In The Loop Episode 15 | Google, OpenAI, Meta, Anthropic & The Three Battles To Own All AI

This week’s discussion focuses on three major AI battlegrounds. First, Google and OpenAI are now taking on Anthropic in the AI coding space. Then there’s Meta, trying to challenge OpenAI’s dominance in the consumer AI market. And finally, we’re seeing OpenAI’s slow but steady push into Google’s search territory.

What we’re witnessing is a high-stakes battle between trillion-dollar companies, all trying to shape the next computing paradigm. They want to own the entire experience—to be the platform you and I visit every single day.

In today’s episode, I’m going to tell that story. This is In The Loop with Jack Houghton. Hope you enjoy the show.

‍

‍

Battle #1: Anthropic vs. Google, the AI coding space

For the past year, the AI coding space has had a pretty clear hierarchy. Claude, by Anthropic, has been the go-to model for most programmers. They’ve been well ahead of the pack, even raising over $3 billion largely because of that advantage.

But that started to shift. On May 6, Google released Gemini 1.5 (often mistakenly called 2.5) just two weeks ahead of their annual developer conference, where they typically unveil their most exciting products. Clearly, they’re gearing up for something big.

Gemini 1.5 immediately overtook Claude in the Arena LLM Leaderboard for web development, becoming the first model to break the 1400 Elo rating barrier—which is seen as a gold standard for coding benchmarks. There’s been strong validation from developers as well. Google shared a quote from Celia Salvatierra at Cognition Labs—the creators of the DevON engineering agent—saying Gemini feels more like a senior developer because of its one-shot prompt ability to do large-scale refactors and make high-level decisions.

That said, I take all this with a grain of salt. The benchmark story is nuanced. While Gemini leads in the LMSys Arena test, it falls behind on others. For example, there's a benchmark called LiveBench, which is contamination-free—meaning the evaluation uses data the model hasn’t seen in training, avoiding regurgitation. On this test, OpenAI’s GPT-3.5 Turbo (also referred to here as “03 mini”) actually outperforms Gemini on coding accuracy.

So yeah, benchmarks can get confusing. Personally, I prefer trying the same one-shot tasks across different models and judging for myself.

Still, the key takeaway is that Google just topped one of the most influential assessments, which will matter to developers. But this isn't just about model performance. Google is playing a bigger game—integration. They’re weaving Gemini into Android Studio, Firebase, Cloud Code—everywhere they already own developer attention. And with their scale, they’re starting to steal the spotlight.

Meanwhile, OpenAI isn’t just sitting back. This week, they made a massive move: a $3 billion acquisition of Windsurfer, an AI coding tool. That’s about 1% of OpenAI’s total valuation—not a small bet. It shows just how strategic this space is. Coding remains one of the top AI use cases. It’s expensive to hire engineers. It’s slow. And if AI can reduce headcount or boost output, the ROI is clear.

‍

‍

Battle #2: OpenAI vs. Meta, the consumer AI space

When it comes to consumer AI—people using it for everyday tasks, often for free or via a basic membership—ChatGPT is way out in front with around 400 million weekly active users.

But Meta is coming for that market share. They’ve just launched the Meta AI app, powered by their latest language model, Llama 3. And they’re already claiming impressive numbers—500 million people have tried its AI features. In their Q1 earnings call, Mark Zuckerberg said Meta is on track to reach 1 billion monthly active AI interactions and that adoption has outpaced their internal projections.

Meta’s advantage is obvious: distribution. They’ve got massive reach across Facebook, Instagram, and WhatsApp. That kind of leverage makes it easier to scale quickly. I wouldn’t be surprised if they integrate personal data into their AI features, offering hyper-personalized experiences—which most people tend to prefer.

That said, one-to-one conversations with a chatbot require trust. And Meta still has a reputation problem after the data privacy scandals. So while the tech is promising, the brand perception might be a barrier.

Also, Zuckerberg’s language matters. He says “500 million people have tried” the features—not that they’re weekly or monthly active users. Until we see those metrics, it’s hard to gauge the real success.

What’s particularly interesting is Meta’s hardware play—Ray-Ban smart glasses, the Metaverse, and other early bets. None have paid off yet, and they may be too early, but they could become the next consumer platform.

Apple recently said it doesn’t think the iPhone will be the main consumer device forever—they believe it’ll be replaced. So could Meta's glasses be that next step? Maybe. But personally, I’m skeptical of Meta building great hardware and software. They’re better at acquiring and integrating—think WhatsApp and Instagram—than building from scratch.

Still, their narrative is compelling. They claim 50% of users are engaging with the voice assistant on the glasses, which is a big deal. That’s literally putting AI on your face—just one step from having it in your ear all day.

Whether Meta can take over OpenAI’s consumer market remains unclear. And whether Apple can successfully enter the space is also up in the air. I don’t naturally associate social media with task-oriented chatbots. Maybe Meta could succeed in companionship and therapy use cases. If you didn’t catch last week’s episode, check it out—I covered that trend in detail.

For me, AI is mostly a work tool. I’d say 70–80% of my usage is task-based. I’m not convinced people will turn to Meta AI for productivity. But personal tasks—shopping, relationships, emotional support—those are tightly linked to social media. That’s where Meta could really win.

As for OpenAI, they’re clearly not giving up ground. Their 2025 playbook is already in motion:

1. More powerful models e.g., GPT-4o.
2. Ubiquity—the best models are now available even at the free tier or supercharged for Pro users.
3. Vertical integration—with web browsing, shopping plugins, and more, making ChatGPT the go-to assistant app.

‍

‍

‍

Battle #3: Google vs. OpenAI, the online search and shopping space

Let’s talk about OpenAI’s distribution and their actual number of active users—because it’s still relatively small compared to Google.

For those who missed it, OpenAI recently rolled out a shopping feature in ChatGPT. You can now ask things like, “What are the top five laptops for video editing under $1000?” and get a curated, ad-free list. It provides detailed reasoning for each recommendation, lets you ask follow-up questions in the same conversation—like for reviews from different sources—and then, if you want, sends you directly to buy the item. All within a single chat.

And if the numbers are accurate, it’s already had a major impact.

OpenAI disclosed that ChatGPT processed over 1 billion web searches in a single week—a first for them. Now, for context, that’s just 1% of Google’s weekly search volume, but it’s a massive leap forward for OpenAI. Internally, they’ve told investors that conversational search is growing faster than anyone anticipated, especially for product discovery and shopping.

I personally found it interesting. I’m not a big online shopper, but what stood out was how easily you could pull in product reviews during the same conversation. That’s exactly how I make purchase decisions—look at a few options, check YouTube videos and reviews. So, if all of that happens in one fluid chat? That’s extremely powerful.

Now, of course, Google isn’t rolling over. They’re investing heavily in search innovation. You might’ve noticed the new AI Overviews feature in Google Search—that’s already been used by over a billion people. And they’ve also launched an AI Mode in the Google app, where you can toggle into an assistant-style interface.

But here’s the issue: to me, it feels like two separate products. It’s disjointed. Like Google hasn’t figured out how to fully merge the traditional search and the new AI experience—and I think that speaks to the bigger problem they face. Google’s entire business model still depends heavily on blue link ad clicks, which they can’t afford to disrupt too much.

And people are noticing. A KMPG survey in April found that 52% of users trust ChatGPT's recommendations more than Google’s, citing Google’s ad-heavy search results as a major reason for that mistrust. Users know Google is trying to sell them something with every query.

That’s why OpenAI, Perplexity, and others are going hard in the opposite direction. They’re not monetizing through ads. In a May 2025 interview, Sam Altman said he believes there’s a better way to connect users with products—without advertising. And you’ve probably picked up throughout this podcast: I totally agree. I’m not a huge fan of ad-driven experiences, so I love that OpenAI and others are challenging that model.

And Perplexity is right behind them. I fully expect them to do the same when they roll out their own web browser, continuing this pressure on Google’s outdated monetization strategy.

We’re starting to see the effects. In May 2025, Apple’s SVP of Services, Eddy Cue, made a major admission: Apple saw its first-ever drop in Safari web searches month-over-month. That’s a big deal because Safari search runs on Google—and Google pays $20 billion a year to stay the default search engine on Apple devices. That’s 18% of Apple’s earnings, and over 50% of Google’s search traffic comes from Apple users.

Even more striking: Cue explicitly said the drop was due to users switching to ChatGPT and Perplexity. Apple is now considering adding those search tools natively to Safari.

So, zooming out, this whole battle between trillion-dollar tech giants isn’t just about building the best coding assistant or chatbot. It’s about owning attention. Whoever shapes how we interact with information is going to define the next decade of tech.

Anyway, I hope you found today’s episode helpful. I’ll be following these stories closely and will see you next time.

‍

‍