Google DeepMind posted a new video today. It shows Gemini 4 Argon, the company's newest AI model, building a 3D periodic table. The table syncs to a mid-century Tom Lehrer song called The Elements. That song names every known chemical element in a fast patter. Gemini built the whole scene inside Antigravity. Antigravity is Google's tool for building apps with AI help. The table shows electrons orbiting each atom. It shows atomic data next to each element. It shows the light each element gives off. All of it moves in time with the song.

The post earned 766 likes in about ninety minutes. It drew 114 replies in that same stretch. Those numbers trail Google's biggest launches. They still show a quick, early crowd. The crowd gathered around something playful.

Chemistry teachers have used Tom Lehrer's song for decades. It turns a dry list into something people remember. Pairing it with Gemini marks a shift. Google wants people to see its AI model as a teaching tool. A model that can answer questions is one thing. A model that can build a class demo is another.

The choice says something about how labs show off new models now. A raw benchmark score rarely spreads far on its own. A finished, playful demo spreads faster. DeepMind built something a teacher could show a class tomorrow. That differs from a leaderboard win.

Wikipedia traffic for Tom Lehrer sat at 509 views the day before the post. Traffic for the Gemini chatbot page sat at 149 views that same day. Both pages held steady. The post is still fresh. A mid-century song and a model built this year now share one screen. That pairing carries the whole idea behind the demo.

Nvidia posted a short teaser today. It points to a keynote tomorrow from CEO Jensen Huang. Huang speaks at NVIDIA's own conference, called GTC, in Berlin. GTC stands for GPU Technology Conference. Nvidia has run GTC stops in four European cities this month. Each stop built toward the same message. Europe wants to build its own AI infrastructure. Nvidia wants to supply the chips for that build.

The post earned 132 likes in under half an hour. It drew 34 replies in that same stretch. Those numbers are small next to Nvidia's biggest posts. A single teaser rarely draws a crowd on its own. The bigger story sits in the pattern behind it.

Nvidia posted about a partnership with Mercedes-Benz last week. It posted about grid-aware AI factories the same week. It posted about industrial AI work in Germany days later. Each post built toward this week's keynote in Berlin. Together, the posts add up to one message aimed at one audience.

Jensen Huang draws real public interest on his own. His Wikipedia page drew 7,696 views on one October day. It drew 7,313 views the next day. Those numbers hold steady day to day. People already watch Huang closely before he speaks. A CEO with that kind of pull needs fewer ads to fill a room.

The steady run of posts shows a company working one region hard. Nvidia wants Europe to see it as a partner first. It wants Europe to see it as a chip seller second. Chips sell faster when buyers trust the seller. That kind of trust takes more than one post to build.

A four-city tour builds that trust in public. Each stop adds proof before the next one starts. Berlin closes out this month's run. Huang carries four weeks of goodwill onto one stage tomorrow. Europe's AI buyers are watching the whole run. The keynote is just one stop in it.