👋 Good Morning! This week’s developments highlight a clear shift in how AI is being absorbed into everyday systems of value creation and control. Consumer behavior is becoming easier to measure through spending rather than hype, creative tools are moving from experimentation toward production reliability, and institutions are adjusting long-standing norms to preserve trust and accountability. The common theme is maturity: AI is no longer proving that it can exist, but proving where it fits, what it’s worth, and how it should be governed.
We’ll be taking a short break until January 5. See you then.
📱ChatGPT Mobile App Surpasses $3 Billion in Consumer Spending
This week, a major commercial milestone arrived for consumer AI: OpenAI’s ChatGPT mobile app has now generated an estimated $3 billion in lifetime consumer spending worldwide, and it did so in just 31 months since its launch in May 2023, a pace that outstrips even TikTok and leading streaming platforms.
The growth curve has sharply steepened this year. In 2025 alone, mobile users spent roughly $2.48 billion, more than four times the $487 million logged in 2024. By contrast, the initial 2023 rollout brought in only about $42.9 million.
Two structural dynamics stand out. First, consumers are willing to pay at scale for mobile AI experiences that feel useful, fluid, and integrated into daily life, from subscription tiers like ChatGPT Plus and Pro to future revenue layers such as in-app commerce or ads. Second, the milestone underscores how mobile remains the primary commercial interface for AI at scale, even if other revenue streams (web subscriptions, APIs) also matter.
In comparative context, the app’s spending milestone came in at half the time it took for top streaming apps like Disney+ or HBO Max to hit the same mark, reinforcing ChatGPT’s unusually rapid monetization trajectory among consumer software products.
The takeaway: consumer willingness to spend on AI is real, not just in trial usage, but in paid engagement at scale, and mobile is where that behavior is most pronounced. As ChatGPT builds out additional monetization vectors (such as its own curated app ecosystem), this milestone sets a high bar for competitors and signals that consumer AI has moved solidly into revenue-driven adoption rather than exploratory usage.
🔧 AI Tools and Updates: Luma’s Ray3 Modify Brings Keyframe-Guided Video Generation
Luma, the a16z-backed AI video and 3D model company, has released a new generative video model called Ray3 Modify that gives creators a much higher level of control over AI-generated footage. At its core, the model lets users specify a start frame and an end frame that guide how transitional video content is generated between those two anchors, effectively letting AI fill in motion and scene changes in a directed way rather than relying on unconstrained generation.
A central challenge in generative video is preserving the continuity of human performance, motion, timing, eye lines and emotive delivery, while introducing transformations. Ray3 Modify addresses this by using character reference images, which preserve performance details from original footage even as the AI generates new content or edits. Creators can change scenes, costumes, or even settings while retaining the fidelity of the recorded performance.
The start/end frame guidance is particularly notable because it lets creators direct how sequences unfold, enabling deliberate transitions and behavior across scenes while maintaining continuity. This moves video AI closer to practical workflows for studios and content teams that need more predictable outputs than unfettered generative models typically provide.
According to Luma, Ray3 Modify blends the expressiveness of generative models with real-world performance capture, allowing teams to shoot once and then leverage AI to adapt that footage for different contexts without re-shooting. The model is available through Luma’s Dream Machine platform, where users can apply it to creative projects alongside Luma’s other video generation capabilities.
💡Quick Hits and Numbers
Oral exams make a comeback: Colleges across the U.S. are reviving face-to-face oral exams to counter AI-assisted cheating and ensure students demonstrate true understanding of course material.
Adobe in ChatGPT: Adobe has integrated Photoshop, Adobe Express, and Acrobat into the ChatGPT interface, letting users edit images, create designs, and manage PDFs directly via conversational prompts.
Denmark deepfake law: Denmark is moving to amend copyright law so individuals can hold legal rights over their own face, voice, and body to curb the creation and spread of AI-generated deepfakes without consent.
🧩 Closing Thought
Taken together, these stories point to AI entering a phase where economic traction, creative utility, and social guardrails are advancing in parallel. ChatGPT’s mobile monetization shows that users will pay when AI delivers consistent, everyday value. Luma’s approach to video generation signals a push toward professional-grade control rather than novelty output. And responses from education systems, regulators, and platform partnerships reveal growing pressure to anchor AI in credibility and accountability. The next chapter of AI adoption won’t be defined by what models can do, but by which uses earn trust, justify cost, and hold up under real-world scrutiny.