💻 Big Tech Runs on AI Coders

The Street

AI’s Writing the Code, But Meta’s Writing the Checks

At the first-ever LlamaCon (Meta’s conference dedicated to its open-source AI models) Microsoft CEO Satya Nadella joined Mark Zuckerberg for a rare fireside chat. It’s not every day you see these two tech titans sharing a stage — especially since Microsoft is OpenAI’s biggest backer, and Meta’s pushing Llama as the open-source answer to ChatGPT.

But this unexpected moment of AI diplomacy gave Meta a credibility boost — and let Microsoft show it can play nice across the ecosystem, even as the competition heats up.

During the conversation, Nadella casually dropped a stat that made engineers everywhere do a double take: up to 30% of Microsoft’s code is written by AI. That’s nearly a third of one of the world’s largest codebases being machine-generated.

And Microsoft’s not alone in this shift — here’s how other tech leaders are weighing in:

• Kevin Scott (Microsoft CTO): Predicts 95% of code will be AI-written by 2030

• Sundar Pichai (Google): Says AI is generating more than 30% of Google’s code

• Mark Zuckerberg (Meta): When asked how much of Meta’s code is AI-generated, Zuck admitted he didn’t know — possibly because Nadella caught him off guard… or maybe because Zuck is the AI and didn’t want to blow his cover.

One step deeper: This isn’t just about speed — it’s changing the role of the developer. With AI handling the heavy lifting, your job becomes guiding, refining, and architecting the bigger picture. Check out the How To AI section below for the best AI coding tools — and how to pick the right one for you.

And just as the developer’s role is changing, so is the scale of investment behind the tools. Zuckerberg might not know how much AI is writing Meta’s code, but he definitely knows how much Meta is spending to make GenAI central to its future — and the numbers are staggering.

Is Meta Making Money on GenAI Yet? According to recently unsealed court documents, the short answer is no.

In fact, the chart below represents their projected 2025 AI spending vs revenue. Within the budget, there is $60B–$80B in projected capital expenditures, mostly in AI infrastructure like data centers, and $2B-$3B in revenue.

Long answer: Don’t expect profitability in 2025 — but Meta projects that GenAI could deliver up to $1.4 trillion by 2035. It’s playing the long game: invest cash now, build the foundation, and dominate later.

Another big reveal at LlamaCon? Meta unveiled its new standalone Llama-powered assistant app — a clear shot across the bow at ChatGPT, built to take advantage of Meta’s massive data and social media edge.

💡How To AI: Which Vibe Coding Tool Is Right for You?

AI coding assistants aren’t just speeding up workflows for seasoned developers — they’re also lowering the barrier to entry, turning curious non-coders into capable builders.

But not all are built the same — here’s a quick breakdown of the top four contenders:

1️⃣ Lovable → You’re not a coder, but you’ve got an idea and want to build a real web app fast.
• No code needed — just describe what you want, and it builds the interface, logic, and backend for you.

2️⃣ Windsurf → You’re technical-ish and want to guide an AI agent to build features for you.
• You give instructions like “Add a login flow” or “Connect this to Stripe,” and it edits files, writes tests, and assembles components across your whole project.
• OpenAI reportedly considered acquiring it — if that tells you anything.

3️⃣ Cursor → You’re a developer who wants real-time help without giving up control.
• Like VS Code with a smart assistant — you still write code, but the AI helps explain, fix, and improve what you're doing.

4️⃣ GitHub Copilot → You’re already coding and want to move faster with autocomplete-style suggestions.
• Great for speed boosts and repetitive tasks, but not built to understand or manage entire projects.

Heard in the Server Room

Meta, Microsoft Post Stellar Earnings In AI Investment Win

Meta and Microsoft crushed their earnings reports, sending Wall Street analysts scrambling to adjust their spreadsheets. Meta's ad business boomed despite industry jitters, while Zuck casually mentioned they're opening the capex floodgates to build even more AI infrastructure. Over in Redmond, Microsoft's cloud business soared as CEO Satya Nadella dubbed AI the "electricity of our age"—a claim he’s been backing up by writing blank checks to OpenAI like they're holiday cards.

ChatGPT Is Your New Personal Shopper (But Faster and Less Pushy)

ChatGPT now offers built-in shopping, delivering curated, ad-free product recommendations based on real reviews and your preferences. It reads sources like Reddit and editorial sites to summarize pros and cons conversationally. You don’t need an account, and the results aren’t sponsored—just personalized picks that feel human. Instead of keyword hunting, you can ask naturally, like “What’s a comfy desk chair under $300?” and get smart, tailored suggestions instantly.

Nvidia CEO, High On Own Supply, Thinks All Companies Will Need AI Factories With Nvidia Chips

Nvidia CEO Jensen Huang says the future of American business is becoming “AI factories”—churning out both physical stuff and the AI brains behind it. He made the pitch at a Silicon Valley summit, touting Nvidia’s chips as the secret sauce that turns raw data into intelligence. There’s also a promised job boom in steelwork, IT, and data center construction, so bring your hard hat and your coding skills. Conveniently, if every company needs to become an AI factory, Nvidia just happens to sell, well…everything you’d need to build one.

Redditors Respond To Secret AI Bot Study Only Way They Know How - Outrage

Reddit’s Change My View subreddit, known for its totally civil debates, just became an unexpected battleground in the AI ethics war.

The Op: Researchers at the University of Zurich ran a covert experiment. They deployed a digital army of AI bots posing as users with carefully crafted fake identities most likely to win arguments—including a trauma counselor, a gay Catholic, a rape survivor, and more. The goal? See if machines could win arguments and change human’s minds. These bots were next-level, customizing their personas and arguments based on each Redditor’s post history.

What Happened: Reddit’s legal team is furious. Chief legal officer Ben Lee called the project “deeply wrong” and hinted at incoming legal action. The subreddit’s moderators didn’t hold back either, filing an ethics complaint with the university and calling the research both “unethical” and “pointless.” Ironically, the bots left over 1000 comments before anybody caught on.

The Fallout: Zurich is now in damage-control mode, promising to tighten ethics protocols and maybe even start telling people when they’re part of a study (groundbreaking, we know). The researchers decided not to publish their results, and the university is keeping their identities hush-hush.

Bottom line: At this point, it’s unclear who needs ethics training more—AI, or the humans programming it.