AI agents are moving from answering questions to doing real work across teams. that’s where productivity gets interesting: fewer steps to manage, more time for the work that needs human touch. 🤖
the 10 reference limit is a really useful upgrade for keeping characters consistent across shots. that’s one of the biggest things that makes AI video harder to use in real projects. excited to see what creators do with 30-second generations 👀
Kling 4.0 is now live on ElevenCreative.
More control over every shot with improved motion quality:
- Native generations up to 30 seconds
- Up to 10 image and video references to keep characters and style consistent
- Improved motion, character consistency, and cinematic storytelling
Try it now: elevenlabs.io/video/kling-4
Sarah B. retweeted
the problem
a long migration doesn’t fit in one agent chat. the context fills up and you start over.
the fix
cursor launched projects in beta on september 10. you talk to a coordinator agent that plans the work and hands it to other agents. the coordinator doesn’t write code itself. the project runs on its own cloud computer, so closing your laptop doesn’t stop it, and it keeps shared context files every agent can read.
action
open the agents window, pick projects in the sidebar and click new project. name it after the outcome, like “checkout redesign”, not the repo. pick the repo and the coordinator’s model. make your first message research only: ask for a plan, a conventions file and a how-to-test file before any code.
tradeoff
pro: when one agent works out how to test a service, later agents reuse those notes.
con: cursor’s docs say five parallel subagents use about five times the tokens of one, and simple tasks can run slower. the “30% more prs merged” figure is cursor’s own.
takeaway
use it for work that takes weeks. a one-hour bug fix doesn’t need a coordinator.
the full-project cding and 3D modeling use cases caught my attention 👀 if the reliability improvements hold up in real workflows, this could be a really useful opion for people building with AI every day.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
the useful part of google’s new gemini agent is that it can work across your business tools and bring back something finished. that’s the kind of AI workflow i want more of: less copying between apps, more time for the work that needs a human brain.
one song on repeat for 3 hours because apparently my brain has decided this is the only music that exists today 🎧😂
my favorite AI tools are the ones that save me 20 minutes without creating 30 minutes of cleanup. that's the bar tbh 😅✨
apple and meta are teaming up so the muse ai model can run on mac computers.
here's why that matters if you're not a developer: if a model runs on the mac itself, your drafts, notes and files don't have to go to a server for it to help you. i haven't tested it yet, so i'm not calling it good or bad. when it ships, try it on one low stakes task first, like summarising your own meeting notes, and see if it holds up.
love this direction from perplexity. i’d feel a lot better letting an agent work with my files if everything stays local.
one thing i keep coming back to with AI tools: impressive demos are easy to remember, but the boring time-savers are the ones that stick.
if a tool can handle meeting notes, organize research or remove repetitive admin from your day, that's a real win.
the goal isn't to use more AI. it's to make work feel a little less exhausting.
people ask me which AI tool they should start with.
honest answer: the one that helps with a task you do every day.
for me, it was meeting notes. i spent 20 minutes after every call writing summaries. otter.ai took that off my plate.
Sarah B. retweeted
meta just released muse spark open, and the label “open weight” is worth understanding.
it means the trained model files are public, so you can download and run it yourself instead of only using it through a website.
this one has 816 billion parameters, which is the model’s size, but only 42 billion are used for each request. think of a huge team where only the right specialists answer each question.
it can run on two Apple M5 Ultra Mac Studios, so a normal laptop is out.
meta’s license also limits commercial use, so check the terms before building a business on it.
research.meta.ai/blog/muse-s…
my favorite AI tools are the ones that remove a step from my day.
not the ones that give me 12 new things to configure, 3 dashboards to check and another workflow to maintain.
if a tool saves me 20 minutes without creating 15 minutes of cleanup, its earned a spot in my routine.
small AI productivity tip: stop asking it to do everything in one giant prompt.
give it the task, your context, and what a good result looks like. then refine from there.
less prompt gymnastics, better output.
Sarah B. retweeted
In February Alphabet sold a £1B bond that matures in 2126, at a 6.125% coupon with £9.5B of orders. It was the first 100-year bond from a tech company since Motorola in 1997
By July it had traded below 90p
Alphabet, Amazon, Meta, Microsoft, Oracle and SpaceX had issued over $180B of USD bonds this year as of early September. 41% of hyperscaler issuance since 2025 runs past 15 years, against 16% for the US corporate index
The servers get depreciated over 5–6 years. The longest bonds funding the buildout run 30 to 100
this is the kind of AI workflow i love seeing. it didn’t just generate a design, it handled the CAD, browser automation, and final video too. that’s a lot of separate steps off someone’s plate.
















