Getting a fight scene to look right isn’t just about the prompt. The characters’ positions, timing, and camera angles all matter. I tried using Pippit’s 3D Director Studio to block out a two-warrior fight scene before generating it with Seedance 2.5. Being able to figure out the scene in 3D first gave me a clearer idea of how I wanted the final shots to look. Here’s how it turned out ⚔️

Your next edit just got cheaper. Save 60% on Pro or 40% on Ultra, Oct 20–30. CapCut Desktop only. I tested a Codex + CapCut PC AI video editing workflow using the Jean Phil trend. #capcutpc

$0.003 per second for video generation, without compromising on speed or visual quality? The supplied comparisons put Vivix W1 alongside leading models on visual quality. It also delivers eight seconds of video in roughly eight seconds. But the bigger story is that W1 is a real-time interactive video model. Every extra minute of interaction means more video to generate and pay for. Making generation this affordable could make interactive stories and live visuals practical to run, not just impressive demos. Live visuals are what I’m most curious about. Generating quickly is one thing; reacting at the right moment is what would make them feel live. @VivixLabs_HQ already offers real-time generation through the API; the browser playground currently produces finished clips. A closer look at what could power those live visuals 👇 platform.vivix.ai/vivix-w1-m…

Deel just shared something pretty wild: $140M+ in ARR added in 90 days without increasing headcount. Behind it: 8,000+ AI agents doing the work of ~600 employees. Akai was built internally to automate Finance, HR, AP, Compliance and other repetitive operations. 600 employees’ worth of monthly hours automated by AI agents Akai was an internal tool built at Deel ($17.3B valuation) to automate back-office operations. You record your screen while doing a task. It watches how you work, learns the workflow and edge cases, then runs it for you. → Works across your existing tools, even legacy systems with no APIs → Workflows can be changed in plain English → Teams can share and improve workflows → Agents keep learning as more cases are handled The results Deel shared: → 85% of payment reconciliation automated → 500+ hours/week of manual work eliminated → Revenue per employee: $130K → $215K Deel's putting a 1,000-hour Automation Guarantee behind it. If you're a 100-person company and they can’t automate 1,000+ hours in 30 days, you get a full refund. If this fits your description, I'd book a Demo ASAP: akai.run $5000 free Akai credits btw.

I spent an afternoon inside openJiuwen's WorkSwarm, an open source multi-agent platform under Apache 2.0, and one test result stuck with me. 5 people shared one agent for 189 turns. It still caught all 8 cross-role conflicts, because Persistent Session tracked who actually had authority to change a decision, even after older context got compressed. I've seen agent setups lose that thread by turn 20, sometimes sooner. Repo 👉 github.com/openJiuwen-ai/jiu… #AI #Agent #openJiuwen #WorkSwarm

This is the part of agentic AI that doesn’t get enough attention. Reasoning is one thing. Actually working across spreadsheets, docs and presentations without constantly losing context is another. Univer is building for that layer.