CEO Nadella Admits to the Temptation of "Token Maxing"

Microsoft CEO Satya Nadella has spoken about "token maxing" — the practice of overusing frontier models (the most powerful AI models) for everyday tasks. While pointing out that "the productivity gains should be worth the token costs," Nadella made the striking admission: "I'm a token maxxer myself. It's addictive."

It's noteworthy that the CEO of a major tech company is himself sharing concerns about AI cost efficiency. That sense of balance — using powerful models where they're the right fit rather than throwing them at everything — may hint at where the industry is heading.

TensorZero Archives Its Repository Right After Raising $7.3M

It caused quite a stir on Hacker News, racking up 104 points and 61 comments. TensorZero, an open-source AI tool, suddenly flipped its GitHub repository into an archived (read-only) state immediately after announcing a $7.3 million seed round.

The development team's intent is unclear at this point, and speculation is flying around the community: "Are they going closed-source after taking funding?" "Is this a business model pivot?" OSS projects changing course after raising money is nothing new, and in the AI space, the balance between openness and commercialization remains a contentious topic.

"Hades" Malware That Tricks AI Security Scanners

According to Tom's Hardware, a new malware campaign dubbed "Hades" employs a clever technique to evade AI-based security scanners. The malware embeds prompts about building nuclear and biological weapons inside the code, tricking AI scanners into aborting the scan for safety reasons.

As AI-powered security tools become widespread, the fact that attack techniques exploiting their weaknesses are already in practical use is a serious concern. How to guarantee the reliability of AI security looks set to become a major challenge going forward.

EU Budget 2027: A Mere €2.5B for AI

It has come to light that AI-related allocations account for just €2.5B (about 0.25%) of the EU's proposed 2027 budget (€200B in total). With €75B earmarked for "unity, resilience, and values," the Hacker News community has reacted with surprise and criticism over the paltry investment in AI.

For a Europe often said to be lagging behind the US and China in the AI race, this allocation speaks volumes about its strategic priorities. Whether it will be revised going forward, among other things, remains to be seen.

US Department of Energy Plans an AI-Powered National Science Platform

The US Department of Energy (DOE) is planning to build "Genesis," a single national platform for AI-driven scientific research. According to an article in Communications of the ACM, the initiative — framed as going "from the Manhattan Project to Genesis" — is an attempt to fundamentally transform how scientific research is done.

Consolidating AI research infrastructure at a national scale could streamline and accelerate scientific research, but concerns over data governance and privacy are also issues that deserve debate.

---