New Policy Replaces Developers With AI Scripts.
Also a Rocket Mishap and a $60B Office Supply Run

Mandatory Training: The AI is the New Operating System

Andrej Karpathy, former AI lead at both Tesla and OpenAI, has released the mandatory 45-minute video presentation outlining the new corporate development philosophy, which he calls Software 3.0. The core message for developers is that the days of writing explicit C++ or Python logic, what he terms Software 1.0, are over; the new normal is what is being called vibe coding through natural language prompts. The new foundation for all software is essentially one large neural network, which Mr. Karpathy refers to as a new type of computing platform, a utility akin to the electrical grid.

The task for the modern engineer is therefore shifting from writing rules to data curation, which sounds suspiciously like the new, highly-paid job title for the person who used to organize the shared drive. The underlying infrastructure is capital-intensive and centralized, meaning the company will now outsource all intelligence to a handful of providers and pay per-token, like a utility bill. The internal consensus is that this is a great step forward, mostly because it means the codebase is now so complex, nobody can be blamed when it goes wrong; you can simply blame the model’s "jagged intelligence," a perfectly normal and acceptable new euphemism for a bug.

$60 Billion Earmarked for Foundational Office Supplies

Texas Instruments, the calculator and chip behemoth, has committed to investing more than sixty billion dollars into U.S. manufacturing facilities across Texas and Utah. This gigantic capital expenditure is not for the headline-grabbing, sub-3nm AI wafers that dominate the news cycle; it is for "foundational semiconductors," specifically analog and embedded processing chips.

Essentially, the company is spending the equivalent of a small nation's GDP to ensure a stable, domestic supply of the components that allow your refrigerator to know what time it is, or your car to reliably turn on. These are mature-node chips, running at the 28nm to 130nm scale, which is the equivalent of buying a gold-plated, artisanal flip phone in the age of the smartphone. The investment is necessary, dependable, and profoundly unsexy, which is exactly how most IT infrastructure purchases tend to go, except this one is designed to produce hundreds of millions of chips daily.

SpaceX Starship Suffers Internal Paperwork Misalignment

The aerospace division at SpaceX filed an incident report stating that Starship 36 experienced a "major anomaly" while on the test stand at Starbase, resulting in the complete loss of the vehicle. This event, which involved a "sudden energetic event" during the loading of cryogenic propellant for a six-engine static fire test, is another example of a pressurized tank failing to respect corporate safety limits.

Initial forensic analysis suggests the failure of a Composite Overwrapped Pressure Vessel, a part which contained gaseous nitrogen and ruptured below its official stress threshold. The entire affair serves as a frustrating setback, particularly since the Federal Aviation Administration had just confirmed the launch date for the tenth flight attempt just hours earlier. The company, however, maintains that no personnel were injured, and that all hazards remained within the designated safety zone, which is the closest the aerospace industry ever gets to an "all clear" email.

Briefs

• End of Life Policy: A new project called End of 10 is helping users migrate their Windows 10 machines to Linux. It is a noble project; the IT team is still debating whether the increased support tickets for broken drivers outweigh the satisfaction of deleting a legacy OS.
• Quick Acquisition: Wix acquired Base44, a solo-founded company, for $80 million cash six months after it was founded. The CEO, probably still in their early twenties, now has enough capital to pay for their AWS bill indefinitely.
• Alternative Protocols: A developer "Show HN" entry presented a new BitTorrent tracker written in Elixir. The goal is to prove that one can still build extremely robust infrastructure without involving a Large Language Model or a $100 million venture round.