Presented by
Good morning, AI enthusiasts. Hollywood finally made a film about the most chaotic week in AI history. Then the studio that funded it decided no one should see it.
Amazon shelved its finished Sam Altman biopic just months after putting $50B into OpenAI. When your new business partner is also the movie's villain, who gets final cut?
In today's recap:
Amazon buries its own Sam Altman biopic
OpenAI's o3 cracks 18 unsolved rare-disease cases
Run Claude Code on GLM-5.2 to cut costs
Lutnick forces Anthropic to cut foreign Fable 5 access
4 new AI tools, prompts, and more
AMAZON & OPENAI
Amazon shelves its own Sam Altman biopic
Getty Images
Recaply: Amazon MGM just dropped "Artificial," a finished film about how Sam Altman got fired from OpenAI and won his job back. It did this months after putting $50B into the company.
Key details:
The film stars Andrew Garfield as Altman and Ike Barinholtz as Elon Musk. Director Luca Guadagnino plays the boardroom drama like a comedic take on The Social Network.
Amazon spent about $40M to make the movie. Then it shelved the film months after pledging $50B to OpenAI, part of a wider $100B, eight-year cloud deal.
Studio boss Mike Hopkins reportedly killed it after watching a cut. The Daily Beast notes Amazon would face a clear conflict showing a big new partner in a bad light.
Amazon confirmed on June 19 it won't release the film. Netflix, Focus Features, and Warner Bros already passed, and now A24 and Mubi are circling.
Why it matters: People keep worrying about AI swallowing media. Here's an early case, just not the kind they pictured. Amazon didn't kill the film with bad reviews. It shelved a finished, well-tested movie because an unflattering Altman story got awkward once OpenAI became a $50B partner. When the same firms fund both the models and the stories about them, the conflicts write themselves.
PRESENTED BY NEO
AI help, without the trust tax.
Most AI tools ask you to trade your data for intelligence. Norton Neo doesn't. It's the first safe AI-native browser built by Norton, and it gives you powerful built-in AI without handing your privacy over to get it. Search, summarize, and write with AI built directly into your browser. Your data stays yours. Your context stays private.
Built-in VPN, anti-fingerprinting, and ad blocking come standard. No add-ons. No setup. No compromises.
Fast. Safe. Intelligent. That's Neo.
AI RESEARCH
OpenAI's o3 cracks 18 unsolved rare-disease cases
OpenAI
Recaply: OpenAI just helped doctors solve 18 rare-disease cases that had gone unsolved for years. Its o3 model re-read hundreds of patient genomes and found leads experts could check.
Key details:
Researchers gave the o3 Deep Research model de-identified genome data, symptom terms, and a short list of gene variants. They asked it to name the most likely cause and show its work.
The model re-checked 376 cases that specialists had already failed to crack. It surfaced leads behind 18 confirmed diagnoses, a 4.8% gain on top of earlier expert review.
In one case, researchers say o3 spotted a missing piece of chromosome 22 that wasn't in the input data. It tied the gap to DiGeorge syndrome, later confirmed by a follow-up scan.
The study ran with Boston Children's Hospital and Harvard and was published June 18 in NEJM AI. The model only suggested ideas. It never made a diagnosis.
Why it matters: Plenty of AI health demos overpromise. This one came with peer review and a hard limit. The model never diagnosed anyone. It read across scattered records and changing research faster than any person could, then handed leads back to doctors. For families like Kyra's, who waited nearly 20 years for an answer, that speed is the whole point. Periodic AI re-checks could turn cold genetic cases into solvable ones.
GUIDES
Run Claude Code on GLM-5.2 to cut coding costs
Recaply: In this tutorial, you will learn how to point Claude Code at Z.ai's GLM-5.2 model through its Anthropic-compatible API, keeping your exact Claude Code workflow while paying a fraction of frontier-model prices.
Step-by-step:
Install Claude Code with Node.js 18 or newer, using nvm on macOS to avoid permission errors. Run "npm install -g @anthropic-ai/claude-code", then confirm it works with "claude --version". If it's already installed, run "claude update" instead.
Get a Z.ai key on a coding plan. Sign up at z.ai/model-api and buy a GLM Coding Plan (Lite, Pro, or Max, from $18 per month), then create and copy a key at z.ai/manage-apikey. Pay-as-you-go keys won't work here.
Open ~/.claude/settings.json and add an "env" block with "ANTHROPIC_BASE_URL" set to "https://api.z.ai/api/anthropic", "ANTHROPIC_AUTH_TOKEN" set to your key, and "API_TIMEOUT_MS" set to "3000000". To skip the manual edit, run "npx @z_ai/coding-helper".
Map the models to GLM-5.2. In that same env block, set "ANTHROPIC_DEFAULT_OPUS_MODEL" to "glm-5.2" and keep "ANTHROPIC_DEFAULT_SONNET_MODEL" and "ANTHROPIC_DEFAULT_HAIKU_MODEL" on "glm-4.7". GLM-5.2 isn't the default, so this step is what switches it on.
Open a new terminal so the variables load, go to your project, and run "claude". Type "/status" to confirm glm-5.2 is active, then use "/effort" and pick "max" for hard, multi-step coding tasks.
Pro tip: GLM-5.2 burns quota two to three times faster at peak hours, so keep the Sonnet and Haiku slots on glm-4.7 and save glm-5.2 with "/effort max" for your hardest tasks.
TOGETHER WITH STORECLAW
Stop Paying for 6 Tools. One AI Does It All.
Most e-commerce sellers juggle 6–8 tools and pay hundreds monthly to keep operations running. StoreClaw replaces the stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks profit 24/7. Connect your store and let AI handle the work — no prompts, no complex setup, no credit card required.
TOOLS
Trending AI Tools
🤖 Sakana Fugu - Multi-agent system that assembles and coordinates frontier models per task through one API, with an Ultra tier for hard problems
🧠 OpenRouter Fusion - Fans one prompt across a model panel and synthesizes a single answer, topping Fable 5 on DRACO at about half the cost
🎬 Palmier - AI-native macOS video editor with Kling, Seedance, Veo, and Grok Imagine built into the timeline, $29 per month Pro
NEWS
What Matters in AI Right Now?
Sakana AI launched Fugu, a multi-agent system that learns to assemble and coordinate frontier models per task through one API. Its Ultra tier scores 73.7 on SWE-bench Pro, pitched as frontier power without export-control risk.
Meta signed new deals to buy about 1.6 gigawatts of AI computing power from data-center developer Crusoe across Texas and Missouri, part of its pledge to invest $600B in US infrastructure over three years.
Google lost two of its biggest names in a week, with AlphaFold Nobel laureate John Jumper joining Anthropic and Transformer co-author Noam Shazeer joining OpenAI as Lead for Architecture Research.
The Atlantic uncovered four datasets holding more than 21 million tracks used to train AI music models like Suno and Udio, then made them fully searchable. Reporter Alex Reisner found work from Radiohead, Lady Gaga, and Bruce Springsteen inside.
Commerce Secretary Howard Lutnick ordered Anthropic to suspend exports of its Fable 5 and Mythos models to all foreign nationals worldwide, the first use of 2018 Export Control Reform Act powers on an AI model. Anthropic cut global access the same day.
Tesla filed a trademark for "Megapod," a modular, self-contained AI data center system it plans to sell externally, bundling servers, networking, power, and cooling. The move comes less than a year after Tesla killed its in-house Dojo supercomputer.
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
