Good morning. Well, the results are in … and don’t worry, you can look — both Meta and Microsoft beat earnings expectations by a healthy margin, and the stocks are soaring, despite all the chaos going on.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

• 🌎 AI for Good: Soil health

• 🚘 Here’s how Waymo is going after Tesla

• 📊 ‘As scientists, we must do better’; the problem with Chatbot Arena

• 💰 Despite it all, Meta and Microsoft have a surprisingly good night

AI for Good: Soil health

There is a significant link between climate change and soil health. 

Healthy soil — rich in a diverse array of nutrients and minerals — better sequesters carbon dioxide, helps reduce the impact of flooding and produces better, more nutrient-rich crops. 

Unhealthy soil can’t hold onto compounds nearly as well as its healthier counterpart, meaning heavy rain or flooding results in runoff, taking minerals, nutrients, compounds, sequestered carbon dioxide and pesticides with it. 

The trouble is that as much as 70% of the soil across the European Union is considered to be unhealthy. 

What happened: A coalition of researchers is trying to turn that ship around, and they’re leveraging AI to do it. 

• The AI for Soil Health project, funded by Horizon Europe, is working to build an AI-powered digital platform for rapid, accurate soil health measurement and testing.

• The core idea is to provide farmers and land managers with clear, data-driven insights to help guide them in new farming directions that will aid soil health without harming their ability to respond to steadily growing food demand. 

Since each little ecosystem is different, the specific solutions will vary from farm to farm, and country to country. 

Why it matters: “We are at a crossroads; we need to do something if we are to preserve European and global soil resources. It is a paradox that on the one hand, soil is part of the solution to reduce greenhouse gas emissions, and at the same time, 60-70% of Europe’s soil is not doing well,” Professor Mogens H. Greve of Aarhus University said in a statement.  

The “ambition” is that these tools will enable the agriculture sector to reduce its climate impact significantly by the end of the decade.

Become an AI Generalist that makes $100K (in 16 hours)

Still don’t use AI to automate your work & make big $$? You’re way behind in the AI race. But worry not:

Join the World’s First 16-Hour LIVE AI Upskilling Sprint for professionals, founders, consultants & business owners like you. Register Now (Only 500 free seats)

Date: 2nd-3rd-4th of May, 11 AM - 7 PM.

Rated 4.9/10 by global learners – this will truly make you an AI Generalist that can build, solve & work on anything with AI.

In just 16 hours & 5 sessions, you will:

✅ Learn the basics of LLMs and how they work.

✅ Master prompt engineering for precise AI outputs.

✅ Build custom GPT bots and AI agents that save you 20+ hours weekly.

✅ Create high-quality images and videos for content, marketing, and branding.

✅ Automate tasks and turn your AI skills into a profitable career or business.

All by global experts from companies like Amazon, Microsoft, SamurAI and more. And it’s ALL. FOR. FREE. 🤯 🚀 

Join now and get $3000+ in additional bonuses: AI community access ($1999), AI Tool Stack ($299), and Workflow Templates ($999)—all unlocked when you sign up and attend! 🎁

Here’s how Waymo is going after Tesla

Waymo, the Google-owned self-driving car company, has spent the past couple of years establishing itself as the only legitimate robotaxi company on the market. 

Scaling with relative caution to cities across the country, Waymo now delivers 250,000 paid autonomous rides each week. But right now, it remains limited by a relatively small total fleet size and incredibly expensive cars — unlike some of its early-stage competition, Waymo retrofits vehicles with a suite of sensors, meaning it has yet to take advantage of the cost savings afforded by high-volume car production. 

What happened: Earlier this week, Waymo entered into a preliminary agreement with Toyota “to combine their respective strengths to develop a new autonomous vehicle platform.”

• The idea is to cross Waymo’s technology with Toyota’s “vehicle expertise” to “enhance next-generation personally owned vehicles.” This marks the first time that Waymo is embarking on an attempt to step outside the robotaxi space and actually make its technology available to consumers. 

• While timing and details remain unclear, the move shortly follows a jab from Tesla Chief Elon Musk, who recently said that the “issue with Waymo’s cars is it costs way more money.” 

Musk has been talking about getting fully autonomous Teslas off the ground for around a decade, something he has yet to achieve (Tesla’s current software requires the hands-on, eyes-on attention of its drivers). Musk’s plan is for end-to-end neural networks and cameras to somehow unlock a fully autonomous vehicle; the problem with this, beyond neural network vulnerabilities, is a lack of redundancy, something that’s pretty important when it comes to cars. 

In addition to neural networks and cameras, Waymo takes advantage of radar and lidar sensors, which inject additional safety layers into the cars. 

Depending on how it plays out, this partnership could certainly eat into Tesla’s only advantage over Waymo: that it can sell its cars and technology to consumers directly.

• No more sycophants: OpenAI, upon rolling back its sycophantic ChatGPT update, said that it had placed too much focus on short-term, rather than long-term, feedback, resulting in the chatbot’s sycophancy. OpenAI said it is evolving its training procedures — and adding a few guardrails — to help avoid this problem in the future.

• Predictive policing: New York City’s Metropolitan Transportation Authority said recently that it is exploring the use of AI technology and cameras to enable "predictive prevention” of crime in the city’s subways. Though the MTA said the system will rely on behavior analytics rather than facial recognition, such a move raises massive ethical questions ranging from data privacy and consent to algorithmic bias in predictive policing and unwanted, but unavoidable, surveillance.

• Millions of Apple Airplay-enabled devices can be hacked via Wi-Fi (Ars Technica).

• If you own Ray-Ban Meta glasses, you should double-check your privacy settings (TechCrunch).

• American Panopticon (The Atlantic).

• Trump-Bezos call sets stage for tense earnings report from Amazon (CNBC).

• Mark Zuckerberg is planning a premium tier and ads for Meta’s AI app (CNBC).

‘As scientists, we must do better’; the problem with Chatbot Arena

We’ve talked often about the trouble with AI benchmarks. 

Often, they don’t measure what they purport to measure, they don’t translate to real-world efficacy and, without effective model transparency, they’re effectively meaningless. Still, the more people work to integrate AI technology into workflows and industries, the more important effective benchmarking becomes; people need to know what these models are good at (and what they’re not good at) to better understand how to use them. 

Of those benchmarks, the Chatbot Arena has become wildly popular. Chatbot Arena scores are included in almost every major model release, and it’s not hard to see why; the Arena allows users to compare and rank responses from anonymous chatbots, seemingly serving as a blind, fair indicator of which is more effective. 

What happened: A team of researchers from Cohere and a number of universities recently published an investigation into the Arena, finding that the benchmark is tilted in the favor of major developers, which are able to privately test multiple versions of a model before releasing one publicly. 

• Models from these major developers — OpenAI, Google, Meta, etc. — are tested at a greater frequency than smaller developers, meaning those companies get access to a far wider quantity of Chatbot Arena data, according to the researchers.

• This data enables those developers to unlock Arena-specific performance gains, resulting in “in overfitting to Arena-specific dynamics rather than general model quality.” 

It has become a somewhat regular occurrence that high-scoring models in the Arena are not the same models that get publicly released, enabling developers to claim high scores that don’t actually align with the model they end up selling. 

“The widespread and apparent willful participation in the gamification of arena scores from a handful of top-tier industry labs is undoubtedly a new low for the AI research field,” the researchers write. “As scientists, we must do better. As a community, we must demand better.”

The Chatbot Arena, which started as a research project in 2023, recently rebranded to the LMArena as part of its transition into a real company. 

Despite it all, Meta and Microsoft have a surprisingly good night

Last night, in the words of perennial tech optimist Dan Ives, kicked off a “major week” for Big Tech earnings. 

Meta and Microsoft, two members of the so-called Magnificent Seven, and two of the most prominent, publicly traded AI names on the market, reported earnings for the first quarter.

The environment in which the reports are entering, however, is a challenging one. 

Macroeconomic concerns have been mounting since President Donald Trump announced his sweeping suite of global tariffs earlier in April; the policy behind Trump’s trade war strapped the markets into a violent rollercoaster ride, with markets entering into deep routs, recovering from those deep routs, then falling again, and so on.  

The interesting thing is that, even before macroeconomics was a concern, Big Tech didn’t have a great start to the year. Tariffs aside, Wall Street has been getting impatient, and maybe a little nervous. Spending on AI infrastructure has reached truly enormous levels, but returns on all that investment have yet to really materialize. 

Adoption where it counts — in the enterprise — has been uneven at best, and slow and cautious at worst. 

Shares of Nvidia were up 171% in 2024. 

We’re four months into 2025, and shares of Nvidia are down 22%. 

In other words, challenging. 

But I digress. Here’s how they did. 

Microsoft

• Microsoft reported earnings of $3.46 per share, above expectations of $3.22, on revenue of $70.07 billion, above expectations of $68.42 billion. Revenue spiked 13% from last year’s numbers, and net income surged 18%. 

• Microsoft’s Azure revenue grew 33% year-over-year; 16 points of that growth were attributed to AI, though specific numbers are unclear. Microsoft’s Intelligent Cloud unit, which includes Azure, raked in $26.75 billion in revenue. 

Shares of Microsoft spiked around 9% in extended trading. CEO Satya Nadella called AI and the cloud “essential inputs” for the continued growth of the company. 

Microsoft said it expects revenue for the second quarter to land between $73.15 billion and $74.25 billion, above analyst expectations. Nadella affirmed that Microsoft is continuing to expand its data center capacity. 

Meta

• Meta reported earnings of $6.43 per share, above expectations of $5.28, on revenue of $42.31 billion, above expectations of $41.40 billion. That revenue number represents a 16% year-over-year increase. 

• The company said that its Meta AI assistant now has a billion users, though how that metric is measured remains unclear. CEO Mark Zuckerberg says he plans to build the product for about a year before monetizing it. 

Meta raised its capital expenditure outlook for the year from $60 to $65 billion, to $64 to $72 billion; the investment is meant “to support our artificial intelligence efforts as well as an increase in the expected cost of infrastructure hardware.” 

“Our business is also performing very well, and I think we’re well-positioned to navigate the macroeconomic uncertainty,” Zuckerberg told analysts on the call. 

Shares spiked as much as 5% in after-hours trading. 

With data center investments affirmed, shares of Nvidia rose roughly 4% in after-hours trading. 

It makes sense; Nvidia was last night’s clear winner. 

The capex push will continue. Massively growing data centers have become integral to business operations, since the Big Tech players are bundling generative AI into their existing platforms … which means greater GPU requirements, which means more and larger data centers, which means revenue for Nvidia. 

The circular AI economy continues. 

And with the technology getting bundled and incorporated into existing platforms, user interest remains super unclear. 

Stock futures rose in the wake of the reports. 

We’ll see if it sticks. 

🤔 Your thought process:

Selected Image 2 (Left):

• “Too much was in focus in Image 1. With the length of the exposure there should have been a lot more of the image out of focus.”

Selected Image 1 (Right):

• “The details had me fooled. I thought the AI was being cheeky with a heavily flared image, even though the sunburst looked more natural in the real image.”

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

Here’s your view on Duolingo:

30% of you think it’s a great idea; 30% of you are immediately deleting Duolingo.

26% think that, AI or not, the app is terrible.

Terrible app:

• “Even without AI, it's already approaching ‘slop’ levels of language learning, as it hit its enshittification phase of just being addicting with gamification over actual content. AI-forced development will just accelerate this desire to get ‘user n+1’ to log in and learn nothing just to keep a streak.”

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.