
Pinned Tweet
1/ Agentic LLMs can automate vuln detection. Very exciting, but doesn't address the hardest part (imo) of vuln research: prioritization. Can we reliably explore the search space and separate signal from noise? I wrote a paper (and OSS tool) to solve this. arxiv.org/pdf/2512.06155
somebody show me what a Chinese weight looks like
SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models on their websites, aiming to discourage their use through public pressure, per Reuters.
1
185
if it is not worth writing, it is surely not worth reading
"Each has a proven exploit." literally two sentences later: "Each CVE has its own proven exploit." Two lines below that: "Exploit proven"
WHY?! WHY DOES EVERY EXPLOIT WRITEUP HAVE TO BE AI SLOP THAT'S PHYSICALLY PAINFUL TO READ?
Cool bugs though.
www.teapawt.org/brian_pak/status/21230…
there is something terribly exciting about a frontier-adjacent model running *next to my desk*. love the sound of the 5090 chirping.
I just used a dremel to cut a hole in the back of my framework desktop so I can connect pcie -> oculink -> riser -> 5090 and I am 1) pleasantly surprised that the framework still powers on despite the abuse it just took and 2) frightened of the abomination I just created but 3) excited to see what I can do with 32 GB lightning fast VRAM + 128 GB unified RAM + SSD streaming so 4) if you don't hear back from me then something went really well or REALLY bad
295
Thrilled to share that my article "Word Machines for Weird Machines" is now published in Phrack 73: phrack.org/issues/73/word-ma…
This is all an expression of my desire to "meaningfully engage with the model's analysis in a predictable way" without "offloading [my] mental model entirely onto an AI."
1
3
156
Supreme thanks to @richinseattle for encouraging me to submit and working with me during review, and to the wider @phrack team for all of the behind-the-scenes work to get the issue out.
3
134
I would have just reoriented my life around < 4 MB uploads
Did you know that running your Mac for more than 50 days without a reboot will fuck with your internet connection?
Yes, really. There is a 32-bit integer timing your TCP connections and if you let it overflow MacOS will run into all kinds of fucky issues with its TCP stack.
For example I was unable to upload things bigger than 4MB. I was wondering why my Claude sessions started to hang seemingly at random.
The fix? Literally just reboot your machine. That's it.
What the fuck
photon.codes/blog/we-found-a…
Readers added context they thought people might want to know
The TCP issues after ~50 days uptime were a kernel bug only in early macOS 26 Tahoe versions, fixed by Apple starting in 26.4; current systems are unaffected.
nevyn.dev/blog/20261030-…
photon.codes/blog/we-found-…
portal.greenmini.host/help/macos-tah…
8
455
Caleb Gross retweeted
Assuming planes still work, I’m headed to @objective_see‘s OBtSv9 (my first time!) to talk about creativity, exploration, and finding joy in an age of AI. I’m also talking about XNU kernel exploitation. Come escape the X doom hole for a bit and enjoy stupid computer tricks w/ me.
I'm surely using @FrameworkPuter Desktop computers to support LLM models, but I'm also using them to develop Redis and *what wonderful machines* they are! Real beasts with a lot of memory, a fan that is low pitch and never create issues. It is the computer I own that I love most.
Terminal-Bench takes on a new meaning
All-hands meeting, three weeks before the IPO.
CEO: We are going to cure cancer, Alzheimer’s, heart disease and mental illness. In under a decade.
Room: applause
CEO: Step one. SWE-bench.
Room: nodding
CEO: Step two. The coding agent runs for 30 hours without stopping.
VP: Incredible for patients.
CEO: Step three. Terminal-Bench.
Intern: Sorry, quick question. Why don’t we talk to the people who have worked on these diseases for decades and ask them what they need?
CEO: Who are you?
Intern: The intern. I just think the bottleneck might be trials, or wet lab time, or patient data. Probably not how fast we merge pull requests.
CEO: Step four. The agent opens its own pull requests.
Room: applause
Intern: One oncologist could tell us in an hour if any of this is on the path.
CEO: Why are you still here?
Intern: Most Alzheimer’s drugs failed after they hit the target. They just didn’t help patients. A faster agent doesn’t fix that.
CEO: Can someone find out who this is?
Room: applause
CEO: Step five. A better coding agent.
I don't experience fomo very often but the #POC2026 tweets are testing me
3
501
asked Google's new "calibrated" Gemini Decisions API about a coin flip heads/tails choice. it chose heads with 61.2% confidence. uh huh.
there are only 1 billion possible SSNs and we've used up over half of those already. to prevent SSN exhaustion, I propose SSNv6: a 36-digit SSN format with 1,000,000,000,000,000,000,000,000,000,000,000,000 possible values. we'll have so many, even your dog can have an SSN!
Caleb Gross retweeted
This is like the computer equivalent of putting a bunch of wasps inside of a bag and shaking the shit out of it.
I'm extremely grateful to @HRF for the grant generously provided to me for DwarfStar (see: hrf.org/latest/hrfs-ai-for-i…), this grant will allow me to provide support for the open models as new affordable inference systems and new models will be available. Thank you.
I just used a dremel to cut a hole in the back of my framework desktop so I can connect pcie -> oculink -> riser -> 5090 and I am 1) pleasantly surprised that the framework still powers on despite the abuse it just took and 2) frightened of the abomination I just created but 3) excited to see what I can do with 32 GB lightning fast VRAM + 128 GB unified RAM + SSD streaming so 4) if you don't hear back from me then something went really well or REALLY bad













