Creative production is going AI-native, from Figma to the terminal

Design is the new code

On the All-In podcast, Figma CEO Dylan Field made the week's sharpest creative-industry argument: as AI coding models improve, the value of design, brand, and emotional reaction is rising as the primary differentiator for software — because code itself is being commoditized. He drew a contrast: AI models show rapid progress in formal domains like math and cybersecurity, while progress in creativity and design remains incremental. And he pointed to the market agreeing with him: Figma grew 48% year over year after Anthropic launched its Design tool. m:…5478

Field's framing, on the show: AI tools are lowering the floor for digital production — coding, motion, design more democratized — but AI is not raising the ceiling of creative output; human intent and focus are. In a market where code is commoditized, he argued, design, brand, and a unique point of view are what differentiate. Figma's own product answer is Weave, a node-based system for connecting AI model outputs. My read: whether or not the thesis is right, it is the thesis the design-tool market is currently pricing in. m:…5478

Consumer photo goes generative

Adobe, meanwhile, is shipping AI where the frustration is most mundane: blinking in group photos. Lightroom on mobile now automatically detects closed eyes and presents an Open Eyes icon as a Quick Action — and the fix can borrow eyes from other photos of the same person in your library. Alongside it, Generative Expand lets users extend the scene beyond the original frame to reframe compositions or adapt images to different sizes. m:…1466

My read: this is the consumer-AI playbook working as designed. The headline-grabbing generative features get the demos, but the features that stick are the ones that fix a ruined photo in one tap. Adobe is betting that AI photo tools land as repairs first, fantasies second. m:…1466

A review desk for AI documents

Google's Gemini Notebook is adding the creative control its Slide Decks and Video Overviews were missing: a Review toggle in the studio panel that opens a review flow before final generation. Users can review outline notes for slides and producer notes for video, steer content and visual style — add an executive summary to the start of a deck, swap the order of a scene in a video — then hit Proceed. Or skip the toggle and generate as usual. The company says the capabilities are rolling out globally over the upcoming week. m:…3018

My read: the interesting move here is steering before generation, not editing after it. The industry spent two years on prompt-and-regenerate; NotebookLM's bet is that creative AI works better as co-creation with a checkpoint — a review desk between the idea and the artifact. m:…3018

Video generation moves to the terminal

The developer surface arrived this week too: Higgsfield shipped an open-source CLI that generates images and video from a terminal, script, or coding agent without opening the web app. It supports 35+ models, the --json flag makes every response machine-readable for scripts and agents, and it integrates with Claude Code, Codex, OpenClaw, Hermes, and Cursor. Pricing is credit-based on the user's existing plan — a 5-second Seedance 2.5 clip at 1080p runs 80 credits for a full render, 15 in draft mode. m:…7609

And the models keep accelerating underneath it all. fal says FLUX 3 Max — post-trained by fal Research in collaboration with Black Forest Labs — ranks #1 for image-to-video quality in Artificial Analysis's preliminary video arena, generating a 10-second clip in about 11 seconds. fal's prompting guide prescribes a five-part prompt structure: camera move, main action, secondary motion, light or time change, and sound. My read on the week: the creative stack now spans every surface — pro design tools, consumer photo, documents, the developer terminal — and the through-line is Field's point. The production floor is automated; the differentiator moved upstairs to taste. m:…9387