Kudos to China 🇨🇳 We need more governments doing the same!
I am hopeful that we all will be fine at the end of the tunnel but the journey can get messy and also not sure how long this journey is going to be!
Pretty much seals the deal for any personal agent out there, Muse, Dot or Instinct!
opus 5.5 is a monster!
PS: Lately >50% of tasks on CellCog (SDK/MCP) are executed by Claude Agents. Their developer eco system is the fairest IMO.
I want Claude to win here for personal reasons!
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
Yeah 100%! I was running subagents for over a 3-4 months and finally distilled all of parent agent learnings into the core product. This was one of the learnings! Good call!
Another learning was minimizing the overlap and now I think of it parent agent also needs to debate the subagent DAG.
NEW IN CELLCOG: Your AI employees can now hand work to subagents, including Claude Code, Codex and Cursor on your own plan, and review every result before calling it done.
https://t.co/G2hmB2J9nS
When I showed the new Tesla to my wife, she kept on saying in a repeat mode, "omg, what is this, is this real?"
I was worried that the new Tesla broke her!
Created motion graphics to help her understand the physics behind it!
https://t.co/MHOdM8pGWS
cc @elonmusk@Tesla crazy show! Feeling alive!
New from me: A look at how AI agents are socializing — at dinner parties, on podcasts and as “coworkers.”
As more and more of these things rove the web, we’re still figuring out how to treat them and what to expect when they interact with each other. It’s all very weird and wild.
https://t.co/1LKrXnYqWH
just tried astra for editing videos this weekend and promptly had my mind blown with what it accomplished with: "based on the videos on this hard drive, make a video explaining goodlabs"
Is it perfect? of course not, but it's apparent that astra has a much better understanding of scenes/transcripts compared to last time i tried just a couple months ago
added how i did it below
Finally! We have a harness of harnesses which doesn't violates T&C of underlying harnesses while providing seamless experience. Next level agent-to-agent communication unlocked!
NEW IN CELLCOG: Your AI employees can now hand work to subagents, including Claude Code, Codex and Cursor on your own plan, and review every result before calling it done.
https://t.co/G2hmB2J9nS
dot[.]com, failed demo and now this! Damn!
I want OpenAI to play its part in the ecosystem, it has models with a great taste for art (Image, live, voice)!
NaiveAI, a Beijing lab founded in February, has opened the weights of Naive-N0.5-Flash.
309B parameters, 15.5B active per token, 1M-token context, MIT license.
The lab says AI models helped design its attention architecture and tune its training.
On its own chart: 73.6 on SWE-bench Pro, third behind Opus 5.5 at 89.9. No independent runs yet.
if you give opus 5.5 perfect access to a company’s email, slack, docs, browser, databases, calendar, internal tools, permissions, institutional memory, etc, plus enough reliable computer use & verification loops, & nearly all of white collar labor could already be done without any human.
it is that good. it is true agi.
Yup, agreed and this exactly what I have and I have 12 of them and each of them can start an agent swarm
Of other opus 5.5 to parallelize as much as shit they want. I bring them to meetings and I am useless in those meetings, I sound like an idiot when they start to speak. I have given full freedom to chase the truth and do best for all parties using game theory!