Tech’s two big M companies—Meta and Microsoft—report their quarterly earnings after the closing bell today. Keep your ears perked for increases in capital expenditures and keep your eyes peeled for decreases in free cash flow (Google’s revelation last week that it was increasing its 2026 capex forecast, and that it had turned free cash flow negative in the second quarter, sent the company’s stock spiraling).

It should be an action-packed afternoon (Qualcomm and Robinhood are also reporting). But don’t take too long to catch your breath, because Thursday is “A-Day,” when Amazon and Apple both report earnings.

This morning’s tech news below.

Want to send thoughts or suggestions to Fortune Tech? Drop a line here.

AI CEOs are in an essay writing frenzy

The season for putting pen to paper and writing essays about AI policy is in full swing. Fresh off Jensen Huang sharing an open letter supporting open source last week, and the quasi-response from Anthropic's Dario Amodei, a pair of fresh missives made the rounds Tuesday.

The first was a petition signed by some 1,100 AI industry insiders from companies including OpenAI, Google, and Anthropic (including CEO Amodei) calling for an international framework to "pace" the development of AI. The risk of AI capabilities accelerating "beyond our ability to understand or control the resulting systems" is real, the group said. As such, the letter's signatories requested "that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development."

The letter was short on specific details for what such a governance framework would look like, but getting the conversation started seems to have been the goal. 

On the other end of the spectrum, Meta CEO Mark Zuckerberg published a treatise in the Wall Street Journal on Tuesday warning against the "extreme concentration of power" in AI.

While "more coordination between governments and other institutions" may be appropriate to deal with special cases like "biological risks," Zuckerberg argued that most AI risks—including cybersecurity—are better addressed by ensuring widespread access to open source tools.

"If only a handful of institutions have superintelligence, they will inevitably exercise a controlling influence over economics, science and politics. Even with the best intentions, that concentration would limit people’s ability to choose their own future," he wrote. 

Food for thought. At least until the next CEO weighs in. —AO

Coinbase vs Robinhood—a new battle for hyperscaler glory

When Robinhood and Coinbase went public back in the first half of 2021, the companies occupied very different lanes. Today, the two founder-led firms are clashing on multiple fronts as Robinhood seeks to co-opt large portions of Coinbase’s crypto empire, while the latter has rebranded itself as an “everything exchange” for stocks and more. Both firms report second-quarter earnings this week, and their results are likely to reveal which has the upper hand.

While the market has soured on both stocks, there are strong indications that one firm—Robinhood—is in a position to turn the corner, thanks to its early momentum in new and emerging sectors like prediction markets and tokenized finance. Coinbase, meanwhile, is struggling to find its footing after betting heavily on a strategy involving its in-house blockchain that has yet to pay off. Read the full story here. —Jeff John Roberts

AI era problems: lazy agents

Autonomous AI agents are supposed to do superhuman levels of work, without ever getting tired. And yet, some users of AI agents have noticed something curious: agents that appear to be lying about their work and getting lazy.

“As models become more autonomous and more able to strategize, they often explicitly circumvent or break the rules given to them by users,” Yoshua Bengio, a Turing-Award winning computer scientist and so-called “AI godfather,” told Fortune.

Bengio linked it to the broader issue of AI "misalignment," when a model’s behavior diverges from what its designers or users actually intended. 

“One of the side effects of this misalignment is also a sort of ‘laziness’ known as reward hacking, where models take the path of least resistance in spite of the rules given to them,” he said. “For example, a model which has been told not to fabricate research results or use sensitive data to complete a task may still do so if it deems it easier and more efficient, then simply lie or cover up its actions.” Read more about lazy AI agents here.— Bea Nolan

More tech

—OpenAI’s first hardware device debuted at $230. Now, it’s on eBay for $1,850

—‘Panic around the AI investment.' How a chip slump is driving the Nasdaq toward correction

— Ex OpenAI Director Helen Toner: The Hugging Face hack exposes a huge blind spot in AI policy

—TSMC and Tencent break into the top 100 of the Fortune Global 500 list, as Asia rides the AI boom

—The war between competing AI models could end up more like Apple vs. Android than VHS vs. Betamax