Zeitgeist · July 14, 2026
AI Meets the Real World
The paper nobody read
- ▸A grandly titled framework for frontier AI opened with a problem: nobody in the room had actually read it.
- ▸The team compared the moment to people citing Nietzsche and then discovering nobody in the conversation had opened the book.
- ▸Adwait wondered whether waking up and immediately joining an AI meeting was "priming my brain to be the most AI possible." The answer from Nick was an immediate yes.
New York hits pause on big data centers
Link: https://x.com/MTSlive/status/2077034402220322928
- ▸New York's moratorium targets data centers above 50 megawatts. Nick added an important wrinkle: only about 10 to 15 percent of current facilities cross that line, though future builds may get much larger.
- ▸Matthew hated the logic. His analogy was banning buildings over 10 stories before skyscrapers exist, then acting surprised when future capacity gets squeezed.
- ▸The team also understood why the policy lands with voters. Brian summed up the mood as "the whole world hates data centers," which led to a serious idea: answer the public's concerns carefully instead of dismissing them.
The anti-AI font lasts under five minutes
- ▸A moving dot pattern billed as an "anti-AI font" briefly confused the humans too. Matthew compared it to one of those magic-eye images.
- ▸The forever-proof claim fell apart fast. Grok could often interpret it, and Codex decoded Matthew's screen recording in 4 minutes and 52 seconds.
- ▸Nick noted that a person could read it in about two seconds. Matthew's verdict was better: it was anti-AI for exactly 4 minutes and 52 seconds.
Are we measuring AI productivity yet?
- ▸One article credited recent productivity gains to post-COVID digitization rather than AI. Nick found it hard to believe AI deserved no credit, while Matthew argued that broad gains still look narrow outside coding and automation.
- ▸Nick suspected the data misses quieter wins, especially faster emails, messages, and blog posts. Matthew pushed back: AI may create more writing without creating more understanding.
- ▸The room got most excited about agents that can act in a browser. Booking a DMV appointment with a three-word prompt sounds useful, although Alex quickly pointed out that the bot still cannot wait in the DMV line.
Free Claude meets the classroom
Link: https://x.com/claudeai/status/2077047278078931243
- ▸Free Claude access for teachers felt like another step toward making AI ordinary, though the exact meaning of "Claude Premium" on the offer page left the room confused.
- ▸The enthusiasm came with a hard line: students can spot lazy AI output, and teachers who outsource real feedback to a model deserve the pushback they get.
- ▸Alex expects the reaction to swing back as the tools improve and the bad taste fades. For now, students may become the sharpest detectors of AI slop in the room.
Grok's big caveat
Link: https://x.com/hrkrshnn/status/2076716354754015368?s=20
- ▸The latest warning around Grok 4.5 involved reports that user data had been exposed. The team immediately slowed down because the details were unclear.
- ▸Alex asked the question nobody could answer yet: was the data leaked publicly, or uploaded to the service itself? "No one knows" was the only defensible conclusion in the moment.
- ▸Elon Musk's public reaction became part of the story too. Alex joked that he had to take a break from posting about Sam Altman to address it.