will notify them directly and provide instructions on mitigating the associated risks.
At the time of writing, no ransomware groups have claimed the attack on the Japanese tech giant.
BleepingComputer has contacted Advantest directly to request more details about the attack, but we have not heard back by publishing time.
Multiple Japanese companies have been the target of cyberattacks recently, as several high-profile entities suffered data breaches and operational disruptions. Notable examples include Washington Hotel, Nissan, Muji, Asahi, and NTT.
Advertisement
Advantest says that the investigation continues and that it will provide updates on the incident when new details emerge.
Modern IT infrastructure moves faster than manual workflows can handle.
In this new Tines guide, learn how your team can reduce hidden manual delays, improve reliability through automated response, and build and scale intelligent workflows on top of tools you already use.
Cybernews found misconfigured database in “Video AI Art Generator & Maker” app
Leak exposed 8.27m media files, including 2m private user photos and videos
Developers secured database after disclosure; similar flaws seen in another Codeway app
Yet another misconfigured database leaking sensitive user data was found, but this one is even more worrying since the data being leaked is – user-uploaded photos and videos.
Researchers from Cybernews recently discovered an Android app called “Video AI Art Generator & Maker” contained a misconfigured Google Cloud storage bucket which was accessible to anyone who knew where to look.
In total, more than 1.5 million user images, and more than 385,000 videos were stored in the bucket. Furthermore, it stored more than 2.87 million AI-generated videos, more than 386,000 AI-generated audio files, and more than 2.87 million AI-generated images.
Several vulnerable apps
The app offered AI-generated makeovers for photos and videos – something that is particularly popular these days. It was launched in mid-June 2023 and according to Cybernews, has been storing the multimedia people upload since then.
So, in total, the exposed bucket contained 8.27 million media files, 2 million of which were private, user-uploaded content.
The app was allegedly developed by Codeway Dijital Hizmetler Anonim Sirketi, a private company registered in Turkey, but we could not find this app on the Play Store, or the developer’s official website. Codeway’s dedicated Play Store page shows only three apps.
Advertisement
However, the official website does demonstrate a different app, called Chat & Ask AI, and this one also had a misconfigured backend using Google Firebase. In early February 2026, an independent researcher found that this app, one of the most popular ones in its category, exposed 300 million messages tied to 25 million users.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Still, Cybernews said it managed to get in touch with the developers, who secured the Video AI Art Generator & Maker database soon after.
“This data leak shows how some AI apps prioritize fast product delivery, skipping crucial security features, such as enabling authentication for the critical cloud storage bucket used to store user data, including images and videos,” the researchers explained.
We’ve been rigorously testing work-from-home gear for years—even prior to the Covid-19 remote work boom—and that includes dozens of office chairs and desks. Branch furniture has made standouts that are highlighted in our guides over and over again. Its Presidents’ Day deals have been extended, bringing some of the better discounts we’ve seen on essentials we’ve tested like chairs and desks.
Check out our other deals coverage for additional discounts on gear we’ve tried and would recommend to a friend.
Branch Ergonomic Chair Pro for $449 ($50 off)
Photograph: Julian Chokkattu
Photograph: Julian Chokkattu
Photograph: Julian Chokkattu
Branch
Ergonomic Chair Pro
This price matches the best we usually see for our very favorite office chair. Out of the dozens we’ve tried, this chair strikes the best balance of features for the price. It’s comfortable, adjustable, and easy to dial in so you can get your perfect ergonomic fit. It also has a solid warranty and isn’t too terribly expensive compared to similar chairs. There are different fabric finishes and colors to choose from, all of which are on sale right now.
Advertisement
Branch Ergonomic Chair for $323 ($36 off)
The best budget office chair is even more affordable right now thanks to this deal. It’s easy to assemble, it has some adjustable elements, it’s comfortable and breathable, and it looks nice with or without the optional headrest. The upholstery is available in several colors, though the fabric does pill and attract pet hair. We still think this is a chair worth checking out if you’re on a tight budget.
Branch Four Leg Standing Desk for $854 ($95 off)
Photograph: Julian Chokkattu
Branch
Four Leg Standing Desk
This is editor Julian Chokkattu’s favorite desk he’s tried. At first glance, it looks like a standard desk, but it’s actually a standing desk that can be raised or lowered with the little control panel. Assembly was easy, the controls are simple, and the shape is elegant. If you want a desk that looks great no matter how tall it is, this is worth checking out, especially at this price.
Branch Duo Standing Desk for $494 ($55 off)
Photograph: Julian Chokkattu
We like this compact, affordable standing desk, which gets you a lot of value for how little you’ll pay. It’s compatible with a lot of add-ons, and the paddle controls are easy to use. There’s even a preset mode so you can press the paddle twice to raise it to your preset height. This desk is compact, but if you don’t need a ton of room for your working setup, it’s a good option even at full-price. (Luckily, right now, you can snag it for less.)
Matt Oppenheimer led Remitly for nearly 15 years as co-founder and CEO. He announced this week that he’s moving into the chairman role. (Remitly Photo)
Build with intentionality. Lead with authenticity. Prioritize customers over your ego. And focus on the problem you’re solving — with flexibility on the solution.
That’s part of the playbook Matt Oppenheimer followed as he helped grow a three-person Techstars Seattle startup into one of the world’s leading remittance platforms.
After nearly 15 years leading Remitly as CEO, Oppenheimer announced Wednesday that he’s stepping down as CEO and moving to a board chair role. He’s passing the baton to Sebastian Gunningham, a longtime tech and finance leader who previously led Amazon’s marketplace and payments businesses.
“I feel wonderful, honestly,” Oppenheimer told GeekWire on Thursday. “One thing that has always driven me from the moment I started the business 15 years ago is impact and purpose and doing things with a sense of intentionality. And I feel like that’s how we’ve done this succession planning.”
The Remitly story began more than a decade ago after Oppenheimer had just returned from Kenya, where he was working for Barclays and realized how hard it was for families to send and receive money overseas.
Advertisement
He teamed up with co-founders Josh Hug and Shivaas Gulati, navigating an early pivot before landing product-market fit and raising around $400 million. The company went public in 2021 at a valuation of nearly $7 billion.
Remitly’s mobile technology lets people send and receive money across borders, eliminating many of the forms, codes, and in-person agents traditionally associated with international transfers. It’s used by more than 9 million people. The company reported revenue of $442.2 million in Q4, up 26% year-over-year, and had its first full year of GAAP profitability in 2025.
We spoke to Oppenheimer about lessons learned from Remitly’s journey and his advice for entrepreneurs. Here are some key takeaways.
Fall in love with the problem, not your product
Oppenheimer remembers the frustration he saw and felt watching families struggle to send money across borders. That sparked the idea for Remitly. The key, he says, was locking onto that problem — not any one product idea. The danger is when founders apply their grit in the wrong place.
Advertisement
“If they channel that perseverance in the wrong area — the product or trying to force something into existence that customers don’t care about — they fail,” he said. “They run out of time, energy, or money.”
Remitly’s booth at Southcenter Mall served as a key customer feedback mechanism. (Remitly Photo)
Get close to customers
In the early days, Remitly set up a booth at Southcenter Mall near Seattle outside a legacy remittance location, complete with scotch-taped signage.
Oppenheimer referenced a phrase from Airbnb co-founder Brian Chesky: “find marketing channels that don’t scale.” The goal wasn’t growth, but rather insight.
They learned why customers weren’t using Remitly. That feedback drove a big pivot from mobile wallets to cash pickup, bank deposit, and door-to-door delivery.
“We had to follow customers,” Oppenheimer said. He added: “If we would have been stubborn about only doing mobile wallets — that’s what our pitch said — then we would have failed.”
Advertisement
Define culture as behaviors — and keep rewriting it
Oppenheimer said many companies stop at a short list of vague values. “Culture is how people in a company or institution interact to deliver for their customers,” he said.
Before Remitly launched its product, the founding team did an offsite to define the culture on a whiteboard. Early values like “relationships” were well-intentioned but too broad. Remitly refreshed its values every six months and now every couple of years, evolving them into more specific behaviors such as “be a compassionate partner,” “lead authentically,” and “constructively direct.”
Customer centricity sits at the top as the single overarching value. Oppenheimer said the test is whether values show up in concrete decision-making: “Once you’ve got it defined, [you embed it] into the interview process and the performance review process.”
Remitly co-founders Josh Hug, Matt Oppenheimer and Shivash Gulati. (Remitly Photo)
Find complementary co-founders
Oppenheimer said Remitly wouldn’t exist without his co-founders, pointing to Hug’s product skills and Gulati’s engineering chops.
“It’s important for all founders to surround themselves with complementary skills and respect those skills deeply,” he said.
Advertisement
In the very early days, his own contribution was often clearing obstacles: money transmission licenses, office leases, even taking out the trash. “My job was to help them build,” he said. Oppenheimer stressed the importance of shared values but different strengths.
Raise more capital than you think you need
Remitly raised hundreds of millions of dollars on its way to an IPO across multiple rounds. None were easy.
“It requires getting a lot of no’s,” Oppenheimer said. “It requires that grit, tenacity and perseverance that is critical for any entrepreneur to be successful.”
He advised treating fundraising as a two-way conversation, not a one-sided pitch. “Investors can sniff desperation,” he said. Make sure investors are asking the right questions, and think about whether you want them on your board.
Advertisement
When the partner is right, Oppenheimer leans toward raising a bit more. “Things always take a little bit longer than you imagine,” he said. For companies pursuing bold visions, “if you’ve got the right partner, you can raise enough capital, then it’s worth the dilution to be able to make progress against accomplishing that vision.”
Oppenheimer on his first trip to the Philippines as Remitly’s CEO. (Remitly Photo)
Treat your own growth like a product, with reviews and roadmaps
As he focused more on management, Oppenheimer built a formal process for his own development as CEO, especially as Remitly grew from a handful of people to more than 3,000.
He started asking each new investor who joined Remitly’s board to run his performance review. “I’d like you to talk to all other board members. I’d like you to talk to my leadership team,” he’d tell them. “And then I’d like your insights.”
He turned that input into a written development plan, shared it with the company, and then found coaches and mentors to help him work on specific gaps. “It took a lot of intentionality to grow as a leader,” he said.
That work continues in his new role as chairman. “After mission and purpose, my second biggest motivator for me personally is growing as a human,” he said. “That’s what I’ve loved about the journey, and it continues in this next role.”
Advertisement
Don’t underestimate the role of community
Seattle is a huge part of Remitly’s story. Techstars Seattle helped launch Remitly (back when it was called Beamit Mobile); talent from the region’s tech ecosystem helped scale it.
“The talent we’ve been able to recruit from some of the largest technology companies has been foundational,” Oppenheimer said. With fewer growth-stage companies in the city than in some other hubs, he believes Remitly could attract people looking to join a mission-driven startup with scale ambitions.
Last year the company moved into a new headquarters in downtown Seattle. Oppenheimer said he and Remitly remain committed to Seattle, noting that he wants to make sure “that’s the case for the next decade to come.”
The summit is the fourth in an annual series that began in UK in 2023, before moving to Korea, and then France.
Global business leaders and heads of states all flocked to New Delhi this week to attend India’s AI Impact Summit 2026, which just concluded today (20 February).
The summit is the fourth in an annual series that began in UK in 2023, before moving to Korea, and then France last year.
India has huge aspirations to become a leader in AI – and overall, the billions in investments announced from Big Tech leaders this past week could mean that the fast-growing economy might be on the right track.
Advertisement
But as The New York Times puts it: “India brims with tech talent but not the companies that command it.” While slow to develop on the technology itself, India offers a very large AI user base (100m weekly ChatGPT users alone are from India) and huge under-employed workforce.
Outside the summit, however, disorganisation was rife. Blocked roads forced delegates to walk kilometres to reach the summit and wait in long queues. Meanwhile, dozens of New Delhi’s poorest accused state leaders of forcibly displacing their makeshift homes to ‘beautify’ the streets for the incoming international guests.
Though as that went on, AI leaders including OpenAI, Microsoft and Nvidia made some big announcements.
Microsoft’s $50bn ‘global south’ pledge
Microsoft said it is on its way to invest $50bn in the “global south” – a term used to refer to the world’s developing countries. Research from the company finds that AI usage in the “global north” is roughly twice that of the “global south”. The company said that the investment will be used to help increase AI usage in the region.
Advertisement
The Bill Gates-founded company wants to build the infrastructure needed for better AI diffusion in the country, help develop multilingual and multicultural AI capabilities, enable local AI innovations, and be able to measure AI diffusion better to guide future policies and investments in the tech.
Last year, the company announced plans for around $17.5bn worth of AI investments in India.
Adani adds $100bn to AI data centre pot
Billionaire business mogul Gautam Adani’s company, the Adani Group, announced direct investments of $100bn to create the “world’s largest integrated data centre platform” in India.
The new investment adds to AdaniConnex’s already existing 2GW national data centre, and expands it to a 5GW target. These AI data centres will be powered with renewable energy, the company claimed.
Advertisement
Its plans for the mega data centre comes with existing partnerships with Google and Microsoft. The company said it is discussing plans for more large-scale campuses in India with other major players.
According to the Adani Group, the $100bn investment is expected to generate an additional $150bn across manufacturing, advanced electrical infrastructure and sovereign cloud platforms in the country by 2035. Together, it projects to create a $250bn AI infrastructure ecosystem in India over the decade.
Telco leader drops $110bn for compute
Indian telecommunications giant Reliance Industries and Jio – its digital business – announced 10trn rupees (around $110bn) in new investments to build AI computing infrastructure in the country.
Owner Mukesh Ambani said that the investment would fund what he described as India’s sovereign compute infrastructure, which would include multi-gigawatt-scale data centres, a nationwide edge computing network and new AI services integrated with Jio.
Advertisement
“This is not speculative investment, this is patient capital to build India,” Ambani said at the summit in New Delhi.
OpenAI becomes a TCS customer
Tata Consultancy Services (TCS) announced that OpenAI will become its first customer in its recently announced data centre business, Hypervault, with an initial commitment of 100MW of AI capacity. The capacity, it said, can be eventually scaled up to 1GW.
The project is a part of OpenAI’s Stargate venture, a $500bn privately-funded initiative to build AI data centres across the globe.
Alongside infrastructure, the partnership will also deploy ChatGPT Enterprise across parent company Tata Group’s other subsidiaries over the next several years.
Advertisement
“Through OpenAI for India and our partnership with the Tata Group, we’re working together to build the infrastructure, skills, and local partnerships needed to build AI with India, for India, and in India, so that more people across the country can access and benefit from it,” OpenAI CEO Sam Altman claimed.
L&T teams up with Nvidia
Larsen & Toubro (L&T), on the other hand, claimed to be building India’s “largest gigawatt-scale AI factory” using Nvidia’s AI infrastructure, which includes its GPUs, CPUs networking and accelerated storage platforms.
The venture will scale Nvidia GPU clusters at the company’s data centres in Chennai and Mumbai.
“AI is driving the largest infrastructure buildout in human history – everyone will use it, every company will be powered by it and every country will build it,” said Jensen Huang.
Advertisement
Its venture with L&T is “enabling AI factories at national scale”, he added, “ready to serve global and domestic AI demand”.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
-In-Use-Reviewer-Photo-(no-border)-SOURCE-Julian-Chokkattu.jpg)
-front-Reviewer-Photo-SOURCE-Julian-Chokkattu-(no-border).jpg)
