Microsoft Copilot struggles with basic functions.
Also the FBI cannot open the file cabinet.

The Case of the Missing Stapler; Or, Microsoft’s Copilot Is Confusing

The core of Microsoft’s new AI-centric identity, the Copilot chatbot, is reportedly hitting the kind of wall that is usually reserved for a QA build that was accidentally shipped to a Fortune 500 client. User adoption among paid subscribers has tumbled, sliding from a somewhat promising 18.8% down to a less-than-thrilling 11.5% in the last six months, according to a Recon Analytics survey. This suggests a significant percentage of the workforce has decided the new digital assistant is simply not worth the trouble; they have gone back to their old ways, like using search engines or asking the intern.

Current and former employees, in conversations with the Wall Street Journal, pointed toward confusing branding and interoperability issues as major culprits. Microsoft, in a panic that only a mega-corp can produce, is now pouring millions of dollars into television advertising, including a Super Bowl slot, presumably to convince everyone that they actually like the product they already paid for. The underlying problem is that users with access to competitive products like Google Gemini or OpenAI’s ChatGPT prefer those tools; Microsoft is spending $60 million on television ads trying to fix a code problem with a commercial.

Bureaucratic Firewall Stumps Elite Hacking Team

In a classic case of an elite security team being defeated by basic user configuration, the FBI ran into a digital brick wall while attempting to access a seized iPhone. The device belongs to a Washington Post reporter, and the government’s failure to get past the lock screen is attributed to Apple’s "Lockdown Mode" being enabled. This is not a zero-day exploit or a custom hardware dongle from a nation state actor; this is an operating system setting that the user, journalist Greg Miller, turned on himself.

The entire debacle proves that sometimes, the most sophisticated security measure available is simply checking the box that says "Increase Protection," which disables a list of obscure, less-common phone functionalities. One can only imagine the look on the federal agent's face when the multi-million dollar forensics tool returned an error code that essentially translates to Permission Denied: User checked the damn box. The world’s top law enforcement agency is defeated by what amounts to a user set a chmod 400 on their contacts.

Anthropic Launches New Conference Room Named Claude

Anthropic, a large language model company, has announced that their AI, Claude, is a "space to think," which sounds like either a corporate wellness initiative or a newly rented, windowless office cubicle. The company is leaning into the concept of "deep reasoning" by positioning Claude as an assistant that is ad-free and focused on clarity and safety. This is the adult version of being told to go to your room and think about what you did, except the room is a high-compute cluster with a "constitution" written by the company.

The primary marketing boast is that Claude does not have incentives based on attention, such as advertisements; it is built on the idea that an AI should aid thought rather than distract from it. However, for most of the industry, the main product of their AI is distraction, so Anthropic is marketing a feature that most executives see as a bug. We are just pleased that somebody is finally offering an AI that is less fun to use.

Briefs

• B2B SaaS Implosion: A new report suggests AI is actively killing the B2B SaaS model. The market is now being cannibalized by AI agents that can replicate core CRUD functions, essentially automating the internal junior developer who was already doing most of the work for minimum wage.
• Postgres Scaling Issue: A deep dive reveals that the Postgres Postmaster process does not scale efficiently. This is what we call an eternal truth; nothing truly beautiful should scale without causing you significant pain first.
• Zendesk Spam Round: Users are reporting another round of Zendesk email spam. The support ticket software appears to be sending out its own support tickets for fun; a perfect ouroboros of digital noise.
• Eradication News: The Guinea worm is on track to be the second human disease eradicated, with only 10 cases reported in 2025. This is the only actual good news, so it is buried at the bottom because it involves neither GPUs nor venture capital.