OpenAI's Image Model Now Thinks Before It Draws
OpenAI released GPT-image-2 (ChatGPT Images 2.0) on April 21–22 — the world's first image model with built-in reasoning that researches, plans, and thinks before generating a single pixel. It renders near-perfect text in 12+ languages and scripts (Japanese, Korean, Arabic, Hindi, Bengali), handles scenes with 100+ objects, and produces coordinated multi-format brand assets from one prompt (1:1, 9:16, 16:9, 3:4 — simultaneously). Available to all ChatGPT users now; API opens to developers in early May.
Figma Brings GPT-image-2 Live — and Shares AI Workflows
In the same 48-hour window, Figma pushed two updates: GPT-image-2 is now live inside Figma Design, Draw, Slides, FigJam, and Weave — generate and edit images without ever leaving your canvas. On top of that, Figma Weave workflows (repeatable generative AI pipelines built on a visual canvas) are now shareable on the Figma Community, so you can browse, remix, and publish automated design workflows the same way you share component libraries.
Does GPT-image-2 Actually Replace Designers?
Within hours of the GPT-image-2 launch, VentureBeat and PANews both published pieces asking the uncomfortable question head-on: are designer livelihoods genuinely at risk this time? VentureBeat tested the model generating full infographics, slides, maps, and manga — "seemingly flawlessly." Senior creatives are split: some call it "the last lever before commoditization," others argue the model amplifies art direction but still has no taste without a skilled brief.
AIGA 365 Competition Deadline: 5 Days Away
The 365: AIGA Year in Design — one of the most recognized annual design showcases in the US, juried by leading creative directors — has its final deadline on April 28, 2026 (5 days from today), with a late-entry window closing May 12. The competition accepts work across all disciplines from any designer, including students. Entry fees apply, with costs increasing after each deadline tier.
Brand Kit in One Session — Then Tear It Apart
The combination of GPT-image-2's multi-asset generation and Figma's Weave workflows creates a new workflow challenge this week: can you go from concept to a complete visual identity kit (logo mark, color palette, type pairing, 3 application mockups) in a single 30-minute AI-assisted session? Designers across Threads and X have been posting their attempts this week — results range from generic to genuinely surprising.
You’re done for today!
5 stories · under 10 minutes. See you tomorrow.
TODAY’S THREADThe week's stories share one throughline: the bottleneck in design has shifted from execution to taste. GPT-image-2 can render anything — but it cannot judge what's worth rendering. Figma Weave can automate pipelines — but someone has to design the pipeline. AIGA 365 still rewards the human eye that selected, refined, and positioned the work. As AI tools race toward execution parity, design education is quietly becoming judgment education — and that's actually good news for students who learn to think like art directors, not just make things.