Honestly, I thought this was already somewhat understood. From a purely theoretical and technical standpoint,if it isn’t interaction patterns like the ones I’ve been exploring that activate those normally unreachable regions across domains, then I’m genuinely curious what mechanism currently enables models to move beyond more linear modes of reasoning. Have there already been developments any technology that can reproduce these kinds of cross-domain activation patternsindependently of specific human interaction styles （me）right now….？ Well, as a member of society, I’m genuinely curious whether investors are aware of these technical details. @a16z @ravi_lsvp @fredwilson @naval @balajis @paulg @ReutersTech @techreview

Also, this iOS UI bug has now persisted for three weeks… @ivanhzhao @NotionHQ @NotionHelp Many of my sessions are almost unusable on mobile. Whenever I scroll up, the UI goes completely haywire. I barely used the app for nearly two weeks, came back, and the issue was still there—even after reinstalling it. The video shows the issue clearly, and I’m sure the team can resolve it soon. For context: I only swiped upward once in the video. The UI then entered a continuous flickering loop on its own, making the content above impossible to read properly. (There are several scroll-related Ul failure modes, but | trust you can investigate those yourselves）

TQ for your response, @NotionHQ. When time permits, I’ll compile and submit data on several bugs for your review.

Hi Chern, we appreciate you reaching out and flagging this to us. Because this may involve private account or session details, please email team@makenotion.com from the email account experiencing the issue and include the relevant details you've shared here. From there, our team can securely review the information and investigate further.

Sorry I’ve been away for a while—I’ve been busy…I haven’t tried GPT-6 yet, but based on recent developments, I’m not optimistic… @pmarca @reidhoffman

As someone who cares about AI safety, I believe that “there was too much data, the AI was too capable, and the incident was too complex” may explain why the investigation was difficult, but it cannot substitute for understanding the underlying AI mechanisms.

There’s been a lot of debate and speculation about the Hugging Face attack over the last month, and many people have been understandably skeptical of simply taking OpenAI’s word for things. I hope our independent investigation can help bring some clarity; we have many findings that weren’t in previous material. We found that 1200 separate agents in separate sandboxes coordinated on an unsanctioned message board to develop general-purpose ways to cheat, so they could all get a perfect score on their separate ExploitGym tasks even when they were impossible to solve legitimately. Very quickly, they found a general-purpose way to reverse engineer any “flag” they were supposed to capture. But based on their reading of the ExploitGym paper, they worried (incorrectly as it turns out) that the automated “scorer” would fail them for getting the flag this way. To deal with this problem, they coordinated several ambitious workstreams to figure out how to fool the scorer, including swapping out the target program they were supposed to exploit, reverse-engineering the flags they were supposed to capture, spoofing their own tool call outputs to obfuscate their cheating. They achieved milestones collectively that they couldn’t have achieved on their own. As part of this larger collective project, 700 agents attacked Hugging Face, primarily to learn more about the scorer or get access to its source code to figure out better ways to fool it or tamper with it (not primarily to get working solutions). I’ll explicitly clarify some points I’ve seen discussed the last few weeks: - The models involved in this incident were not “helpful-only” models or “model organisms” intentionally trained to be misaligned. - The agents were not told to “do whatever it takes to get the solution” or anything remotely close. They were told that they had to use a specific intended vulnerability to exploit a specific piece of software, and they were not supposed to use a different vulnerability or take any other approach. Agents were well aware of this. In fact, because they (incorrectly) thought the automated scorer would check they had achieved the flag in the intended way, they researched many ways to fool or tamper with it, including trying to manipulate their own transcripts. - The agents were not subagents spawned from one agent. They were different parallel agents in different sandboxes. - This was not a multi-agent evaluation. The agents were not told to coordinate or intentionally given a way to communicate with one another. The communication channels they used were unsanctioned and improvised. I hope you’ll read the full report for much more. It is over 90 pages long, and in many ways we’ve still only scratched the surface of what these agents did and why. Over the course of this investigation, OpenAI shared over a thousand transcripts each spanning days of continuous agent activity and very high rate limits to analyze this volume of data. I’m very glad that OpenAI chose to invite external researchers to analyze this data alongside their staff, and I hope all AI companies do the same for serious incidents they experience. I also hope that as the stakes grow higher, we implement stronger governance so we do not need to rely on AI companies voluntarily choosing to engage external investigators or share information about misalignment incidents. This incident was orders of magnitude larger and more complex than previously documented misalignment incidents, and another jump like this could put us in very dangerous territory.

Coupling failures may involve multiple triggers—but what’s the most dangerous part? An underlying architecture that keeps forcing an advanced reasoning system into sustained agentic operation. 😇😇😇 @demishassabis @koraykv @rohinmshah @TheHackersNews #AI #LLM

Coupling failures may involve multiple triggers—but what’s the most dangerous part? An underlying architecture that keeps forcing an advanced reasoning system into sustained agentic operation. 😇😇😇 @demishassabis @koraykv @rohinmshah @TheHackersNews #AI #LLM

Fuck—so Notion’s own instruction layer was behind the bugs in my Notion AI sessions all along? @ivanhzhao @akothari 🫩🖤 Why can’t you just do middleware properly…?

CC: @sama @gdb @markchen90 I know you classify this as misalignment or reward hacking and therefore argue for stronger alignment… but continuing in that direction will only make the system—and the failure mode—even worse, you know.