A big week for AIC
Happy Monday and hello new subscribers!
Last week was a doozy for AI Clambake. I broke a story over on the AIC website and it has gone somewhat bonkers on the Internet. See below.
Kos
Tom Di Mino, an AI engineer, may have cracked Linear A
Tom Di Mino may have cracked one of the biggest linguistics puzzles out there: Linear A, a Bronze Age writing system discovered on Crete about 120 years ago. His work is currently being reviewed by experts in the field. If it’s verified it will be one of the biggest science stories of the year.
Not much happened for a few days, but on Friday night Boris Cherny (who invented Claude Code) tweeted a link to the story, and Boris’s tweet was retweeted by Marc Andreessen, one of Silicon Valley’s most prominent venture capitalists (and a self-professed enemy of introspection).
Tom did the hard part himself but said Claude Code facilitated his efforts.
Adobe has vastly expanded its AI offering
Adobe has been busy. They have stuck proprietary AI agents into their software. There used to be an agent that sat above the software programs, now there are agents baked in. Adobe also announced that they will build custom AI models for Walt Disney Imagineering’s R&D arm. They also launched Brand Visibility, a generative engine optimization tool. Last but not least, they partnered with LinkedIn to launch AI Essentials for Marketers, a set of role-based courses available in 47 languages. Presumably not Minoan.
It’s initiatives like this that seem to me to pose a serious threat to OpenAI and maybe to Anthropic. Although Anthropic released a new version of Claude Design that seems to offer similar services. They have the sexy LLMs (for now), but legacy companies have existing relationships and distribution. I think OpenAI is going to get MySpaced.
"It's like being surrounded by a million dolphins"
Midjourney, the AI-image generator, is getting into ultrasounds. They announced their new machine will do a full-body scan, and is “as powerful as an MRI,” but only takes 60 seconds instead of 60 to 90 minutes with an MRI. Midjourney will also open spas where you can get it done. #Dolphins
Yoroll.ai lets you create video games from text prompts
Or from photos or videos. Featured titles include “Beyond the Grid” (“Power fades. Legends Rise.”) and “Cyber Trap”!
Brands: they are using AI-generated influencers. Sorry, human influencers.
This practice will be banned in the EU beginning in August, so if you want to get on the AI influencer train, the time is now.
Internal drama at Apple's design team
Not AI related, but a great read regardless. Jony Ive was very influential in the company and had a seat on the executive team. His replacement did not. She instead reported to the chief operating officer, Jeff Williams, who was Tim Cook’s best pal at Apple.
Meta employees are mad as hell and can't take it any more
Between the layoffs, the keystroke monitoring, and the company using their work to train AI models, Meta employees are starting to push back. They, along with other tech workers around the world, are starting to kick the tires on the U word. No, not uxorious. Unionization!
Alarming data point: Since January 2025 almost 400,000 tech workers have been laid off, 150,000 this year. Not clear if that’s global or just US. Regardless: yikes.
TikTok shows 3x more AI slop than Youtube
If you create a new account on TikTok, 59% of the videos it serves you will be AI slop. So, now you know where to look if you’re jonesing for that type of content. Kids are fed a lot of slop. OK, well what about videos tagged as #healthtips? 75% of those are healthslop.