FLUX 3 Image: compose a picture box by box, then prompt each one
Black Forest Labs makes layout a first-class input for image gen. Plus: Copilot code review gets a REST/GraphQL API and GitHub curbs AI-slop security reports.
Copy markdown
Lay out the image, then prompt each box
Black Forest Labs' FLUX 3 Image drops the single-prompt lottery: pick an aspect ratio, draw bounding boxes on a 0–1000 grid, describe what goes in each, and add one scene prompt to tie them together — regional, box-level control instead of re-rolling the whole frame. It makes new weekend builds practical, like a UI-mockup or ad-layout generator where you place every element yourself. Commercial weights are available on request, with an open-weight FLUX 3 Dev variant on the way.
Copilot code review gets a REST/GraphQL API
GitHub made Copilot code review callable through its REST and GraphQL APIs, with Balanced now the default effort level — so you can trigger automated reviews from your own CI, bots, or scripts instead of clicking through the PR UI.
GitHub clamps down on AI-slop security reports
Private vulnerability reports now have daily per-user submission caps plus structured forms that demand a proof of concept and specifics — GitHub's answer to the flood of low-quality, AI-generated security filings hitting open-source maintainers. If you run a public repo, your triage queue should get quieter.