An Ultrasound Machine from Midjourney
Midjourney stepped outside its image-generation identity to introduce a new venture: Midjourney Medical. Their device uses sound waves produced by half a million sand-grain-sized mini sensors inside a water-filled tub to complete a full-body scan in 60 seconds. By analyzing the refraction patterns of those waves, it can generate a complete 360-degree map of the body. They’re aiming to make it available at spa centers in San Francisco by the end of 2027. We find it an exciting development: AI finally making a direct impact on other areas of life, presenting an experience straight out of science fiction. Their existing users, who have been pushing for an API for years, had a rather different reaction.
Google & A24 Partnership
Google DeepMind announced a first-of-its-kind partnership with A24, the studio that recently rose to fame with The Backrooms. DeepMind CEO Demis Hassabis said that for this collaboration, AI tools for the creative industry can only be built by working alongside people from that industry. According to the WSJ, the deal involved a $75 million investment from Google, though the details remain unclear. We’re curious to see whether AI will play a visible role in upcoming A24 films.
Claude Introduces @-Tagging
With Claude Code’s new tag feature, you can now assign tasks to Claude directly in Slack or your internal communication tool by typing @claude. This is actually how we’ve been working for a while already, using a Slack-Claude integration. But before the tag feature, everyone had to connect their own Claude account, and Claude would only take direction from the account owner. Anthropic spotted this gap quickly and shipped the fix: Claude is now a first-class participant in team conversations. Any team member can now talk to Claude about the same thread, and Claude can look back through Slack history to catch itself up on what it needs to know.
According to Karpathy, this is the 3rd major shift in how we interact with language models. First we talked to them through chat interfaces. Then we installed apps and started interacting through tools like Claude Code, right where the work actually happens. Now we’re bringing language models into our internal conversations like a new team member, assigning tasks directly and expecting results delivered from the cloud.
Sakana Fugu
Japan has officially entered the language model race with Sakana Fugu. Sakana works as an orchestration layer that routes queries to the best available language models through a single API. Rather than relying on one company’s models, it’s a language model trained specifically to direct queries to whichever model fits best. When it can’t answer something itself, it reaches out in the background to OpenAI, Claude, or other models to construct a response. The core pitch is avoiding vendor lock-in, especially for large enterprises and institutions. It received pushback for being unavailable in the EU and UK, but we can access it from Turkey.
⚡️ Quick Bits
• Superhuman acquired GPTZero, the AI detection startup. TechCrunch →
• SpaceX signed an infrastructure deal with open-source model startup Reflection AI. Reuters →
• Google announced Android 17. Google →
📩 Liked this newsletter? You can forward it to your friends
if you are feeling nostalgic, though it’s always better if
they just subscribe on their own.
