Day Old

AI news · Sunday, July 26, 2026

Anthropic’s Opus 5 tops benchmarks as Microsoft struggles with compute

Anthropic just set a new bar for machine reasoning with its Opus 5 model, which cleared the ARC-AGI-3 benchmark with a 30. 2% score—a significant jump over OpenAI’s GPT-5. 6 Sol (Max), which hit just 7.

8%. This lead is attributed to better autonomous planning and self-correction, though some skeptics argue the model may have been tuned specifically for these puzzle formats. While Anthropic celebrates these technical gains, Microsoft is feeling the physical constraints of the arms race.

The company is set to report earnings this Wednesday under intense pressure, as its stock has slid 24% over the last year. Microsoft’s massive $190 billion in capital expenditures—mostly for data centers—hasn't been enough to keep up with demand, forcing executives to prioritize their own internal products like M365 Copilot and GitHub Copilot over paying Azure customers. The situation is so tight that Microsoft has reportedly looked to competitors like Amazon to bail out its struggling infrastructure.

As the company tries to manage these capacity bottlenecks, CEO Satya Nadella is moving toward a more aggressive, fear-driven management style, recently overhauling performance reviews to resemble the old-school “stack ranking” system of the Ballmer era. Meanwhile, a proxy war for the industry's political future is intensifying, with $65 million already funneled into super PACs. Tech heavyweights are splitting into camps: one side, led by OpenAI’s Greg Brockman and Marc Andreessen, is pushing a pro-AI, deregulation agenda through the network 'Leading the Future,' while Anthropic-backed groups like 'Public First' are campaigning for stricter safety guardrails.

Meta is also jumping into the fray with its own state-level lobbying efforts in California and beyond to prevent regulatory fragmentation. This political spending comes as companies continue to link job cuts to AI efficiency; Monday. com is the latest, slashing 20% of its workforce while citing an 'AI-driven growth strategy.

' Despite the corporate anxiety, some developers are pushing back against the tech-evangelism, leading Linus Torvalds to tell holdouts who refuse to use AI coding tools that they can 'go fork themselves,' framing the technology as a technical tool rather than a political statement. Apple, meanwhile, is trying to thread a different needle, delaying its smart glasses until 2027 to ensure it can sell them as a privacy-first product, hoping to avoid the 'pervert glasses' stigma currently dogging Meta.

The quick hits

Sources

Get the day's AI news in one calm read, every day.

Get the app on Google Play Get the app on the App Store Or read today's brief in your browser