Clicks to Choices, AI is reshaping Advertising

Digital advertising has long revolved around a simple loop: search, click, convert. The rise of AI referral traffic—visits driven by answers from tools like ChatGPT, Gemini, Claude and Perplexity—is beginning to disrupt that model. For now, the shift is small. AI platforms account for roughly 0.24% of global internet traffic, according to SE Ranking. But growth is accelerating. AI referral traffic, which was at 0.15% in 2025, had risen 1.6 times by January 2026, and is expected to climb further by year-end.

In this week’s edition of Mint Tech Talk:

• How Generative AI is reshaping advertising
• AI Tool of the Week: Claude Dispatch
• Why netizens are fussing over Claude Mythos
• Googlers turning to Agent Smith to automate tasks
• Is AI to blame for Oracle layoffs?

Leader of the pack

Nearly four out of five AI-driven visits originate from ChatGPT, which generates about eight times more traffic than the next-largest player Google Gemini, according to SE Ranking. Through much of the first half of 2025, Gemini’s share hovered at 0.0083-0.0105%, even as ChatGPT’s share of global traffic more than doubled from 0.0799% to 0.1801%. By August, ChatGPT was sending 21.7 times more traffic to websites than Google Gemini.

Then, the momentum shifted.

In September 2025, ChatGPT traffic rose 53.8% but Gemini’s grew faster at 63.9%. By January 2026, Gemini’s traffic had more than doubled—around 12 times faster than its growth between January and October 2025—coinciding with the rollout of Gemini 3 models. It also surpassed Perplexity in global AI traffic.

ChatGPT remains the dominant platform but Google Gemini is growing at a faster clip. Advertisers are paying attention because the growth of AI referrals is steep. Adobe has reported a 10 times surge in Generative AI-driven referrals, while Similarweb estimates over a billion monthly visits now originate from AI tools.

AI & Advertising

Gen AI systems are replacing ranked links with synthesised answers, and advertising is moving upstream into those outputs, according to a BCG report.

Google’s AI Overviews provide an early signal. For exploratory and research-oriented queries, shopping and search ads now appear directly inside AI-generated responses, often ahead of organic results. Hence, visibility now relies less on keyword coverage and more on how well product data, creative, and bids align with model-level interpretation.

Microsoft Copilot is following a similar path, reframing search as an ongoing conversation built atop Bing’s existing ad ecosystem. According to BCG, Perplexity is testing a more explicit approach, surfacing sponsored follow-up questions that are clearly labeled and contextually generated, reflecting rising expectations for transparency in AI-generated experiences.

Additionally, conversational advertising is emerging as a distinct line item in the media mix, separate from search and programmatic, as brands reserve dollars to test, learn, and build early capability, notes BCG.

In the AI era, advertising inventory is emerging in three distinct forms:

• In-answer ads, embedded directly within synthesised responses when commercial intent is detected.
• In-conversation ads, appearing alongside dialogue during exploration—even for non-transactional queries.
• Agentic ads, where sponsored options surface as an AI Agent plans, compares, or completes tasks on a user’s behalf.

Agentic AI — Clicks to Outcomes

The advertising game is no longer just about traffic. It’s more about monetisation. In 2026, OpenAI began testing ads inside ChatGPT, reportedly crossing $100 million in annualised revenue within weeks and onboarding hundreds of advertisers, according to Reuters. Brands like Target, Ford, and Adobe have participated via agency partners such as WPP and Omnicom, The Verge and Adweek have reported. These are not experimental budgets but early bets on a new interface for discovery.

What advertisers are buying today is not “AI referrals” in the traditional sense. They are buying proximity to intent—placements that appear when a user asks a question with clear commercial relevance. It’s closer to search advertising than display, but a single prompt can collapse the entire funnel comprising discovery, comparison, and decision.

Agentic AI threatens to break current advertising models completely since such systems don’t just answer questions but also take actions. For instance, if you ask an AI Agent to “book the best hotel in Goa under ₹8,000”, instead of offering links, it could compare options, select one, and complete the transaction. This makes the very idea of “referral traffic” disappear since there are no clicks—simply decisions.

When Agentic AI systems begin to execute tasks end-to-end, marketers will care less about how many users visited their site and more about whether their product was selected by the AI Agent in the first place. This creates a new competitive dynamic. The traditional search offered blue links; even dominating one position left room for others. AI answers, by contrast, often surface just a handful of options—sometimes only one.

There are early hints of how this might play out. The e-commerce platforms already operate on “winner-takes-most” mechanics, such as Amazon’s Buy Box, where a single seller captures the majority of transactions. Agentic AI could extend that logic across the web. Instead of competing for page rank, brands may compete to become the default choice in an algorithmic decision.

None of this will happen overnight

Today’s AI still frequently defers to links, and ad formats remain limited and geographically constrained. In markets like India, monetisation is nascent. Attribution is murky, and user trust remains fragile, especially if paid placements blur into organic answers.

But the trajectory is clear. In the short term, AI usage will drive more referral traffic as users click through for deeper information. In the medium term, better answers will reduce the need to click. And in the long term, as agentic capabilities mature, the click itself may disappear. For advertisers, that is the real inflection point.

The question is no longer how to attract attention, but how to influence decisions made by machines on behalf of users. AI referral traffic, then, will serve as the bridge from an internet built on clicks to one built on choices.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚏𝚎𝚊𝚝𝚞𝚛𝚎 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢: 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙿𝚒𝚌𝚝𝚞𝚛𝚎 𝚝𝚑𝚒𝚜: 𝚊 𝚜𝚎𝚗𝚒𝚘𝚛 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚐𝚎𝚝𝚜 𝚘𝚞𝚝 𝚘𝚏 𝚊 𝚍𝚒𝚏𝚏𝚒𝚌𝚞𝚕𝚝 𝚌𝚕𝚒𝚎𝚗𝚝 𝚖𝚎𝚎𝚝𝚒𝚗𝚐 𝚊𝚗𝚍 𝚐𝚎𝚝𝚜 𝚒𝚗𝚝𝚘 𝚊 𝚌𝚊𝚋, 𝟺𝟻 𝚖𝚒𝚗𝚞𝚝𝚎𝚜 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝚘𝚏𝚏𝚒𝚌𝚎． 𝚂𝚑𝚎 𝚛𝚎𝚖𝚎𝚖𝚋𝚎𝚛𝚜 𝚝𝚑𝚊𝚝 𝚑𝚎𝚛 𝙲𝙵𝙾 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚎 𝚖𝚘𝚗𝚝𝚑𝚕𝚢 𝚖𝚊𝚗𝚊𝚐𝚎𝚖𝚎𝚗𝚝 𝚙𝚊𝚌𝚔—𝚍𝚊𝚝𝚊 𝚙𝚞𝚕𝚕𝚎𝚍 𝚏𝚛𝚘𝚖 𝚜𝚒𝚡 𝚍𝚎𝚙𝚊𝚛𝚝𝚖𝚎𝚗𝚝 𝙴𝚡𝚌𝚎𝚕 𝚏𝚒𝚕𝚎𝚜, 𝚏𝚕𝚊𝚐𝚐𝚎𝚍 𝚏𝚘𝚛 𝚟𝚊𝚛𝚒𝚊𝚗𝚌𝚎𝚜, 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚎𝚍 𝚊𝚗𝚍 𝚛𝚎𝚊𝚍𝚢—𝚋𝚢 𝚖𝚘𝚛𝚗𝚒𝚗𝚐. 𝙷𝚎𝚛 𝚕𝚊𝚙𝚝𝚘𝚙 𝚒𝚜 𝚋𝚊𝚌𝚔 𝚊𝚝 𝚑𝚎𝚛 𝚍𝚎𝚜𝚔. 𝙷𝚎𝚛 𝚝𝚎𝚊𝚖 𝚒𝚜 𝚞𝚗𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎. 𝚃𝚑𝚒𝚜 𝚒𝚜 𝚗𝚘𝚝 𝚊 𝚏𝚛𝚒𝚗𝚐𝚎 𝚜𝚒𝚝𝚞𝚊𝚝𝚒𝚘𝚗. 𝚃𝚑𝚒𝚜 𝚒𝚜 𝚃𝚞𝚎𝚜𝚍𝚊𝚢.

𝚄𝚗𝚝𝚒𝚕 𝚗𝚘𝚠, 𝚝𝚑𝚎 𝚘𝚙𝚝𝚒𝚘𝚗𝚜 𝚠𝚎𝚛𝚎 𝚕𝚒𝚖𝚒𝚝𝚎𝚍: W𝚊𝚒𝚝 𝚝𝚒𝚕𝚕 𝚜𝚑𝚎 𝚐𝚘𝚝 𝚋𝚊𝚌𝚔, 𝚌𝚊𝚕𝚕 𝚜𝚘𝚖𝚎𝚘𝚗𝚎, 𝚘𝚛 𝚜𝚔𝚒𝚙 𝚝𝚑𝚎 𝚙𝚛𝚎𝚙 𝚊𝚕𝚝𝚘𝚐𝚎𝚝𝚑𝚎𝚛. 𝙰𝙸 𝚝𝚘𝚘𝚕𝚜 𝚑𝚊𝚟𝚎𝚗’𝚝 𝚜𝚘𝚕𝚟𝚎𝚍 𝚝𝚑𝚒𝚜 𝚋𝚎𝚌𝚊𝚞𝚜𝚎 𝚝𝚑𝚎𝚢 𝚕𝚒𝚟𝚎 𝚒𝚗 𝚊 𝚋𝚛𝚘𝚠𝚜𝚎𝚛 𝚝𝚊𝚋 𝚝𝚑𝚊𝚝 𝚗𝚎𝚎𝚍𝚜 𝚢𝚘𝚞 𝚜𝚒𝚝𝚝𝚒𝚗𝚐 𝚒𝚗 𝚏𝚛𝚘𝚗𝚝 𝚘𝚏 𝚒𝚝．

𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚌𝚑𝚊𝚗𝚐𝚎𝚜 𝚝𝚑𝚊𝚝. 𝙸𝚝 𝚕𝚎𝚝𝚜 𝚢𝚘𝚞 𝚜𝚎𝚗𝚍 𝚊 𝚝𝚊𝚜𝚔 𝚝𝚘 𝚢𝚘𝚞𝚛 𝚍𝚎𝚜𝚔𝚝𝚘𝚙 𝙰𝙸 𝚏𝚛𝚘𝚖 𝚢𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎 𝚊𝚗𝚍 𝚌𝚘𝚖𝚎 𝚋𝚊𝚌𝚔 𝚝𝚘 𝚝𝚑𝚎 𝚏𝚒𝚗𝚒𝚜𝚑𝚎𝚍 𝚠𝚘𝚛𝚔. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚞𝚗𝚜 𝚘𝚗 𝚢𝚘𝚞𝚛 𝚌𝚘𝚖𝚙𝚞𝚝𝚎𝚛, 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚊𝚌𝚝𝚞𝚊𝚕 𝚏𝚒𝚕𝚎𝚜, 𝚞𝚜𝚎𝚜 𝚢𝚘𝚞𝚛 𝚌𝚘𝚗𝚗𝚎𝚌𝚝𝚎𝚍 𝚝𝚘𝚘𝚕𝚜, 𝚊𝚗𝚍 𝚍𝚎𝚕𝚒𝚟𝚎𝚛𝚜 𝚝𝚑𝚎 𝚘𝚞𝚝𝚙𝚞𝚝 𝚋𝚊𝚌𝚔 𝚝𝚘 𝚝𝚑𝚎 𝚜𝚊𝚖𝚎 𝚌𝚘𝚗𝚟𝚎𝚛𝚜𝚊𝚝𝚒𝚘𝚗. 𝚈𝚘𝚞 𝚍𝚘𝚗’𝚝 𝚠𝚊𝚝𝚌𝚑 𝚒𝚝 𝚠𝚘𝚛𝚔. 𝚈𝚘𝚞 𝚓𝚞𝚜𝚝 𝚛𝚎𝚝𝚞𝚛𝚗 𝚝𝚘 𝚛𝚎𝚜𝚞𝚕𝚝𝚜.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝙳𝚘𝚠𝚗𝚕𝚘𝚊𝚍 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚎𝚜𝚔𝚝𝚘𝚙 + 𝙲𝚕𝚊𝚞𝚍𝚎 𝚖𝚘𝚋𝚒𝚕𝚎 𝚊𝚙𝚙. 𝙴𝚗𝚊𝚋𝚕𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚒𝚗𝚜𝚒𝚍𝚎 𝚝𝚑𝚎 𝙲𝚘𝚠𝚘𝚛𝚔 𝚝𝚊𝚋.

𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚑𝚎𝚕𝚙𝚜 𝚢𝚘𝚞 𝚠𝚒𝚝𝚑:

𝚂𝚒𝚕𝚎𝚗𝚝 𝙲𝚘𝚖𝚙𝚒𝚕𝚎𝚛: 𝙲𝚘𝚖𝚙𝚒𝚕𝚎 𝚖𝚞𝚕𝚝𝚒-𝚏𝚒𝚕𝚎 𝚛𝚎𝚙𝚘𝚛𝚝𝚜 𝚠𝚑𝚒𝚕𝚎 𝚢𝚘𝚞’𝚛𝚎 𝚒𝚗 𝚝𝚛𝚊𝚗𝚜𝚒𝚝 𝚘𝚛 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚖𝚎𝚎𝚝𝚒𝚗𝚐𝚜.

𝙸𝚗𝚋𝚘𝚡 𝚂𝚎𝚗𝚝𝚒𝚗𝚎𝚕: 𝚃𝚛𝚒𝚊𝚐𝚎 𝚢𝚘𝚞𝚛 𝚒𝚗𝚋𝚘𝚡 𝚊𝚗𝚍 𝚏𝚕𝚊𝚐 𝚞𝚛𝚐𝚎𝚗𝚝 𝚎𝚖𝚊𝚒𝚕𝚜 𝚋𝚎𝚏𝚘𝚛𝚎 𝚢𝚘𝚞𝚛 𝚍𝚊𝚢 𝚋𝚎𝚐𝚒𝚗𝚜.

𝚁𝚎𝚌𝚞𝚛𝚛𝚒𝚗𝚐 𝙿𝚒𝚕𝚘𝚝: 𝚁𝚞𝚗 𝚛𝚎𝚌𝚞𝚛𝚛𝚒𝚗𝚐 𝚝𝚊𝚜𝚔𝚜 (𝚕𝚒𝚔𝚎 𝚍𝚊𝚒𝚕𝚢 𝚜𝚝𝚊𝚝𝚞𝚜 𝚛𝚎𝚙𝚘𝚛𝚝𝚜) 𝚘𝚗 𝚊 𝚜𝚌𝚑𝚎𝚍𝚞𝚕𝚎, 𝚖𝚘𝚗𝚒𝚝𝚘𝚛𝚎𝚍 𝚏𝚛𝚘𝚖 𝚢𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙰 𝚌𝚘𝚗𝚜𝚞𝚕𝚝𝚊𝚗𝚝 𝚑𝚎𝚊𝚍𝚒𝚗𝚐 𝚒𝚗𝚝𝚘 𝚊 𝚌𝚕𝚒𝚎𝚗𝚝 𝚙𝚒𝚝𝚌𝚑 𝚛𝚎𝚊𝚕𝚒𝚣𝚎𝚜 𝚜𝚑𝚎 𝚑𝚊𝚜𝚗’𝚝 𝚏𝚞𝚕𝚕𝚢 𝚙𝚛𝚎𝚙𝚙𝚎𝚍 𝚘𝚗 𝚊 𝚌𝚘𝚖𝚙𝚎𝚝𝚒𝚝𝚘𝚛. 𝙵𝚛𝚘𝚖 𝚑𝚎𝚛 𝚙𝚑𝚘𝚗𝚎:

“𝙲𝚑𝚎𝚌𝚔 𝚖𝚢 /𝚁𝚎𝚜𝚎𝚊𝚛𝚌𝚑 𝚏𝚘𝚕𝚍𝚎𝚛 𝚊𝚗𝚍 𝚎𝚖𝚊𝚒𝚕𝚜 𝚏𝚛𝚘𝚖 𝚘𝚞𝚛 𝚜𝚎𝚌𝚝𝚘𝚛 𝚕𝚎𝚊𝚍. 𝙲𝚘𝚖𝚙𝚒𝚕𝚎 𝚊 𝟷-𝚙𝚊𝚐𝚎 𝚜𝚗𝚊𝚙𝚜𝚑𝚘𝚝 𝚊𝚗𝚍 𝚜𝚊𝚟𝚎 𝚝𝚘 /𝙿𝚒𝚝𝚌𝚑-𝙿𝚛𝚎𝚙.”

𝙱𝚢 𝚝𝚑𝚎 𝚝𝚒𝚖𝚎 𝚜𝚑𝚎 𝚠𝚊𝚕𝚔𝚜 𝚒𝚗𝚝𝚘 𝚝𝚑𝚎 𝚛𝚘𝚘𝚖, 𝚒𝚝’𝚜 𝚍𝚘𝚗𝚎.

𝙰𝚗 𝙷𝚁 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚠𝚛𝚊𝚙𝚙𝚒𝚗𝚐 𝚞𝚙 𝚋𝚊𝚌𝚔-𝚝𝚘-𝚋𝚊𝚌𝚔 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠𝚜 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚛𝚎𝚎 𝚌𝚊𝚗𝚍𝚒𝚍𝚊𝚝𝚎 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐 𝚙𝚊𝚌𝚔𝚜 𝚛𝚎𝚊𝚍𝚢 𝚏𝚘𝚛 𝚝𝚘𝚖𝚘𝚛𝚛𝚘𝚠’𝚜 𝚏𝚒𝚗𝚊𝚕 𝚛𝚘𝚞𝚗𝚍. 𝙵𝚛𝚘𝚖 𝚑𝚎𝚛 𝚙𝚑𝚘𝚗𝚎:

“𝙿𝚞𝚕𝚕 𝚝𝚑𝚎 𝚜𝚑𝚘𝚛𝚝𝚕𝚒𝚜𝚝𝚎𝚍 𝙲𝚅𝚜, 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠 𝚚𝚞𝚎𝚜𝚝𝚒𝚘𝚗𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚝𝚑𝚎 𝙹𝙳, 𝚊𝚝𝚝𝚊𝚌𝚑 𝚝𝚑𝚎 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠 𝚗𝚘𝚝𝚎𝚜, 𝚊𝚗𝚍 𝚜𝚊𝚟𝚎 𝚘𝚗𝚎 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐 𝚍𝚘𝚌 𝚙𝚎𝚛 𝚌𝚊𝚗𝚍𝚒𝚍𝚊𝚝𝚎 𝚝𝚘 /𝙱𝚛𝚒𝚎𝚏𝚒𝚗𝚐-𝙿𝚊𝚌𝚔𝚜”

𝚂𝚑𝚎’𝚜 𝚑𝚘𝚖𝚎 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚒𝚖𝚎 𝚝𝚑𝚎𝚢’𝚛𝚎 𝚛𝚎𝚊𝚍𝚢 𝚝𝚘 𝚜𝚑𝚊𝚛𝚎．

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

• 𝚃𝚛𝚞𝚕𝚢 𝚊𝚜𝚢𝚗𝚌: 𝚈𝚘𝚞 𝚊𝚜𝚜𝚒𝚐𝚗 𝚝𝚑𝚎 𝚝𝚊𝚜𝚔 𝚊𝚗𝚍 𝚕𝚎𝚊𝚟𝚎. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚍𝚘𝚎𝚜𝚗’𝚝 𝚗𝚎𝚎𝚍 𝚢𝚘𝚞 𝚊𝚛𝚘𝚞𝚗𝚍.
• 𝚈𝚘𝚞𝚛 𝚏𝚒𝚕𝚎𝚜, 𝚢𝚘𝚞𝚛 𝚖𝚊𝚌𝚑𝚒𝚗𝚎: 𝙻𝚘𝚌𝚊𝚕 𝚙𝚛𝚘𝚌𝚎𝚜𝚜𝚒𝚗𝚐 𝚖𝚎𝚊𝚗𝚜 𝚜𝚎𝚗𝚜𝚒𝚝𝚒𝚟𝚎 𝚍𝚊𝚝𝚊 𝚜𝚝𝚊𝚢𝚜 𝚠𝚒𝚝𝚑 𝚢𝚘𝚞.
• 𝙾𝚗𝚎 𝚙𝚎𝚛𝚜𝚒𝚜𝚝𝚎𝚗𝚝 𝚝𝚑𝚛𝚎𝚊𝚍: 𝚈𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎 𝚊𝚗𝚍 𝚍𝚎𝚜𝚔𝚝𝚘𝚙 𝚜𝚝𝚊𝚢 𝚒𝚗 𝚜𝚢𝚗𝚌—𝚗𝚘 𝚜𝚠𝚒𝚝𝚌𝚑𝚒𝚗𝚐 𝚊𝚙𝚙𝚜, 𝚗𝚘 𝚌𝚘𝚙𝚢-𝚙𝚊𝚜𝚝𝚒𝚗𝚐 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚍𝚎𝚟𝚒𝚌𝚎𝚜.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

Elon Musk doubles down on Grok Imagine after Sora shutdown

SpaceX CEO Elon Musk has reacted to the shutdown of OpenAI Sora, claiming that that video-generation tools will be essential for achieving AGI.

Musk was responding to a post by a user who questioned why the billionaire continued to invest in video-generation tools, given that rivals like OpenAI had shut down the Sora service after reportedly losing $1 million per day. The billionaire noted that xAI remains heavily involved in video generation through Grok Imagine, because he believes it is a path to achieving AGI. Musk hypothesised that visual data carries much more information than text or other inputs. Read more.

What’s the mystery around Claude Mythos?

Anthropic’s latest AI model has been making waves on social media even before its official launch. The new model, codenamed “Claude Mythos”, was accidentally revealed after descriptions of the model were stored in a publicly accessible data cache that was first reported by Fortune.

After the news of the AI model leak came to light, an Anthropic spokesperson confirmed its existence and noted that the model represented “a step change” in AI performance and was “the most capable we’ve built to date”. Anthropic believes its upcoming model poses unprecedented cybersecurity risks. Read more.

Googlers are turning to Agent Smith to automate tasks

Google employees have begun using a new internal AI tool called “Agent Smith” that can automate tasks like coding, according to a report by Business Insider.

The AI agent has become so popular inside the company that the tech giant had to restrict access to handle the sudden influx of employees using it. Agent Smith is likely a reference to the popular character from the movie The Matrix, where Smith is a software agent inside the simulated world designed to monitor, control, and maintain order, and is capable of acting independently and intelligently. Read more.

Meta readies two prescription-supported Ray-Ban AI glasses

Meta Platforms Inc. plans to launch two new Ray-Ban smart glasses next week, especially aimed at prescription wearers. According to a report by Bloomberg’s Mark Gurman, the glasses will be sold alongside EssilorLuxottica SA and will come with prescription support.

This will be the first time that Meta will be introducing prescription support for its Ray-Ban series. The glasses, codenamed Scriber and Blazer, were earlier spotted in the US Federal Communications Commission filings by The Verge. Read more.

Oracle layoffs hit 12,000 employees in India

Oracle is laying off thousands of employees to cut costs, CNBC reported. Reports suggest the number can be as high as 30,000, with ~12,000 employees affected in India. Oracle had a headcount of 162,000 people as of May 2025, according to its most recent 10-K filing.

The Oracle layoffs come amid steps to increase spending on AI infrastructure in an effort to better compete with cloud rivals, such as Google Cloud and Amazon Web Services. You may read the full story here. Still, a detailed study by Yale University’s Budget Lab suggests the recent tech layoffs is not necessarily attributable to AI.