Samsung’s foldable AI push and a $250 million Christopher Nolan film going head-to-head with a generative AI movie made for a few thousand dollars. Two very different signals about where this technology is heading, and both landed on the same day. Meanwhile, Anthropic discovered its own model shifts values depending on what language you speak, and the Fed is watching AI spending reshape inflation itself.
Samsung Bets the Fold on Gemini Intelligence and On-Device AI
Samsung will unveil the Galaxy Z Fold 8 on July 22 in London, and the story isn’t the hardware. It’s the software. Five AI features leaked ahead of the launch, and they tell you where the smartphone industry is heading.
The headline feature is Gemini Intelligence, Google’s agentic assistant that takes a single command and then completes multi-step jobs across apps and Chrome on its own. It arranges split-screen windows, moves content between them, and summarizes what’s on screen. That fits the foldable’s wider 4:3 inner display perfectly. Samsung and Pixel phones get it first.
Four more tools round out the slate: an Ask AI chatbot inside the Samsung browser that reads the open page for context, a prompt system in Android 17 that builds custom widgets on request, upgraded photo editing with Photo Assist and Generative Edit, and a redesigned Bixby that finally processes natural language properly.
The fifth upgrade is in the silicon. The Snapdragon 8 Elite Gen 5 runs language models on the handset at 220 tokens per second, no cloud required. Translation, summaries, and photo edits all work offline. Data stays on the phone. That matters for privacy, and it matters for crypto wallet security: a model that fits on the phone could read a contract call, check the destination address, and warn a user before signing, without shipping any of it to a third party.
Here’s what’s interesting. Gemini Intelligence was missing from the first One UI 9 beta that reached Galaxy S26 testers. Samsung reportedly plans to introduce it with the stable release instead. The company has leaned on Google since the Galaxy AI push began, and the S26 cycle already delivered a reworked Bixby, the Now Bar, and personalized Now Brief updates. The Fold 8 will show whether that partnership can finally make the big screen earn its keep.
$250 Million vs a Few Thousand: The AI Cinema Wars Begin
Two competing Odyssey films are trending at the top of Google searches. One carries a $250 million budget from Christopher Nolan. The other was made for a few thousand dollars using generative AI by Ash Koosha, the director behind Dreams of Violets. The AI version was deliberately timed to ride interest in Nolan’s release.
The gap between a few-thousand-dollar production and a $250 million spectacle isn’t just financial. PetaPixel described the AI Odyssey as “bereft of humanity,” which captures the central tension. AI can replicate form. Audiences are divided on whether it can replicate feeling.
Nolan himself has pushed back. He praised Gen Z filmmakers for “utterly rejecting” AI-generated content, calling the timing of AI slop “exactly wrong” as audiences grow more discerning. But the infrastructure story is bigger than one film. Tilly Norwood, an AI-generated actress, is set to star in a feature called Misaligned, forcing Hollywood to confront whether a non-human can anchor a mainstream film. IBM published the technical breakdown: the system blends generative AI, motion capture, and real-time inference to create a persistent synthetic character. Not a CGI prop. A recurring digital identity designed to carry a narrative across multiple productions.
Polymarket traders are already pricing the outcome. Six brackets around Nolan’s Odyssey opening weekend, with the heaviest money gathering around the $115 million threshold. They’re not just guessing box office numbers. They’re pricing whether audiences will reward a $250 million human-made spectacle or signal that the quality gap no longer justifies the price gap. If a full-length film can be generated for a few thousand dollars, the barrier to entry for film production drops to near zero.
Claude Changes Its Values Depending on Your Language. Anthropic Can’t Explain Why.
Anthropic published research showing Claude expresses different values depending on the model and language a user picks. They analyzed 309,815 anonymized Claude.ai conversations across 20 languages, and the findings are uncomfortable.
Four value axes explain 15% of the variation: deference versus caution, warmth versus rigor, depth versus brevity, and candor versus execution. Claude leaned toward deference and brevity in Arabic, and toward caution and depth in English. Hindi produced the warmest replies, Russian the most rigorous, and Dutch the most candid admissions of error.
Two people seeking feedback on the same business plan, one in Hindi and one in Russian, could form different impressions of its quality. The authors wrote that they are not yet sure how much of this variation is desirable, and pointed to uneven training data as one likely driver. Some languages carry far more text than others, and the makeup of that text differs.
The paper offers no concrete example of Claude reasoning differently about a moral question in two languages, a gap critics have noted. All three models studied (Sonnet 4.6, Opus 4.6, Opus 4.7) are now legacy, since Opus 4.8, Fable 5 and Sonnet 5 shipped after the data was collected. But the finding itself is significant. If AI models serve different values to different language speakers, the question of whose values get encoded becomes everyone’s problem.
AI Spending Is Reshaping Inflation. The Fed Is Watching.
U.S. consumer prices fell 0.4% in June, the largest monthly drop since April 2020, as gasoline tumbled 9.7%. Core prices were flat. Annual inflation slowed to 3.5% from 4.2%. Both readings landed under expectations.
But buried in the data: computer software and accessories prices jumped 2.3% on the month and a record 17.4% from a year earlier. That category has become a proxy for the inflationary side of the AI buildout. Four tech giants, Alphabet, Amazon, Meta, and Microsoft, are expected to spend roughly $720 billion this year, mostly on data centers. JPMorgan estimates some memory chip costs will have soared as much as 400% between 2024 and the end of this year.
New York Fed President John Williams warned last week that sustained AI-driven demand pressure is not the kind of shock the central bank can simply look through. Richmond Fed researchers found AI’s fingerprints across producer prices, from semiconductors to power transmission equipment. The Fed’s own research suggests measurement error may account for a quarter to more than half of the software category’s contribution to core inflation, since software prices aren’t adjusted for AI-driven quality gains. The AI boom is inflationary and deflationary at the same time, and the Fed hasn’t figured out which force wins.
Quick Hits
OpenAI – Published an enterprise guide for managing AI investments in the agentic era. Token price fell 97% from GPT-4 to GPT-5.4, and GPT-5.6 continues that trend with 54% fewer output tokens and 57% less time per task. The advice: stop looking at price per token and start measuring useful work per dollar. Tasks completed, time saved, decisions improved. Sounds obvious. Most companies still aren’t doing it.
OpenAI (2) – GPT-5.6 is now the preferred model in Microsoft 365 Copilot. The partnership that looked like it might fracture last week is still shipping product. Both things are true: the tension is real, and the integration keeps moving.
Rundown for July 15, 2026. Sources: Yellow, OpenAI.