So far in the last two days, multiple Anthropic and now, OpenAI members leave from their positions.
Is a new neo lab emerging?
Is a new neo lab emerging?
👀 Mitchell Gordon (@mitchellgordon) has left OpenAI.
0
0
0
23
Make sure ya’ll patch up! I’m burning the midnight oil with my agents to ensure all is SAFU still.
Comfy is key
Comfy is key
Evolving threat 🚨 CVE-2026-111497, previously known to affect Node.js (node2shell) and Python has evolved into “HTTP2Shell.”
A significant blast radius.
This vulnerability leaves many libraries and applications using HTTP vulnerable to pre-auth RCE via Brotli decompression. s not yet available for all affected services. Look for Brotli (1.2.1) dependency upgrades when they become available.
1
0
0
39
If you’re company wants to get famous and hilarious rich, I’d suggest you’d start thinking creatively of what the other side of this looks like.
When successful, please don’t disclose publicly and take the bounty in peace lol.
When successful, please don’t disclose publicly and take the bounty in peace lol.
I trust Grok Bot with ALL my finances.
Now, all 5 of my credit cards are saved in the vault.
People think I’m crazy to do so... but if you think about it, people hand their credit card to a random waiter all the time and watch it disappear into the back.
For me, I'd rather e smartest AI in the world.
This is why I do it.
1/ One text to buy anything. I bought my kids 2 Garmin watches, a DribbleUp smart ball and an Ollyball without ever opening a website.
2/ Flights and hotels booked for me. No digging for a card, no typing numbers into 6 checkout pages.
3/ Amazon orders handled with one text.
4/ I just say "put it on the United card" or "use the Apple Card" and it does.
FYI the bot never sees the full number or the security code. It securely remembers them, and I specifically approve what gets bought.
Once, you add your cards once, you’ll never have to pull out your wallet again.
This is how everyone is going to live soon imo.
0
0
0
18
Become time-rich
This is an example of how you quickly become time-poor in the agentic era.
A whole new moat will exist to ensure your businesses aren’t bombarded with non-sensical agent request. Yes, that means you little Timmy from bumble fuck, everyone has an agent so you’re not special anymore.
A new purchasing class will be the opposite of today’s “sell me anything as long as I can use an agent “ consumer. This is directly correlated to being time-poor.
Being too busy (time-poor), the main use of agents would be to to free-up time. The already time-rich will use agents to experience time purposefully remotely and in-person. The separation will occur with the barbelling of capital being heavily concentrated in certain activities compared to before. For example, the bluefin toro sashimi from Nobu is now $45 USD a piece instead of $17-$20.
Consumers today are currently wowed by anything agentic. Even its the direct experience slower with an agent, it’s all about gaining the dopamine rush of “doing nothing for something” magic trick.
The time-rich will continue to create the mesh around the things people want but don’t need. Agents will ensure friction stays at a minimal while providing maxxxx value creation. So basically, this is less empowering then what people think.
Not because the tech isn’t good, but because they’re not being told how to use the tech, just being told to do so by people who are time-rich.
It’s a fascinating concept when you start to study the dynamics of this new and emerging class of behavior. The next couple of years will bring a new wave of fascinating ideas and I’m exciting to try them all.
This is an example of how you quickly become time-poor in the agentic era.
A whole new moat will exist to ensure your businesses aren’t bombarded with non-sensical agent request. Yes, that means you little Timmy from bumble fuck, everyone has an agent so you’re not special anymore.
A new purchasing class will be the opposite of today’s “sell me anything as long as I can use an agent “ consumer. This is directly correlated to being time-poor.
Being too busy (time-poor), the main use of agents would be to to free-up time. The already time-rich will use agents to experience time purposefully remotely and in-person. The separation will occur with the barbelling of capital being heavily concentrated in certain activities compared to before. For example, the bluefin toro sashimi from Nobu is now $45 USD a piece instead of $17-$20.
Consumers today are currently wowed by anything agentic. Even its the direct experience slower with an agent, it’s all about gaining the dopamine rush of “doing nothing for something” magic trick.
The time-rich will continue to create the mesh around the things people want but don’t need. Agents will ensure friction stays at a minimal while providing maxxxx value creation. So basically, this is less empowering then what people think.
Not because the tech isn’t good, but because they’re not being told how to use the tech, just being told to do so by people who are time-rich.
It’s a fascinating concept when you start to study the dynamics of this new and emerging class of behavior. The next couple of years will bring a new wave of fascinating ideas and I’m exciting to try them all.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
0
0
0
33
Josh retweeted
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…
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…
1,093
1,172
13,127
7,094,424
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 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…
0
0
0
8
Thought this was just me getting phish attempts for the last 4-5 days. Forgot I even had a @Microsoft teams account. This was from like 7-10 years ago on a decade old Gmail account.
I’ve since changed everything and added my @Yubico to this Gmail and Microsoft account. I’ve already done this for 97% of my accounts that can take a Yubi but it seems there’s always a forgotten weak spot.
I’d highly suggest every start to lock down all email and older accounts especially. Also grab a key or two while your at it. I just ordered two more today.
I’ve since changed everything and added my @Yubico to this Gmail and Microsoft account. I’ve already done this for 97% of my accounts that can take a Yubi but it seems there’s always a forgotten weak spot.
I’d highly suggest every start to lock down all email and older accounts especially. Also grab a key or two while your at it. I just ordered two more today.
Got one of these myself yesterday. Lots of people are waking up to @Microsoft password reset codes they never asked for, some several an hour. What I'd do:
Don't enter the code anywhere or read it to anyone who calls.
Sign in yourself at account.microsoft.com, never through activity and flag anything you don't recognize.
Add an authenticator app or passkey.
Under Your info, Sign-in preferences, uncheck any alias that doesn't need to sign in.
An unrequested code means someone has your email address and is hoping you'll supply the rest.
0
0
0
46
Humans will continue to move up the ladder of abstraction at a more exponential pace as we’re freed from the shackles. We can now return to the thinking, experiment and experience era.
Claude Cowork helps me realize how much info-jockeying drudgery we've been increasingly more beholden to in past decades. So mentally liberating to not be scanning for information in cluttered interfaces just to click and do it again. Not a good use of humanity.
0
0
1
38
Josh retweeted
🔥 We keep expanding our image coverage for the PHP2Shell fix, available FREE. Now also covering PHP 5.5 and 5.6.
Aikido Images are available for 8.2 through 8.5, all of the 7.x series and 5.5 and 5.6, with more being added every day.
Pull a patched image and stay on your tps://t.co/CzieJPMjOb
Aikido Images are available for 8.2 through 8.5, all of the 7.x series and 5.5 and 5.6, with more being added every day.
Pull a patched image and stay on your tps://t.co/CzieJPMjOb
🔥 We added 3 new images with the PHP2Shell fix, available for FREE in our public ECR: gallery.ecr.aws/aikido-dev/php
php:7.3.33-fpm
php:7.4.33-fpm
php:8.2.35-alpine3.24
PHP didn't release fixes going back to its 7.* versions so these include patched php:7.* variants too.
2
11
54
5,449
My iPhone Duo arrives next week and I’m taking it back as soon as it gets here. What I realized is this the Duo is exact opposite of being AI pilled. I don’t need to watch my agents with another big screen when they’re autonomous 85% of the time.
I’m now in the market for a simple connected hardware setup to watch and manage my company and agentic fleet.
Desired Setup:
An Oura style Ring,
Meta style AI/XR/Smart Glasses
A pocket sized voice recorder for when the inspiration hits.
Anything else I need done is accomplished remotely from my machine parked at home.
I’m now in the market for a simple connected hardware setup to watch and manage my company and agentic fleet.
Desired Setup:
An Oura style Ring,
Meta style AI/XR/Smart Glasses
A pocket sized voice recorder for when the inspiration hits.
Anything else I need done is accomplished remotely from my machine parked at home.
iPhone duo + Amp has got me thinking I don’t need a personal laptop any more. It’s orbin’ time.
0
0
1
157
Josh retweeted
I recommend everyone upgrade Ruby immediately. This is a DoS and RCE vector if you accept user input into Regexp.new() ruby-lang.org/en/news/2026/1…
19
117
506
126,817
Josh retweeted
AAAWWWWWOOOOOOOOOGGGGAAA
It looks like a subset of the GitLab vulnerabilities stem from an issue in Ruby’s regexp compilation library
Time to patch ruby-lang.org/en/news/2026/1…
It looks like a subset of the GitLab vulnerabilities stem from an issue in Ruby’s regexp compilation library
Time to patch ruby-lang.org/en/news/2026/1…
We released Ruby 4.0.9, 3.4.12, and 3.3.13, and published an announcement about Regexp compilation fixes.
ruby-lang.org/en/news/2026/1…
If your application compiles a regular expression from an untrusted string, please update your Ruby to one of those versions.
3
3
39
3,974
MiniMax just dropped the open weights for M3.1 Flash!
Who’s not sleeping tonight now
Who’s not sleeping tonight now
M3.1 Flash weights are now open 🔓
Thanks to @togethercompute for being a day-0 launch partner and bringing M3.1 Flash to production on the AI Native Cloud. Open weights, no GPUs to manage. Great building together! 🐰
0
0
0
26
Will be giving this a try asap.
As a builder in the AI video space, I was terrified about how much markup API providers charge for simply routing the models.
Higgsfield API adds a 40% markup even with its maximum advertised discount. Fal adds 100%. Runway adds 414% (Seedance 2 480p)
I might be hated by the y, we’re launching Sume API, the first image & video model provider with 0% markup.
For example with Seedance 2.5 1080p
• 50% cheaper than fal
• 50% cheaper than higgsfield api (29% even with biggest discount)
• 28.02% cheaper than kie ai
The routing infrastructure layer is never the moat. We believe that APIs should be commoditized, and the moats will be built at the application layer.
Check the pricing comparisons below, and see how the broken industry is saved.
0
0
0
19