Claude Code takes the wheel
The day's biggest shift is a default setting. Anthropic says that from August 14, "auto mode" becomes the default in Claude Code, its AI coding tool, for customers on Pro, Max, and Team plans. In practice that means the software takes more actions on its own instead of pausing to ask permission at each step. Defaults matter more than features: most users never change them, so this quietly moves millions of coding sessions from ask-first to act-first.
It lands alongside a second change pointing the same direction. Claude Code version 2.1.224 adds a way for separate sessions to send messages directly to one another, according to a report surfaced via Bing News. Put the two together and the shape of the product is clear — less a tool you operate keystroke by keystroke, more a set of workers that act and coordinate. Whether that feels like leverage or like losing the thread will depend on how tight a leash teams keep.
The bill arrives as electricity
The hardware story has moved downstream of the chip. According to the-decoder.com, Nvidia and Amazon are putting billions not into silicon or models but into the power that runs them — AI's appetite for electricity is now a capital allocation problem for the industry's biggest names.
That appetite has a local politics, and it is getting ugly. In Emporia, Kansas, the city has moved council meetings online after death threats against city leaders intensified over a proposed gigawatt-scale AI data center, per Tom's Hardware. A side effect of the move: public comment at those meetings has ended. A project pitched on jobs and investment has, at least here, cost the town the ordinary machinery of arguing about it.
Robots need chips before they need legs
Analysts at Barclays, cited by Yahoo Finance, argue humanoid robots represent the next wave of demand for AI chips and data centres — compute for machines that have not shipped yet. It is a reminder that in this cycle, the demand forecast usually precedes the product.
What the products actually do is more modest. Business Insider reports that several heavily funded robotics startups have converged on folding laundry as the first task for home robots. Towels are not glamorous, but they are a genuinely hard test: soft, shapeless objects that never present the same way twice. If a machine can fold a pile of them reliably, a lot of other household work becomes tractable.
At the other end of the ambition scale, China plans to put robot dogs on the Moon. Interesting Engineering reports the country intends to integrate four-legged robots into the International Lunar Research Station, its future lunar base, by 2035 — the same quadruped form factor now patrolling warehouses, sent somewhere with no atmosphere and no help.
Deals, orbits, and places
OpenAI has acquired NextSlide, a startup working on presentation software, TechCrunch reported on August 8. The most concrete detail comes from NextSlide itself: the team is already building on ChatGPT. As acqui-hires go, the tell is where the people went, not what the product was.
Elon Musk expects SpaceX's Starlink to grow dramatically as AI and robotics spread, according to a Benzinga report surfaced via Google News — the argument being that autonomous systems everywhere need connectivity everywhere, including the places fiber never reached.
Geography is being rewritten on the ground too. TechCrunch traces the remaking of King's Cross, which two decades ago was one of London's seediest districts and a notorious red-light area, into a home for frontier technology. Few urban turnarounds are that complete, or that fast.
And the map keeps widening: the University of Melbourne has signed a memorandum of understanding with the Indian state of Andhra Pradesh to form a knowledge partnership in quantum computing, per BioSpectrum Asia. Quantum diplomacy is increasingly done state to state, not just nation to nation.
One for the other side
Finally, a small piece of counter-programming. A security researcher has built an algorithm that generates patterns designed to make people, faces, and vehicles invisible to surveillance cameras, TechCrunch reports. These are adversarial patterns — images engineered to exploit how vision models see rather than how humans do. On a day defined by AI systems acting more autonomously and watching more of the world, it is worth noting that the same math also works in reverse.