AGI

Was All the Discussion on AGI Part of a Broader Industry Pattern? Jensen Huang Weighs In

Was All the Discussion on AGI Part of a Broader Industry Pattern? Jensen Huang Weighs In

Are we actually close to cracking AGI, or is that only a fantasy world that tech enthusiasts continue to expend billions into? Jensen Huang has an opinion.

NVIDIA’s CEO believes that they have “sort of” achieved AGI. You know, the tech dream- Artificial General Intelligence, AI that is on par with the human brain.

The claim.

It’s quite a recent but bold claim that Huang’s making. On the Lex Fridman Podcast, he states, “I think it’s now. I think we’ve achieved AGI,” in response to whether AI will finally come to match or surpass human-level intelligence.

Note: To offer readers context, Fridman frames AGI as an AI system capable of building and running a billion-dollar company.

However, what’s surprising isn’t the topic of AGI.

The “backtracking.”

It’s that Huang didn’t wait long before walking his claim back in the same conversation. He highlighted that Fridman was talking about running a $1 billion company, but he didn’t specify for how long. And with that, NVIDIA’s CEO elaborates that it’s not out of the question that someday Claude could create a web service or interesting app that a few billion people use briefly for $0.50 before it goes out of business.

He further comments, “A lot of people use it for a couple of months, and it kind of dies away,” saying the odds of AI agents “building NVIDIA is 0%.”

That sounds less like a backtrack and more like a sleight of hand. Because if AI can spin up, go through this entire cycle, and end up producing $1 billion in revenue even once? That reframes AGI, not as a durable future, but as a short-term commercial flash.

And that’s not what the tech leaders or investors thought AGI was ever about.

The market opinion and critics.

The opinions on AI aren’t in sync with the direction of the actual spending on AI infrastructure. Could it be that building a narrative around ‘imminent’ AGI will help justify all the ‘enthusiastic’ resource allocation? Well, all that depends on how you define imminent.

But all of this is also part of a well-known industry pattern. Huang called it a commercial flash; Altman says they’re very close to it, while Nadella disagrees that we could even imagine what AGI would be like at this point in time.

In short, Huang definitely agrees with Fridman’s narrow, commercially defined benchmark for AGI.

Maybe the chip leader realized mid-conversation that the current AI can’t sustain the kind of complex, stable institution that NVIDIA represents. So, how can we even assume we’re close to achieving AGI?

Retail Media Ecosystem Explained

The Retail Media Ecosystem Explained: Components, Players, and Why Everyone Wants In

The Retail Media Ecosystem Explained: Components, Players, and Why Everyone Wants In

Most explainers on retail media describe what it is and call it a day. This one goes further: who’s actually in the room, what each player wants, and why the whole thing works when it works.

Retail media is one of those terms that gets thrown around in planning meetings by people who have slightly different definitions of it in their heads and nobody stops to compare notes.

So before the strategy conversation: what is it, who builds it, who buys it, and what does each party actually get out of it.

The short version

A retailer has customers. Those customers have purchase histories, browsing behavior, basket data, loyalty card activity. That data is valuable, extremely valuable, to the brands whose products those customers buy or might buy.

Retail media is the business of a retailer monetizing that data, a model that sits at the core of modern retail media advertising and adtech evolution, by selling advertising to brands, letting them target those customers with ads that appear across the retailer’s owned properties and increasingly across external channels too.

Amazon built the template. Now 277 retailers globally are running some version of the same model, forming a rapidly expanding landscape of retail media networks.

The global retail media market hit $140 billion in 2024, reflecting the accelerating future of retail media. It’s projected to reach $165 billion by the end of 2026. For context, it’s growing faster than the total digital ad market and is on track to account for a quarter of all US ad spend by 2028.

That’s the short version. Now for the parts that actually matter.

The components: what the ecosystem is actually made of

Retail media is actually made

On-site inventory

This is where most people’s mental image often shaped by early exposure to retail media network platforms. of retail media begins and ends. Sponsored product listings on Amazon. Promoted placements on Walmart’s search results. Banner ads on a retailer’s homepage.

On-site is the foundation because it captures buyers at the sharpest point of intent: they’re already on the platform, already searching the category, already in purchase mode. The brand pays to show up at the exact right moment.

Sponsored product ads alone are projected to account for $38 billion in advertiser spend in 2026. For most brands, it’s the entry point into retail media and often the highest-performing format in the mix.

The limitation is ceiling. On-site inventory is finite. There are only so many sponsored slots on a search results page, and as more brands compete for them, CPCs rise and efficiency tightens.

Off-site inventory

This is where the ecosystem opened up. The data doesn’t have to stay on the retailer’s platform.

Off-site retail media uses a retailer’s first-party shopper data, extending into environments similar to modern social media marketing ecosystems. to target those same shoppers on external publisher sites, social platforms, programmatic display, and connected TV. The retailer’s audience, everywhere else they go online.

It’s why Amazon DSP is a meaningful product independent of Amazon’s own pages. It’s why Walmart Connect can reach Walmart shoppers on third-party inventory. The data travels; the real estate doesn’t have to be owned.

Over 20% of US retail media spend now goes to off-site channels, aligning closely with cross-media advertising strategies. Amazon DSP advertisers grew spend 31% year over year in Q4 2025 as impressions climbed 32%.

In-store

The most underinvested component for most brands, despite its importance in the evolving future of retail media networks. and the one catching up fastest.

Digital screens at the endcap. Checkout lane displays. Store-mode features in retailer apps that serve offers based on what’s physically around the shopper. Audio in the aisles. Scan-and-go integrations that serve a competing brand’s ad the moment you scan their competitor into your basket.

80% of consumer spending still happens in physical stores. The retail media ecosystem is finally building the infrastructure to monetize that attention the same way it monetizes digital.

The data and measurement layer

This is the component that makes the whole thing worth anything: closed-loop measurement.

A brand runs a sponsored product campaign. A customer clicks, or doesn’t. A purchase happens, or doesn’t. The retailer knows, because the purchase happens on their platform or through their loyalty system. The attribution loop closes.

No other advertising channel has this by default. Google knows if someone clicked. It doesn’t know if they then went and bought the product in-store three days later. Retail media does, or can, which is why the ROI case for brands is structurally stronger than almost anything else in the media mix.

The catch: every retailer runs their own attribution model, their own definition of conversion, their own reporting format. Comparing ROAS across four different networks is still a manual exercise that requires reconciling fundamentally different methodologies. It’s the biggest infrastructure problem in the space.

The players: who is actually in this ecosystem

Who Is Actually in This Ecosystem

The retailers

The retailers are the asset owners. They have the data, the inventory, and the customer relationships.

Amazon sits at the top with $60 billion in ad revenue in 2025 and roughly 75% of the US retail media market by some measures. Walmart Connect is the fastest-growing major network. Target’s Roundel, Kroger Precision Marketing, CVS Media Exchange, Instacart Ads, Home Depot’s Orange Apron Media, each one is a distinct network with its own audience, its own data set, and its own ad tech stack.

What’s changing in 2026 is the mid-market. Smaller retailers who can’t build proprietary ad tech are licensing Amazon’s Retail Ad Service or partnering with platform providers to stand up competitive networks. The ecosystem is decentralizing faster than most brands have adjusted for.

The brands and advertisers

For brands, retail media sits at an awkward intersection similar to how budgets are split across B2B media partnerships and digital channels. of trade marketing and brand marketing budgets. Historically, money spent at the retailer level came from trade. But off-site CTV campaigns using Kroger data? That’s a brand media buy.

Most large CPG brands now allocate 39% of total advertising spend to retail media. The ones getting strong returns are running it as a performance channel with real incrementality testing, not just renewing sponsored product budgets because they always have.

The ones getting mediocre returns are treating the network’s self-serve dashboard as the strategy.

The technology layer

Behind every retailer’s network is an ad tech stack similar to infrastructures discussed in retail media adtech ecosystems, demand-side platforms, supply-side platforms, data clean rooms, measurement and attribution tools, and creative optimization systems.

Some retailers built their own. Most are assembling from vendors: The Trade Desk, LiveRamp, Epsilon, Criteo, CitrusAd, Quotient. Data clean rooms have become the infrastructure, especially as privacy concerns grow alongside synthetic media and deepfake technologies. for brands and retailers to collaborate on audience data without the retailer handing over raw customer records, which they won’t and legally often can’t.

Only 12% of commerce media decision-makers describe themselves as having reached an advanced state with full-funnel capabilities across on-site, off-site, and in-store. The tech exists. The integration is the hard part.

The agencies

Agencies occupy an awkward position as responsibilities overlap with modern media buying processes. in the retail media ecosystem. Retail media buying historically happened through retail or shopper marketing teams. Digital media buying happened through media agencies. They were separate workflows with separate briefs and separate relationships.

Retail media collapsed that division. requiring integration similar to cross-media ad strategies. A brand now needs someone who understands Amazon’s auction mechanics, Walmart’s audience segments, programmatic DSP buying, and CTV creative requirements, simultaneously. The agencies that built that capability are worth a lot right now. The ones that haven’t are billing for it anyway.

The benefits: what each player actually gets

What Each Player Actually Gets 2

For retailers: the margin business they always needed

Core retail is a margin-thin business. making diversification into retail media advertising models essential. Grocery runs at 1% to 3% net margin on a good year. Retail media generates 50% to 70% operating margins on ad revenue.

That’s not a side business. For retailers operating at scale, media revenue is becoming a meaningful offset to the structural cost pressures in their core operations. Walmart’s media and data business is a strategic asset in a way their apparel category will never be.

For brands: the data they can’t get anywhere else

What brands are actually buying are often tied closely to data-driven lead generation services. When they buy retail media, it is not impressions. It’s access to purchase data.

Behavioral targeting on social is based on inference: this person liked three fitness posts, so they might buy protein supplements. Retail media targeting is based on purchase history: this person bought protein supplements from this retailer twice in the last 60 days. One is a guess. The other is a record.

For brands trying to reach buyers at the moment of highest commercial intent, and measure whether the ad actually drove a sale, retail media is the sharpest tool available.

For the shopper: relevance instead of noise

This one gets skipped in most ecosystem explainers, but it matters.

When retail media works as it should, similar to effective social media marketing strategies. the shopper sees ads for products genuinely relevant to what they buy, when they’re already thinking about buying. That’s a different experience from a retargeted ad following someone around the internet for a product they bought three weeks ago.

The ecosystem earns consumer tolerance by being useful. The moment it tips into surveillance-feeling or repetitive, that tolerance disappears fast.

Where the ecosystem goes from here

The three pressure points that will shape retail media through 2026 and beyond are measurement standardization, off-site scale, and consolidation.

Three pressure point shaping the ecosystem

Measurement first. The IAB has pushed for standards. Individual networks have their own incentives to keep methodologies proprietary. Until a brand can compare incremental ROAS across Amazon, Walmart, and Kroger in a consistent format, the budget allocation decisions happening inside brands will keep being made on incomplete information.

Off-site second. The ceiling on on-site inventory is real. The networks that figure out how to extend their first-party data into premium off-site environments, CTV especially, will hold advertiser budgets through the next phase. The ones that stay purely on-site will commoditize.

Consolidation third. 277 retail media networks is not a stable number. Most of them lack the scale, the data infrastructure, and the advertiser relationships to compete long term. The market will concentrate. The question is whether it concentrates around retailers or around the ad tech layer that retailers increasingly depend on.

The ecosystem is not finished being built. That’s either a problem or an opportunity, depending entirely on which side of the budget you’re sitting on.

Meta

Meta Spent $80 Billion on a World Nobody Wanted to Live In. Now It Wants Your AI Budget.

Meta Spent $80 Billion on a World Nobody Wanted to Live In. Now It Wants Your AI Budget.

Meta scrapped its $80B metaverse bet and is pivoting to AI. Here’s everything Zuckerberg is asking you to trust him with next.

Mark Zuckerberg renamed his company after a virtual world in 2021- the Metaverse. This week, Meta confirmed it has stopped expanding that world. Horizon Worlds survives in reduced form. The Metaverse, as a strategic vision, does not.

The bill is $80 billion. That bought a virtual environment with roughly 200,000 monthly active users at its peak. A mid-size city newsletter outperforms that number.

The failure was not technical. Zuckerberg confused infrastructure ambition with human desire. People did not want legless avatar meetings. They wanted to call someone, share a photo, buy something. The platforms that won met people where they already were. The Metaverse asked them to relocate.

The people who built it deserve to be named separately from the decision that sent them there. Many believed in it genuinely. Some still do. They are not the story. The judgment that deployed them is.

Now Zuckerberg is pivoting to AI. The infrastructure investment is serious. The model work is competitive. The distribution across Facebook, Instagram, and WhatsApp is an advantage almost no one else holds.

He is asking the same public that watched $80 billion disappear to trust that this conviction is different. The reading on human behavior is better this time. That the room he is building is one people will actually want to enter.

He may be right. A track record, though, does not disappear because the next bet is more plausible. It sits on the table. It is sitting there now.

The AI Agent the West Banned Just Became China's Hottest Product

Google’s Firebase Studio Sunset is a Lesson in Modern Tech Whiplash

Google’s Firebase Studio Sunset is a Lesson in Modern Tech Whiplash

Google is shutting down Firebase Studio less than a year after launch. While core services remain, developers must migrate to AI Studio or Antigravity soon.

Google has a reputation for the shiver it sends down a developer’s spine whenever the word sunsetting appears in an inbox. This time, the target is Firebase Studio.

The tech giant is pulling the plug despite launching the platform as a full-stack AI workspace only in 2025.

The official message is corporate optimism. Google claims it’s “simplifying” the offerings by folding the lessons learned from this preview into flagship tools like Google AI Studio and the new Antigravity IDE. They want to streamline the path from a simple prompt to a production-ready app.

If you were using Studio for its browser-based ease, you’re headed to AI Studio; if you wanted deep, local code control, Antigravity is your new home.

There is some nuance to be found in the wreckage. Unlike many of Google’s past “kills,” this isn’t a total abandonment of the underlying tech. Core services such as Firestore, Authentication, and App Hosting aren’t going anywhere.

Your actual databases and user data are safe; it’s just the “Studio” environment, i.e., the UI and the agentic workflow, that is being dismantled and reassembled elsewhere.

However, the logic remains frustrating. Firebase Studio was originally the evolution of Project IDX, offering a low-barrier way for developers on underpowered hardware to build complex apps.

By pushing users toward Antigravity, which favors a local, “code-first” workflow, Google is subtly raising the bar for entry again. It’s a move toward consolidation that prioritizes high-velocity professional workflows over the experimental, accessible middle ground that Studio briefly occupied.

Googles Firebase Studio Sunset is a Lesson in Modern Tech Whiplash

The AI Agent the West Banned Just Became China’s Hottest Product

The AI Agent the West Banned Just Became China’s Hottest Product

Google blocked OpenClaw. Tencent just placed it inside WeChat for a billion users. The same tool, but two very different bets on what AI agents are worth.

While Google was suspending accounts and Meta was blocking access, Tencent was opening a door.

On Sunday, Tencent launched ClawBot, an integration that puts OpenClaw directly inside WeChat as a contact. Over a billion monthly users can now send commands to an AI agent the same way they text a friend. No new app. No friction. Just a conversation that automates your email, moves your files, and runs your errands.

The contrast with the Western response to OpenClaw is not subtle. It is the whole story.

Alibaba moved the week before with Wukong, an enterprise platform built to coordinate multiple agents. Baidu followed immediately with OpenClaw tools spanning desktop, cloud, mobile, and smart-home devices.

Three of China’s largest tech companies deployed major AI agent products within weeks of each other. The China Development Forum, held this weekend in Beijing, centered on industrial AI in its theme for the country’s next Five-Year Plan.

The government has a phrase for what it is building- a new form of intelligent economy.

Regulators in China have flagged security concerns around agent products. That is worth noting. It did not slow anyone down.

What we are watching is the same open-source tool generating opposite responses on either side of a geopolitical line.

In the West, OpenClaw triggered platform bans, account suspensions, and cease-and-desist letters. The concern was real: unauthorized infrastructure access, subsidized token arbitrage, and security vulnerabilities that researchers documented in writing.

In China, those same properties became a feature. An open agent that connects to any large language model through an API is exactly what you want if you are trying to move fast and embed AI into a consumer base of a billion people before your competitors do.

The second-order question is about the users.

A billion people on WeChat can now delegate tasks to an agent that operates on their behalf, within an app that is already their wallet, social graph, work communication, and government interface.

The convenience is genuine. So is the surface area.

OpenClaw taught the West something about platform control. It is teaching China something about distribution. Both lessons are being applied at speed. The conclusions are not the same.

Google

Google Might Use AI to Overwrite Original News Headlines. What Could This Possibly Mean?

Google Might Use AI to Overwrite Original News Headlines. What Could This Possibly Mean?

Well, publications no longer need to put heed into their headlines, because Google might change news search forever- with AI.

At a time when content saturation has reached an unsalvageable juncture- AI is only adding to the humdrum.

Users are fatigued- first it was the snacky content that captured their attention spans, then it was AI-generated slop. All of these have fed into a long and curious case of brain rot.

But content purely for entertainment isn’t the only one that’s losing its quality. It’s as if tech companies have taken a pledge to run all platforms amok with AI- that’s the true potential of this modern tech.

The next target is news headlines.

That was brought to attention by The Verge. The publication noticed that one of their headlines wasn’t the same as what they had written it to be. The original headline (the image below) became “‘Cheat on Everything’ AI Tool.”

image

The five-word headline contributed nothing to the article (whether that’s clickbait or short-lived curiosity). However, it did take away something from the piece- it took away the nuance and meaning that the article actually meant to discover.

It’s all owing to Google.

Google is conducting a narrow experiment and attempting to revamp its search queries. What they are starting with is a news space in Google Discover. The tech giant replaced The Verge’s headlines with ones that aren’t written by the publication itself. And that isn’t minute tweaking a really lengthy headline as they used to do, but overwriting the original.

That might be a small experiment according to Google.

However, there’s no indication that Google’s making such changes. That’s how AI is being incorporated lately in content-driven spaces. And the company isn’t stopping here. It’s also planning to transform how websites show up in Search altogether. It’s influencing how publications and brands market content.

Doesn’t this say a lot about authority?

For Google, the objective is simple: elevate user satisfaction by better matching headings with precise queries. The overall intention is to work on web engagement.

But to what extent will this affect brands and publications such as The Verge?

The extent of it seems uncertain. Because it’s not merely influencing content quality, it’s directly about the future of real, unbiased journalism. We all know that AI models are trained on limited datasets.

When AI in search masks the intent by altering all the headlines, it’s the essence of journalism that takes the brunt. The pathway between a reader and a news piece is severed; clickbait will replace it. What happens then?

It’s a future that we can forecast. But when it’s here, it’ll be a trend that snuck up on us.