steveharing1

Found this and it actually solves something that's been bugging me. Running multiple AI coding agents is great until you're juggling five terminal windows.

Emdash is basically a dashboard for that. Each agent gets its own Git worktree so nothing bleeds into each other, you can see diffs before committing, and you can throw GitHub, Jira or Linear issues directly at an agent.

It Supports Claude Code, Codex, Cursor and 20+ others.

YC backed & 60k downloads already. Worth a look if you're running more than one agent at a time. Has anyone used this in a real workflow yet?

https://github.com/generalaction/emdash

Found this on GitHub and couldn't stop reading. It's from a Texas A&M and YouTube/Google team and what got me is it's not your typical blind upscaler where you just cross your fingers and hope. You actually pick a few keyframes, upscale those yourself however you want, and it propagates your choices across the whole video. That level of control is something I haven't really seen before in VSR. Clever idea honestly. Apache 2 licensed too so worth a look. Has anyone tried it yet? https://github.com/taco-group/SparkVSR

Microsoft's second generation image model hits #3 on Arena.ai, delivers strong photorealism and text rendering, but ships with a 1:1 resolution lock, 30 second cooldowns, and no editing features yet. You can find everything about MAI-Image 2 from its impressive capabilities to limitations here https://firethering.com/microsoft-mai-image-2/