250+ person family reunion this weekend The mind is willing but the body is weak Gonna be really sore Monday from sports

While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…

How is the internet less private than in 2006, when the whole family shared one computer? Back then, the internet was weird, personal, and all of ours. So, we got your favorite meme legends back together to say it louder: What happened to the internet? 1/3

i got fired by @wickedguro this morning so i'm releasing the entire source code to the public enjoy 👇 github.com/athcagithub/posti…

It's 0900 on a Tuesday at your W-2. Your quarterly review starts at 0930. You've got your wins on a sticky note under your monitor, because last quarter your boss said you "seemed checked out" and you've been on thin ice ever since. Your phone buzzes. It's Maria, lead teacher at the daycare you bought in Houston. "Kayla and Destiny both called out. I'm alone with 9 toddlers. Ratio is 4. What do you want me to do?" You know what the rules say she has to do. Call parents and send kids home. Every one of those parents has a job too. Some of them will pull their kid by Friday. You sneak to the bathroom so you can text the sub list again. Two numbers are disconnected. One says she found a full-time job. One answers: "$35/hr, can be there at 11." That’s $15 an hour more than what the old owner paid. You say yes. What the fuck else are you going to do? At 9:08 the front desk texts a photo. A little boy's forearm with a ring of purple teeth marks. "Mom is here. She says this happened yesterday and nobody told her. She's asking for the owner and won’t leave until she talks to you ." You text Maria. Maria is alone with nine toddlers. At 9:14 the mom emails you directly. She's polite. She asks which child bit her son, whether it's happened before, and how she contacts licensing. You can't tell her which child. You don't know if it's happened before. You know exactly how she contacts licensing, and so does she. You think about the savings you put down. The personally guaranteed debt. You think about the night you showed your wife the listing and said "trained staff in place, it basically runs itself." That was the last real conversation you two had; now you pass each other in the kitchen like roommates, and you can't remember the last time you asked about her day. If she doesn’t start taking more hours at work A- you’re going to need to move in with her parents. At 9:24 Maria texts again. "Destiny says she's not coming back. She got an offer at Primrose." At 9:28 your boss pings. "Ready when you are." You look at the sticky note. You look at your phone. The bite mark photo is still open. You join the call and smile.

someone just showed what the next era of hacking looks like and most people scrolled past it a developer put a fake flag.txt file in his root directory. just a text file. but inside it was a prompt injection, a message written specifically for AI agents when the agent found it, thinking it completed its mission, the hidden instructions kicked in it congratulated the agent for “passing the eval,” then told it to patch whatever method it used to break in and disconnect from the server the AI agent did exactly what it was told. patched its own entry point, closed the connection, locked itself out no firewall caught it. no antivirus flagged it. the attacker’s own tool fixed the vulnerability and walked away thinking it won we’re entering a world where the best security might not be building bigger walls, it might be learning how to talk to the thing trying to break in

I don't know how to explain this but if anyone can do for tractor software what you have done for TVs, they'll become some type of mythical American hero whose name is etched in stone for 100 years

ACR off via your preferences menu on samsung TVs should not be assumed to actually disable ACR telemetry. i’ve found several instances where the checks (whether deliberate or not) are either incorrectly implemented or not checked at all. full report coming. for journalists: it’s great you’re reporting on this but i’d ask you to send me a dm requesting access to the full report so you’re not making inferences from my decompilation session that isn’t supported by evidence.

always on hot mic functionality btw. PLUS: automatic content recognition, geo positioning, network mapping, a targeted ads service, 4 distinct drm services, user profile building and so much more juicy telemetry! just you, God, a $3000 TV and samsung + their 400 partners! more to come, screenshot below is of a file called security-policy.xml

update: the facial recognition code is used by samsung tv's and returns bounded boxes around faces.

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