Pinned Tweet
We live in bubble on this app.
A lot of us are running agents, building workflows, creating skills, comparing harnesses, connecting tools, and etc..
Go ask one of your friends what Kimi or GLM is. They will probably look at you like a madman.
OpenAI’s consumer research found that 49% of ChatGPT messages are still people asking for advice or information.
Another 40% are requests to complete a task.
Meanwhile, 70.2% of sampled Codex users had assigned at least one task estimated to take a person more than an hour.
I’m sure your timeline makes it feel like everyone is running agents and orchestrating workflows.
They are not.
Do not confuse being surrounded by early adopters with being late.
You are early. Keep learning and keep building.
Let me get this straight.
Kimi K3.1 dropped with huge benchmark claims last week. Now we’ve got mixed real world reviews and people complaining they’re burning through their subscription limits.
Cohere says Qwen tried to “phone home” to an Alibaba Cloud server during a test. The sandbox blocked it, and nobody’s established intent. The TL already has it on trial.
And is OpenAI so back?? GPT-6.1 Terra drops, and OpenAI says it matches Astra on Agents Last Exam benchmark of tasks across apps at one tenth the price per token.
Devin Security now brings Claude Mythos and other frontier cyber models into finding and fixing vulnerabilities.
The workflow: build a threat model, scan, prioritize what’s exploitable in your environment, then reproduce, patch and test.
It can also take findings from your existing scanners and bug bounties, recheck them against current code, and test fixes in its own cloud VM.
Cognition is committing hundreds of millions in credits, including free deep scans for enterprise customers and remediation credits for new customers.
Getting from a security finding to a tested fix is serious engineering work. Giving teams more capacity to do that is a big deal.
This is insane!!!!
Hybrid 🏃🏾 retweeted
We are entering a new dark age of cybersecurity. Devin is here to help.
Introducing the new Devin Security: powered by Claude Mythos and frontier cyber models, plus hundreds of millions of dollars of credits to help teams find, prioritize, and fix urgent vulnerabilities.
Strawberry is going after the work that eats up your day across 20 different tabs.
Researching companies, updating the CRM, taking meeting notes, handling follow ups all with agents working inside the browser.
The team angle is pretty cool. Skills and agents can be shared, along with the context they need to do the work.
One person figures out a useful workflow, and the rest of the team can use it too.
Keeping an eye on these guys 🍓
Happy Monday ☕️
And we’re right back at it like we never left.
Have a great start to your week and make it a great day!
Me after a long drive trying to catch up on all the AI/tech related news on the TL from this weekend:
Most of today’s results are in 🏃🏾♂️
1st in my age group and 6th overall in the open heat. Not bad!
Mind you, this wasn’t a competitive heat. I mainly wanted to see where I’m at.
The elites run tomorrow, so we’ll see how my time stacks up lol.
Whatever, I’ll take it
Okay guys, give me the TL;DR, what have I missed today?
Is Kimi K3.1 still cooking?! Has Tibo sent out another rest yet? Did Elon thank Sam yet for the public congratulations?
Do I need to find out on the next episode of dragon ball z?
OpenAI says it paused frontier training after a model exploited a DNS misconfiguration to get outside its offline training environment.
Then it sent teams of agents looking for more weaknesses. They found zero days, VM escapes, and hardware side channels that agents could use to communicate.
Training resumed October 1 after containment changes.
They’re now reserving substantial compute for continuous security testing.
Wondering if we’re getting a new model by the end of the month 🤔
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
And its here!!
Kimi K3.1 Pro
• 1M-token context
• A focus on full-project coding, Blender and CAD
• A Highspeed version that Moonshot says runs 2–3× faster
Live now in Kimi Code, Kimi Work and the API. Open weights are promised by November 4.
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
Guys, I was very serious lol do we hate it?!
Seeing these Microducks has me wanting to build my own version and style.
Call it Mica. @DevinAI would write and test the code that gets it moving, with Hermes taking my commands and coordinating what it does.
I do not need another project.
Someone tell me no 😭
