Second is advertiser tools. If you’re a small business, you’re not thinking about all the queries people are going to type in. AI is great at figuring out which keywords to use, what’s the optimal creative, and generating all of that.

The third piece is the most nascent: ads in new experiences. The general philosophy we have is to build a great consumer product, then figure out monetization. Because the business is so strong and healthy, that’s a luxury we have.

What have you learned from experiments around ads in AI Mode?

Ads are always separate from organic results and clearly labeled. If we don’t think any ad is relevant, we don’t show any ads. Probably the biggest principle of all is that ads should be useful.

What [ads in AI Mode] have shown is mostly intuitive things. If it’s relevant, a user will click on it. If not, they won’t.

At Davos, Demis Hassabis said that Google has no plans to bring ads to Gemini. How are you thinking about it now?

The reason we focused on ads in AI Mode and AI Overviews is because we see them as extensions of the Search experience. It’s the most natural place for us to do initial experiments.

I would expect that the learnings that we get from ads in AI Mode would likely carry over to what we might want to do in the Gemini app down the road. We’re able to get all those learnings within a context and a construct that users are already aware of ads. It’s an odd thing to say, but our research shows that users actually like ads within the context of Search. Over time, we’ll figure out what makes sense in the Gemini app.

So you guys aren’t ruling out ads in Gemini completely?

No, we’re not ruling them out. It’s just not where we’ve been focusing.

Gemini is a massive product now—it’s grown quite fast. OpenAI is already trying out ads in ChatGPT. What makes you think Gemini isn’t ready for ads?

We’re super happy with how well it’s growing. I would say it’s more of a prioritization question—what’s the right area to focus right now?

Do you think OpenAI introduced ads too early?

It’s hard for me to say. I think it will really depend on how they do it. What we’ve seen is when we do ads right, it’s accretive to the product experience. But the really critical thing is to do it right, and that means relevance, quality, and not putting ads where users don’t want to see ads.

That’s hard to do. We have over 20 years of experience learning how to do that. It’s less a question of timing and more a question of doing it right, and in a way that’s respectful of users.

Google recently launched Personal Intelligence in Gemini and AI Mode. I imagine advertisers would love to get their hands on that data. How are you thinking about that?

Personal Intelligence is incredibly useful. For example, I was skiing and couldn’t see well through my goggles—it was cloudy out. I asked AI Mode a fairly vague question about what lens I needed for the conditions. From my email, it was able to say which resort I was staying at, which mountain I’d be skiing, and what the weather would be. It also pulled in a receipt my wife had forwarded me for my goggles—turns out she’d bought me an extra lens. It’s like subtle magic.

The Australian-founded collaboration software provider wants to become ‘an AI-first company’.

Collaboration software provider Atlassian is to cut around 1,600 roles, amounting to 10pc of its workforce, its CEO and co-founder Mike Cannon-Brookes announced in a written blogpost yesterday (11 March).

Cannon-Brookes said the layoffs were necessary to “self-fund further investment in AI and enterprise sales, while strengthening our financial profile”, with an aim to change the way the Australian-founded company works and reorganise to move faster.

He said that although Atlassian fundamentally believes that people and AI create the best outcomes, and its approach is not for AI to replace people, “it would be disingenuous to pretend AI doesn’t change the mix of skills” needed or “the number of roles required in certain areas”.

“This is primarily about adaptation. We are reshaping our skill mix and changing how we work to build for the future,” he said.

Cannon-Brookes said “a thoughtful and incredibly thorough approach to determining impacted roles” was used to make structural and organisational changes, with a focus on retaining employees with the necessary skills for “an AI-first company” – namely “strong performers, graduates”, and those “with transferable skills”.

All Atlassian employees were due to receive an email update on their individual and regional statuses within 20 minutes of the post going live.

The CEO’s blogpost, which also included a video message to his staff, detailed the compensation arrangements for impacted employees, which are subject to local laws and consultation requirements.

Sydney-headquartered Atlassian was founded in 2002 by Cannon-Brookes and Scott Farquhar, who each have a current net worth of around $7bn, according to Forbes.

Its products include collaboration, communication and project management applications such as Trello, Jira, Confluence and Bamboo, and it has employees across 14 countries.

In a quarterly financial letter to shareholders last month, Cannon-Brookes said Atlassian was “building a bloody great business” and “had a fantastic Q2”, and that he was “convinced AI is great for Atlassian”.

Headline figures from the letter noted that the company has 350,000 customers and revenue of $1.6bn, up 23pc year-on-year.

However, its shares prices have been affected by the software slump brought on by the emerging threat of AI tools, as noted by Forbes.

So far this year, companies such as Amazon and Block have made significant layoffs in tandem with plans to spend more on AI.

Lat September, Atlassian announced plans to acquire productivity tech start-up DX and The Browser Company.

In 2023, the company cut 500 roles in a reorganisation move.

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.

Caller identity platform Truecaller recently launched a new feature that lets one person become an admin of a family group, get alerts about fraud calls received by other members, and even end a call on their behalf if they suspect a family member might get scammed.

The company, which has over 450 million users, first launched the feature in December in a handful of countries like Sweden, Chile, Malaysia, and Kenya. Truecaller said that after seeing promising results, it decided to roll it out worldwide, including in India, the company’s biggest market. The feature is free, and users can create groups even if they are not on a paid Truecaller plan.

With this feature, the tech-savvy member of a family or friends group can become the admin of an up to five-member group. Once the other members join the group, the admin can get alerts about potentially fraudulent calls those members receive. If the admin believes that the call could harm the member, they can remotely end the call as well. While the admin can get alerts for fraud calls when a member is using iOS or Android, they can only end calls for members on Android.

On Android, members can also grant permission to the admin to detect real-time activity such as walking or driving, battery level, and phone sound settings (to check if the phone is in silent mode). Truecaller said this is helpful for admins to keep tabs on elderly members and to only call them when they are not walking or driving.

The admin can also block certain numbers and international calling codes, and share a blocklist with group members.

Truecaller noted that the admin can’t see the non-spam call history or SMS history of group members.

“I think, unfortunately, all of us know somebody or another in our families or friends who have been impacted by fraud,” Kunal Dua, Chief Product Officer, at Truecaller, told TechCrunch over a call. “In that sense, it’s a fundamental shift for Truecaller in terms of what we’ve been focusing on as a problem,” he added.

Last year, Truecaller introduced a voicemail feature for Indian users featuring an AI assistant that listens for calls when a user is unavailable and provides a summary of the transcript. The company is exploring a similar AI approach for family protection to potentially alert the admin about what kind of fraud call a group member is receiving.

The company is also exploring using AI to screen calls and automatically disconnect them when certain words associated with scams are detected, such as “digital arrest” — a tactic in which perpetrators impersonate law enforcement officers to extract information or money from call recipients.

In India, scam calls have risen over time and caused financial losses across the country. Truecaller said that it identified over 7.7 billion fraud calls last year. Indian authorities have launched multiple initiatives, including a controversial policy called SIM binding that could hamper the working of apps like WhatsApp and Telegram.

Truecaller is facing headwinds. Its stock has dwindled by over 80% in the last 12 months. During its Q4 2025 report, the company said that its EBDITA– a measure of operating profitability — dipped 49% year-on-year, with ad revenue declining 31%. The company is also facing challenges from India’s Caller Name Presentation (CNAP) system, which displays the name of the caller as registered with their phone carrier. Truecaller has maintained that just displaying a caller’s name won’t reduce spam calls, arguing that its platform goes further by offering community-based reports.

“In India, there has been much talk about the imminent rollout of CNAP,” Truecaller CEO Rishit Jhunjhunwala said during the Q4 2025 earnings call. “CNAP is partially rolled out, and so far, the impact on our user growth is limited. As we have said in the past, we expect that CNAP might have some impact on user growth, but that remains to be seen as CNAP reaches a full rollout.”

Continued Jhunjhunwala, “Our focus continues to be on delivering a superior product, and as you are aware, the consumer can choose to have CNAP and Truecaller in parallel, where we provide a lot more information and a lot more context and various other solutions, for the consumer.”

The newly verified X account for Iran’s supreme leader could be putting the company on the wrong side of US sanctions, according to a watchdog group. The Tech Transparency Project, which last month published a report on X granting premium perks to sanctioned officials in Iran, now says that the verified account for the country’s new leader raises fresh questions about the issue.

The TTP notes that the X account for Iran’s new supreme leader, Mojtaba Khamenei, appears to be paying for an X premium subscription despite being on the US government’s list of sanctioned individuals since 2019. As the group points out, the Iran-based account was created this month and currently bears a blue checkmark, which typically indicates the account holder is paying for a subscription.

The account belonging to Mojtaba Khamenei has been boosted by other state-linked accounts in Iran, including the one that previously belonged to Khamenei’s father. That account has had a gray checkmark, which indicates it belongs to a verified government official. Verified accounts on X are rewarded with extra visibility on the platform, along with other perks. The younger Khamenei’s verified account has already gained more than 20,000 new followers in the hours since TTP first posted about it.

“The new Supreme Leader’s account is just the latest account for a sanctioned entity apparently paying X for premium services,” TTP director Katie Paul said in a statement to Engadget. “TTP has identified dozens of accounts, many linked to designated terrorists, that subscribed to X premium over the past three years. What’s more concerning than the blatant disregard for U.S. sanctions law is the fact that Musk’s companies have a contract with the Pentagon while X is actively profiting from U.S. adversaries.”

As Paul notes, this isn’t the first time TTP has raised questions about whether X is running afoul of US sanctions via its premium service. In 2024, the group published a report noting that X was accepting paid verification from more than two dozen sanctioned individuals and groups. The company said at the time that it had a “a robust and secure approach in place for our monetization features.”

X didn’t immediately respond to a request for comment. The company removed blue checks from a handful of Iran-based accounts flagged by TTP last month following reporting from Wired.

Google Maps just released the most significant navigation update in over a decade, and Gemini is at the center of it all. Two new features take center stage: a conversational tool called Ask Maps and a revised driving mode dubbed Immersive Navigation, which finally incorporates super-realistic 3D views with turn-by-turn directions.

Ask Maps replaces the old search bar with something far more powerful. Users hit this icon near the top of the app to begin asking actual inquiries in clear language. Gemini handles the rest, gathering data from all of the locations, ratings, and user tips contained in Maps. So, if you live in San Francisco and want to travel to Muir Woods for breakfast along the way, simply ask the question, and the system will be developed. It will provide you with a detailed route with stop-by-stop directions, as well as suggestions based on previous searches or remembered preferences… such as vegetarian options. Just ask follow-up questions, and the plan will be updated on the fly. If you want, recommendations will take you directly to directions or allow you to save places to lists without having to navigate. Ask Maps is available today in the United States and India for Android and iOS users, with desktop support coming soon.

Google Pixel 10a – Unlocked Android Smartphone – 7 Years of Pixel Drops, 30+ Hours Battery, Camera Coach…

• Google Pixel 10a is a durable, everyday phone with more[1]; snap brilliant photography on a simple, powerful camera, get 30+ hours out of a full…
• Unlocked Android phone gives you the flexibility to change carriers and choose your own data plan; it works with Google Fi, Verizon, T-Mobile, AT&T…
• Pixel 10a is sleek and durable, with a super smooth finish, scratch-resistant Corning Gorilla Glass 7i display, and IP68 water and dust protection[4]

Immersive Navigation completely overhauls the way you actually follow directions. You can now see the road you’re traveling in 3D, rather than just lines and icons. Buildings, overpasses, and road landscape seem with realistic depth because Gemini models evaluate Street View images and aerial shots to determine what’s crucial to show. See-through structures protect things from becoming congested, and when it comes down to it, lane markers, crosswalks, traffic signals, stop signs, and other elements are always visible when needed. The map scales up and down to provide a clear glimpse of what’s coming up, avoiding last-minute panics.

Voice instructions keep up with the upgrade and sound much more natural today. So, on the highway, it might instruct you to pass the next exit and then take the following one instead than simply counting down the distance. Alternate routes put out the advantages and cons for you, so you get a longer travel with less traffic or no tolls, and you can decide whether it’s worth the trade-off. As you get closer to your location, Street View shows you entrances and parking areas to help you arrive stress-free.

Google claims that this is the biggest shift in driving direction since the app’s inception, and early feedback suggests that it significantly reduces confusion in congested cities or unfamiliar areas. Both updates, for Immersive Navigation (which will be available on additional devices in the coming months, including CarPlay, Android Auto, and built-in auto systems) and Ask Maps, will be available in the United States and India. Android and iOS phones receive the initial peek, however full coverage will take some time.

Eighteen months after selling his startup to chipmaker AMD for $665 million, Finnish entrepreneur Peter Sarlin has left his role as CEO of the unit now known as AMD Silo AI. He is now chairman at two new ventures: physical AI lab NestAI, and QuTwo, an AI startup aimed at helping companies prepare for the era of quantum computing

Currently fully funded by Sarlin’s family office, PostScriptum, QuTwo describes itself as “an AI lab for the quantum era.” Rather than waiting for quantum computing to mature, however, it is already working with enterprise customers — including European fashion retailer Zalando, with which it is developing what the two companies call “lifestyle agents,” AI tools designed to go beyond product search and proactively suggest products and experiences.

QuTwo is built on the premise that AI is hitting an efficiency wall that quantum computing may eventually help solve. But the company is not betting on when that will happen, Sarlin told TechCrunch. Instead, the startup is building QuTwo OS as an orchestration layer that allows companies to shift from classical to quantum computing — making use of hybrid computing along the way.

Sarlin invested in Finnish quantum companies IQM and QMill through PostScriptum, and is one of a growing number of investors who believe it will eventually outperform classical computers in a wide range of industry applications while easing AI’s energy demands. But he also thinks that initial use cases will require mixed hardware environments, and that enterprises would rather focus on their business problems while QuTwo OS takes care of the routing.

In that respect, the potential advantage of the middle ground known as “quantum-inspired” computing is that it is already viable today, because it uses classical hardware while simulating quantum behavior, working around the hurdles that still hinder quantum hardware. Meanwhile, QuTwo OS is designed to be flexible, supporting quantum or non-quantum algorithms and chips alike.

QuTwo’s team brings experience on both sides of the quantum-AI divide. On the quantum side, there’s IQM cofounder Kuan Yen Tan and board member Antti Vasara, also chair at SemiQon, a Finnish semiconductor startup focused on quantum chips. The enterprise side is equally represented, by Sarlin himself and Kaj-Mikael Björk, one of his former cofounders at Silo AI. Pekka Lundmark, the former CEO of Finnish telecom giant Nokia, also joined QuTwo’s board.

Across both areas, the team counts over 30 quantum and AI scientists, and Sarlin is clear where the company stands. “We’re building for the quantum world, but QuTwo is an AI company,” he said, meaning that QuTwo is “pushing AI workloads from classical to quantum.”

This also means that its customer base could be quite broad. Beyond Zalando, QuTwo also launched a joint quantum AI research initiative with OP Pohjola, a major Finnish financial services provider.

From the outset, QuTwo has been commercially minded and already has “large design partnerships which are in the tens of millions,” Sarlin said. Design partnerships — in which a vendor co-develops its product alongside enterprise customers — are a way for QuTwo to learn what those customers expect as it builds its product. They are also a bet from enterprises looking to establish early footing when and if quantum computing does arrive.

The software engineering world is currently wrestling with a fundamental paradox of the AI era: as models become more capable, the “systems problem” of managing them has become the primary bottleneck to real-world productivity. While a developer might have access to the raw intelligence of a frontier model, that intelligence often degrades the moment a task requires a long horizon or a deep context window.

But help appears to be on the way: San Francisco-based, Y Combinator-backed startup Random Labs has officially launched Slate V1, described as the industry’s first “swarm native” autonomous coding agent designed to execute massively parallel, complex engineering tasks.

Emerging from an open beta, the tool utilizes a “dynamic pruning algorithm” to maintain context in large codebases while scaling output to enterprise complexity. Co-founded by Kiran and Mihir Chintawar in 2024, the company aims to bridge the global engineering shortage by positioning Slate as a collaborative tool for the “next 20 million engineers” rather than a replacement for human developers.

With the release of Slate V1, the team at Random Labs is attempting to architect a way out of this zone by introducing the first “swarm-native” agentic coding environment. Slate is not merely a wrapper or a chatbot with file access; it is an implementation of a “hive mind” philosophy designed to scale agentic work with the complexity of a human organization.

By leveraging a novel architectural primitive called Thread Weaving, Slate moves beyond the rigid task trees and lossy compaction methods that have defined the first generation of AI coding assistants.

Strategy: Action space

At the heart of Slate’s effectiveness is a deep engagement with Recursive Language Models (RLM).

In a traditional setup, an agent might be asked to “fix a bug,” a prompt that forces the model to juggle high-level strategy and low-level execution simultaneously.

Random Labs identifies this as a failure to tap into “Knowledge Overhang”—the latent intelligence a model possesses but cannot effectively access when it is tactically overwhelmed.

Slate solves this by using a central orchestration thread that essentially “programs in action space”. This orchestrator doesn’t write the code directly; instead, it uses a TypeScript-based DSL to dispatch parallel worker threads to handle specific, bounded tasks.

This creates a clear separation between the “kernel”—which manages the execution graph and maintains strategic alignment—and the worker “processes” that execute tactical operations in the terminal.

By mapping onto an OS-style framework, inspired by Andrej Karpathy’s “LLM OS” concept, Slate is able to treat the limited context window of a model as precious RAM, actively, intelligently managing what is retained and what is discarded.

Episodic memory and the swarm

The true innovation of the “Thread Weaving” approach lies in how it handles memory. Most agents today rely on “compaction,” which is often just a fancy term for lossy compression that risks dropping critical project state. Slate instead generates “episodes”.

When a worker thread completes a task, it doesn’t return a sprawling transcript of every failed attempt; it returns a compressed summary of the successful tool calls and conclusions.

Because these episodes share context directly with the orchestrator rather than relying on brittle message passing, the system maintains a “swarm” intelligence.

This architecture allows for massive parallelism. A developer can have Claude Sonnet orchestrating a complex refactor while GPT-5.4 executes code, and GLM 5—a favorite for its agentic search capabilities—simultaneously researches library documentation in the background. It’s a similar approach taken by Perplexity with its new Computer multi-model agent

By selecting the “right model for the job,” Slate ensures that users aren’t overspending on intelligence for simple tactical steps while still benefiting from the strategic depth of the world’s most powerful models.

The business of autonomy

From a commercial perspective, Random Labs is navigating the early beta period with a mix of transparency and strategic ambiguity.

While the company has not yet published a fixed-price subscription sheet, the Slate CLI documentation confirms a shift toward a usage-based credit model.

Commands like /usage and /billing allow users to monitor their credit burn in real-time, and the inclusion of organization-level billing toggles suggests a clear focus on professional engineering teams rather than solo hobbyists.

There is also a significant play toward integration. Random Labs recently announced that direct support for OpenAI’s Codex and Anthropic’s Claude Code is slated for release next week.

This suggests that Slate isn’t trying to compete with these models’ native interfaces, but rather to act as the superior orchestration layer that allows engineers to use all of them at once, safely and cost-effectively.

I’ve reached out to

Architecturally, the system is designed to maximize caching through subthread reuse, a “novel context engineering” trick that the team claims keeps the swarm approach from becoming a financial burden for users.

Stability AI

Perhaps the most compelling argument for the Slate architecture is its stability. In internal testing, an early version of this threading system managed to pass 2/3 of the tests on the make-mips-interpreter task within the Terminal Bench 2.0 suite.

This is a task where even the newest frontier models, like Opus 4.6, often succeed less than 20% of the time when used in standard, non-orchestrated harnesses.

This success in a “mutated” or changing environment is what separates a tool from a partner. According to Random Labs’ documentation, one fintech founder in NYC described Slate as their “best debugging tool,” a sentiment that echoes the broader goal of Random Labs: to build agents that don’t just complete a prompt, but scale like an organization.

As the industry moves past simple “chat with your code” interfaces, the “Thread Weaving” of Slate V1 offers a glimpse into a future where the primary role of the human engineer is to direct a hive mind of specialized models, each working in concert to solve the long-horizon problems of modern software.