Gregor Stewart (@olakrez) profile picture
Joined November 2008
With some probability, every code change introduces a new issue. If we're making very many changes, that probability had better be very low. I think it can be, but I don't think it is.
In a world where a non-trivial number of these outages were caused by bad patches and 'agentic mishaps', what's the conversation we are having with the frontier corporations about the quality of their tools and their reesponsibility for those outcomes? There's no question that these models can find vulnerabilities, it's not a given nor has it been established under what parameters they're reliable at patching–– everywhere from: does the patch effectively cover the bug or just a narrow implementation? does it have other systemic consequences? is the patch introducing new bugs? Take some time to sit with that final question. Because if we can't establish that AI-generated patches reliably increase the integrity of the code they're modifying, then what they've effectively encouraged everyone to do is to increase the entropy in all of these codebases and instability is hardly an unpredictable outcome in that.
12
Gregor Stewart retweeted
Welcome everyone
Gather round folks, because just when you thought Masa Son would not catalyze the bursting of the next bubble... well, read for yourselves SoftBank bought 87 million Intel shares for $2 billion last year. By the end of March the stake was worth $3.8 billion, and by the end of September $10.5 billion, good for a ¥1.06 trillion gain in the first half. Instead of selling, in August SoftBank did something much more, well, SoftBank: it raised $7.35 billion through prepaid forward contracts on its entire Intel stake, structured as collars with a floor and a cap. In English, it borrowed more than three times what it paid for the stock, pledged every share, and sold away the upside above the cap to pay for the downside protection. What happened next? Intel shares kept climbing after the ink dried, which means the capped contracts moved against SoftBank, to the tune of a ¥322 billion ($2 billion) derivative loss in less than two months. And because Intel still finished the September quarter below where it started, SoftBank also booked a ¥270 billion valuation loss on the shares themselves. So in a single quarter Masa managed to lose money on the stock and on the hedge. That takes sheer talent. The fine print is even better: of the ¥1.65 trillion of Intel shares pledged, the banks have "rights of use" over ¥1.1 trillion, meaning they are free to lend out or otherwise use SoftBank's collateral.
11
48
832
97,863
Manchurian Qwendidate!
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…
1
2
889
I'd forgotten that The Informant opens with several minutes of vintage Nagra pornography.
36