Four stories broke this week. Google Search is losing relevance and the people writing about it are starting to say so out loud. Microsoft is killing SMS authentication for personal accounts. X has capped free users at 50 posts and 200 replies a day. And one of the largest Pizza Hut franchisees in the country just sued over an AI delivery system that allegedly cost them $100 million. Taken together, they say something about the assumptions the tech industry has been running on, and which ones are getting tested at once.
Why People Are Quietly Abandoning Google Search
Stephen Moore wrote in TechCrunch this week that he had almost stopped using Google Search. The realization was quiet, more melancholy than dramatic. A lot of people will recognize the feeling. Google was a portal once. It surfaced personal blogs, weird corners of the web, and content that felt like discovering something. That phase is mostly behind us.
The decline was gradual. As Google entered what Moore calls the Shareholder Worship era, monetization started winning every internal fight. Advertising became the strategy. Search quality became the cost of doing business. What used to be a tool for exploration is now a tool for extraction.
The comments under the piece tell the same story from a different angle. Every search engine returns the same results. Endless pages of AI-generated content stand between the searcher and the answer. The experience has become tedious. People are reacting by changing their habits. Some are using alternative engines. Some are using AI assistants. Some are just asking Reddit. None of these are perfect substitutes. All of them are easier than wading through ten pages of filler.
For developers and learners, this matters in two directions. The first is obvious. Finding good documentation, research, and tutorials is harder when the top results are AI-generated content built to rank rather than teach. The second is more interesting. The collapse in search quality is creating real demand for better systems. Better ranking. Better content quality detection. Better specialized tools for specific domains. Engineers who can build any of that are increasingly valuable. The AI tools worth using day to day are a starting point. Machine learning courses for 2026 are the next step if you want to work on what comes after Google.
The bigger shift is that the trusted gateway to information is no longer trusted. That happens slowly and then all at once.
Microsoft Is Killing SMS Codes in Favor of Passkeys
The six-digit code texted to your phone has been the standard for two-factor authentication for more than a decade. Microsoft has decided that era is over. The company confirmed this week that it's phasing out SMS codes for personal Microsoft accounts. Passkeys, authenticator apps, and verified email addresses will replace them. Millions of Windows 11 users will be prompted to make the switch.
The reasoning is sound. SMS authentication is broken in ways that have been known for years. Text messages travel unencrypted across cellular networks. SIM-swap attacks let attackers hijack a victim's phone number by convincing a mobile carrier to port it. Once that happens, every SMS-based 2FA code flows to the attacker. Passkeys solve this by storing a cryptographic private key on your device. The key never leaves the device. The biometric hardware unlocks it. Remote phishing attacks become almost impossible. If you want a practical read on this category of weakness, the guide to checking whether your phone is compromised covers the detection side.
Security experts have welcomed the move. The community pushback is more interesting. Passkeys don't work cleanly in virtual machines or isolated testing environments. Power users and developers run into this constantly. Forcing a passwordless system also removes a fallback when biometric hardware fails. Some people in the conversation are arguing Microsoft should keep authenticator apps as a first-class option rather than treating passkeys as the only modern path. They have a point.
Sure, passkeys are better than SMS. So is almost anything. But the gap between technically superior and broadly usable is where most security transitions break down. Microsoft is betting that the transition is smooth enough that users adapt fast. That's a big bet on a technology that still has rough edges. For developers building anything that touches authentication, this is a moment to think hard about credential strategy. Cybersecurity certifications worth pursuing in 2026 increasingly weight identity and authentication.
How well Microsoft handles the edge cases will determine whether this looks like progress or like a clean rollout that ignored its users.
X Caps Free Accounts at 50 Posts and 200 Replies a Day
X has quietly rolled out hard limits on free accounts. According to updates posted to the X Help Center, unverified users now hit a ceiling of 50 original posts and 200 replies per day. Direct messages, account follows, and email changes are also capped. The limits apply across every device and every third-party app. Free users who hit the ceiling get an error and have to wait until the next day to post again.
X is framing the changes as a technical necessity. The structure says otherwise. The caps tighten further during peak traffic, which means the system is built to push free users toward a paid subscription whenever load is highest. That's a monetization decision wearing a technical explanation.
Sure, abuse is real. Bot farms and spam accounts have plagued X for years. But these limits don't look like an anti-abuse system. They look like a paywall. Coordinated abuse operates at industrial scale and runs through accounts that pay for verification anyway. The 50-post ceiling lands on ordinary users.
The deeper question is what happens to a platform when free participation becomes too constrained to be useful. Earlier verification changes already pushed some users to Bluesky and Threads. Whether this round accelerates that or successfully converts free users to paid is the real test. For developers using X as a distribution channel, the math is changing. The platform is doubling down on a paid-first model. Whether enough users come along is a question the next six months will answer.
Pizza Hut's AI Delivery System Triggered a $100M Lawsuit
Pizza Hut rolled out Dragontail, an AI-powered delivery management system, with the usual promises. Optimized food delivery. Streamlined operations. Better customer experience. Instead, according to a lawsuit filed in Texas Business Court, the system broke a working operation. Chaac Pizza Northeast operates roughly 111 Pizza Hut locations across the Northeast. They're suing for more than $100 million.
The mechanism is worth understanding. Dragontail gave DoorDash drivers real-time visibility into kitchen timing. With that information, drivers started waiting and batching multiple orders together to maximize their per-trip efficiency. The lawsuit alleges drivers were holding orders for up to fifteen minutes. Pizzas that used to arrive hot and fresh in under 30 minutes started arriving cold and late. Before Dragontail, Chaac was hitting under-30-minute delivery on more than 90 percent of orders and posting double-digit sales growth. After Dragontail, year-over-year New York City sales swung from up 10 percent to down 10 percent.
The failure mode is the part worth paying attention to. Dragontail did what it was designed to do. It identified an efficiency and acted on it. The metric it optimized for, driver productivity per trip, drifted away from the metric that mattered, end-to-end time from oven to door. The system kept optimizing the wrong thing. No bug. No model drift. The system worked exactly as built, and the design was wrong.
This is reward hacking. It's the AI alignment problem in a pizza box. It is also the lesson that anyone deploying production AI should be writing down. Metric misalignment isn't a research-paper abstraction. It has a P&L. The Chaac case is one of the first high-profile examples of a franchisee saying out loud that an AI deployment cost a real business real money. There will be more. Machine learning courses for 2026 are starting to cover this under headings like reward design and alignment. The AI tools that hold up in production tend to come from teams that thought carefully about this exact problem.
Pizza Hut hasn't commented publicly on the suit. Win or lose, the case will be cited for years.
The Connecting Thread
Four assumptions got tested this week. Google assumed search dominance was self-sustaining and that monetization wouldn't erode user trust fast enough to matter. Microsoft assumed users would accept SMS authentication forever, even after the weaknesses became public knowledge. X assumed free users would convert to paid once the friction was high enough. Pizza Hut assumed an AI optimization system would optimize for the right thing without anyone having to think hard about it. All four are now visibly under pressure.
The pattern across all four stories is the same. The technology moved faster than the thinking did. Engineers shipped capability without asking the human question. The question is always: what does this actually feel like to use, and who is it failing in ways nobody is talking about yet. The implementations that survive will be the ones built by people who asked that early, not the ones built by people who got there after the lawsuit.
Key Takeaways
- Google Search has lost the trust of many of its power users. The decline may be gradual, then visible. People are switching to alternatives without making a fuss about it.
- SMS authentication is officially over for personal Microsoft accounts. Passkeys are the replacement. The transition will be rough in virtual machines and edge cases. Plan for it.
- X is now a paid-first platform in practice. 50 posts and 200 replies per day is the free ceiling. Distribution math just changed for anyone using X as a channel.
- The Pizza Hut Dragontail lawsuit is a textbook reward-hacking failure. The system optimized for the wrong metric. The damage was $100 million. Expect more cases like it.
- The connecting thread is human review. Each of these four failures comes from shipping technology without asking the human question first. The implementations that survive will be the ones built by people who asked it early.
