[{"author": "Paul Wouters", "text": "<p>Does the more well apply to my agent ? <span aria-label=\"stuck out tongue wink\" class=\"emoji emoji-1f61c\" role=\"img\" title=\"stuck out tongue wink\">:stuck_out_tongue_wink:</span></p>", "time": "2026-11-18T17:02:16.000Z"}, {"author": "Paul Wouters", "text": "<p>S/more/note.</p>", "time": "2026-11-18T17:02:44.000Z"}, {"author": "Brian Carpenter", "text": "<p>@Paul, surely you are responsible for your agent's actions?</p>", "time": "2026-11-18T17:09:11.000Z"}, {"author": "Brian Carpenter", "text": "<p>meetecho, can we see the speaker please?</p>", "time": "2026-11-18T17:10:03.000Z"}, {"author": "Dan York", "text": "<p><span class=\"user-mention silent\" data-user-id=\"209\">\u00c9ric Vyncke</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244839\">said</a>:</p>\n<blockquote>\n<p>I was expecting a full room on such an AI topic :-O</p>\n</blockquote>\n<p>The name of the group did not include \"AI\" or \"Agent\" <span aria-label=\"smiley\" class=\"emoji emoji-1f603\" role=\"img\" title=\"smiley\">:smiley:</span>.  You or your agent had to read the agenda.</p>", "time": "2026-11-18T17:10:35.000Z"}, {"author": "Nick Doty", "text": "<p>I expect the rasprg folks, or anyone still working on the bigbang mailing list analysis software could appreciate the extraction library work</p>", "time": "2026-11-18T17:14:50.000Z"}, {"author": "Colin Perkins", "text": "<p>Yeah - that's a hard problem</p>", "time": "2026-11-18T17:15:03.000Z"}, {"author": "Robert Carolina", "text": "<p>My AI told me I wanted to be here</p>", "time": "2026-11-18T17:15:19.000Z"}, {"author": "David Lawrence", "text": "<p>I've occasionally considered doing a pecha kucha talk ranting about email hygiene and how a lot of ietfers really ought to know better</p>", "time": "2026-11-18T17:15:46.000Z"}, {"author": "Roman Danyliw", "text": "<p>email-reply-extractor = <a href=\"https://github.com/JayDaley/email-reply-extractor\">https://github.com/JayDaley/email-reply-extractor</a></p>", "time": "2026-11-18T17:15:54.000Z"}, {"author": "Martin Thomson", "text": "<p>I've seen humans producing implausible speed, pre-AI.</p>", "time": "2026-11-18T17:15:58.000Z"}, {"author": "Jen Linkova", "text": "<p>David: now I have in writing. You are on the PK-128 agenda...</p>", "time": "2026-11-18T17:16:15.000Z"}, {"author": "David Lawrence", "text": "<p>Oh noes</p>", "time": "2026-11-18T17:16:25.000Z"}, {"author": "Jen Linkova", "text": "<p>Robert: how did we get from 'I tell my agent what to do' to 'my agent tells me what to do'?</p>", "time": "2026-11-18T17:17:16.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"26\">Martin Thomson</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244940\">said</a>: IETF people can be very smart and fast, and always have been, even before AI.</p>", "time": "2026-11-18T17:17:40.000Z"}, {"author": "Adrian Farrel", "text": "<p>smart, fast, and cheap?</p>", "time": "2026-11-18T17:18:12.000Z"}, {"author": "Pete Resnick", "text": "<p><span class=\"user-mention\" data-user-id=\"8361\">@Mark Atwood</span> Though not always both. ;-)</p>", "time": "2026-11-18T17:18:19.000Z"}, {"author": "Mark Atwood", "text": "<p>ive never been cheap</p>", "time": "2026-11-18T17:18:36.000Z"}, {"author": "Jen Linkova", "text": "<p>Adrian: no. I object to cheap</p>", "time": "2026-11-18T17:19:17.000Z"}, {"author": "Suresh Krishnan", "text": "<p>Even the bots don\u2019t care about dkim</p>", "time": "2026-11-18T17:20:37.000Z"}, {"author": "Martin Thomson", "text": "<p>I'm feeling validated by this list.  I participate in only one of those lists.</p>", "time": "2026-11-18T17:20:38.000Z"}, {"author": "Robert Carolina", "text": "<p><span class=\"user-mention silent\" data-user-id=\"720\">Jen Linkova</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244949\">said</a>:</p>\n<blockquote>\n<p>Robert: how did we get from 'I tell my agent what to do' to 'my agent tells me what to do'?</p>\n</blockquote>\n<p>I've spent the last 10 months training my AI how to deal with a lawyer. I suspect that my AI is now trying to train me on how to deal with an AI.</p>", "time": "2026-11-18T17:21:01.000Z"}, {"author": "Martin Thomson", "text": "<p>I'd be curious as to what people think is the cause of the drop in AI mail.</p>", "time": "2026-11-18T17:21:23.000Z"}, {"author": "Colin Perkins", "text": "<p>Agents need a summer vacation too?</p>", "time": "2026-11-18T17:21:46.000Z"}, {"author": "zhang jun", "text": "<p>have you checked the statistics for mails before 2022?</p>", "time": "2026-11-18T17:21:57.000Z"}, {"author": "Steve Traugott", "text": "<p><span class=\"user-mention silent\" data-user-id=\"26\">Martin Thomson</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244975\">said</a>:</p>\n<blockquote>\n<p>I'd be curious as to what people think is the cause of the drop in AI mail.</p>\n</blockquote>\n<p>I'd equally suspect a decline in ability of pangram to detect AI-generated text as frontier models have evolved.</p>", "time": "2026-11-18T17:22:09.000Z"}, {"author": "Robert Carolina", "text": "<p><span class=\"user-mention silent\" data-user-id=\"26\">Martin Thomson</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244975\">said</a>:</p>\n<blockquote>\n<p>I'd be curious as to what people think is the cause of the drop in AI mail.</p>\n</blockquote>\n<p>Drop in detection rate?</p>", "time": "2026-11-18T17:22:16.000Z"}, {"author": "Behcet Sarikaya", "text": "<p>I have some theories on why not crowded?</p>", "time": "2026-11-18T17:22:18.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>It is percentage and not absolute numbers, i.e., when human cam back from vacations, the AI percentage dropped?</p>", "time": "2026-11-18T17:22:18.000Z"}, {"author": "Brian Carpenter", "text": "<p>\"Drop in detection rate?\"</p>", "time": "2026-11-18T17:22:55.000Z"}, {"author": "Martin Thomson", "text": "<p><span class=\"user-mention silent\" data-user-id=\"5020\">Steve Traugott</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244982\">said</a>:</p>\n<blockquote>\n<p><span class=\"user-mention silent\" data-user-id=\"26\">Martin Thomson</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/244975\">said</a>:</p>\n<blockquote>\n<p>I'd be curious as to what people think is the cause of the drop in AI mail.</p>\n</blockquote>\n<p>I'd equally suspect a decline in ability of pangram to detect AI-generated text as frontier models have evolved.</p>\n</blockquote>\n<p>But there is also a drop in overall traffic, which suggests that this is not just that.</p>", "time": "2026-11-18T17:23:05.000Z"}, {"author": "Eric Rescorla", "text": "<p>I probably won't talk about this in my slides, but i pretty routinely use AI tools to help me refine my ideas at this point. I had it look at my slides for DTLS 1.3 last night and it found an error and suggested a new clever idea</p>", "time": "2026-11-18T17:23:10.000Z"}, {"author": "Nick Doty", "text": "<p>I think a lot of IETF communication is bursty, so I'm not sure absolute numbers or percentages reflect underlying trends</p>", "time": "2026-11-18T17:23:13.000Z"}, {"author": "Brian Carpenter", "text": "<p>for sure detection will decline as LLMs get better</p>", "time": "2026-11-18T17:23:37.000Z"}, {"author": "Behcet Sarikaya", "text": "<p>I think it is because people do not expect to learn something from this meeting</p>", "time": "2026-11-18T17:23:40.000Z"}, {"author": "Nick Doty", "text": "<p>has anyone looked at the IETF github repos?</p>", "time": "2026-11-18T17:24:15.000Z"}, {"author": "S Moonesamy", "text": "<p><a href=\"https://mailarchive.ietf.org/arch/msg/scitt/ZWCYAJWfKZ2gVTInQsYDd9yLVOk\">https://mailarchive.ietf.org/arch/msg/scitt/ZWCYAJWfKZ2gVTInQsYDd9yLVOk</a></p>", "time": "2026-11-18T17:24:29.000Z"}, {"author": "Behcet Sarikaya", "text": "<p>+1 to Eric</p>", "time": "2026-11-18T17:24:33.000Z"}, {"author": "Beno\u00eet Claise", "text": "<p>Would be nice for NOMCOM to have access to this data.</p>", "time": "2026-11-18T17:24:44.000Z"}, {"author": "Robert Carolina", "text": "<p>Keep in mind the new(-ish) European regulation on AI mandates efforts to watermark AI generated outputs.</p>", "time": "2026-11-18T17:24:46.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>@Benoit why ?</p>", "time": "2026-11-18T17:25:05.000Z"}, {"author": "Martin Thomson", "text": "<p>I appreciate getting the information.  I share Cullen's concerns about what to do with the information.</p>", "time": "2026-11-18T17:25:11.000Z"}, {"author": "Jen Linkova", "text": "<p>I'd love to see the result of that dashboard for my emails!</p>", "time": "2026-11-18T17:25:31.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"6518\">Beno\u00eet Claise</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245016\">said</a>: A vendor's guess about how each person writes shouldn't feed leadership selection. Judge candidates on their work, which NomCom can already read.</p>", "time": "2026-11-18T17:25:52.000Z"}, {"author": "Cullen Jennings", "text": "<p>I have people beg me to please put my emails through AI before sending :-)</p>", "time": "2026-11-18T17:25:55.000Z"}, {"author": "Ionu\u021b Mihalcea", "text": "<p>+1 Ekr - as well as formatting things that I have an informal write-up for, but it would take me ages to write \"properly\".</p>", "time": "2026-11-18T17:25:59.000Z"}, {"author": "Martin Thomson", "text": "<p>It would be really nice to have some adversarial research on Pangram if anyone is aware of some.  Like with watermarking, it seems to be something that an adversary can use Pangram as an oracle for.</p>", "time": "2026-11-18T17:26:29.000Z"}, {"author": "Eric Rescorla", "text": "<p><span class=\"user-mention silent\" data-user-id=\"2916\">Ionu\u021b Mihalcea</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245033\">said</a>:</p>\n<blockquote>\n<p>+1 Ekr - as well as formatting things that I have an informal write-up for, but it would take me ages to write \"properly\".</p>\n</blockquote>\n<p>I also had it do the diagrams for my slide. I gave it ascii art and had it render good looking diagrams</p>", "time": "2026-11-18T17:26:41.000Z"}, {"author": "Glenn Deen", "text": "<p>My spelllling has improved greating by first running through AI filtering</p>", "time": "2026-11-18T17:27:06.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>I do not care too much on the numbers but more about detecting whether AI-generated IETF content is a real thing or a fantasy</p>", "time": "2026-11-18T17:27:22.000Z"}, {"author": "Eric Rescorla", "text": "<p>One might ask why the chairs haven't thrown people off</p>", "time": "2026-11-18T17:27:35.000Z"}, {"author": "Andrew Campling", "text": "<p>This could be good to feed into RASPRG to explore the topic of detection methodologies etc further</p>", "time": "2026-11-18T17:27:48.000Z"}, {"author": "Jon Geater", "text": "<blockquote>\n<p>One might ask why the chairs haven't thrown people off</p>\n</blockquote>\n<p>Let me forward you some emails...</p>", "time": "2026-11-18T17:28:10.000Z"}, {"author": "Jen Linkova", "text": "<p>my spelling has been declining since I started using spellcheckers....;(</p>", "time": "2026-11-18T17:28:12.000Z"}, {"author": "Phillip Hallam-Baker", "text": "<p>my spelling improved with spell check</p>", "time": "2026-11-18T17:28:29.000Z"}, {"author": "Phillip Hallam-Baker", "text": "<p>Having dyslexia makes spelling difficult</p>", "time": "2026-11-18T17:28:47.000Z"}, {"author": "Eric Rescorla", "text": "<p>I am starting to wonder if we need to offer filtered mailing list feeds</p>", "time": "2026-11-18T17:28:49.000Z"}, {"author": "Jon Geater", "text": "<p>I now delberatley throw spelling errors into my documents/emails to look less AI ;-)</p>", "time": "2026-11-18T17:29:01.000Z"}, {"author": "Hesham Moussa", "text": "<p>I wonder what is the ground truth here if we have any?</p>", "time": "2026-11-18T17:29:07.000Z"}, {"author": "Nick Doty", "text": "<p>I appreciate the disclosure that the slides are AI-generated, although it's obviously pretty clear these days</p>", "time": "2026-11-18T17:29:24.000Z"}, {"author": "Eric Rescorla", "text": "<p><span class=\"user-mention silent\" data-user-id=\"6163\">Jon Geater</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245051\">said</a>:</p>\n<blockquote>\n<blockquote>\n<p>One might ask why the chairs haven't thrown people off</p>\n</blockquote>\n<p>Let me forward you some emails...</p>\n</blockquote>\n<p>Thanks for pushing back. That's my bad</p>", "time": "2026-11-18T17:29:25.000Z"}, {"author": "Joel Halpern", "text": "<p>I appreciate Robert's data.</p>", "time": "2026-11-18T17:29:26.000Z"}, {"author": "Jen Linkova", "text": "<p>spelling of my emails improved, but my spelling declined..;(</p>", "time": "2026-11-18T17:29:35.000Z"}, {"author": "Steve Traugott", "text": "<p>/me wonders how much of the trend might also be indicative of people's writing styles shifting as they become more intertwined with and influenced by their own AI use.</p>", "time": "2026-11-18T17:29:38.000Z"}, {"author": "Pete Resnick", "text": "<p><span class=\"user-mention\" data-user-id=\"7990\">@Hesham Moussa</span> \"ground truth\"?</p>", "time": "2026-11-18T17:29:43.000Z"}, {"author": "Robert Carolina", "text": "<p><span class=\"user-mention silent\" data-user-id=\"71\">Cullen Jennings</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245032\">said</a>:</p>\n<blockquote>\n<p>I have people beg me to please put my emails through AI before sending :-)</p>\n</blockquote>\n<p>I have the opposite problem with my students. Great writing skill for my students (MSc programme in cyber security) is the first sign of AI-generated content. More seriously, finding an acceptable use policy is difficult because the AI models do not seem to distinguish between \"make grammar and word choice corrections\" and \"re-write this to strengthen the argument\"</p>", "time": "2026-11-18T17:29:43.000Z"}, {"author": "Mark Atwood", "text": "<p>The answer to AI volume is AI review. Models are good at finding real defects in drafts: missing case arms, codepoint collisions, broken security claims. They're bad at telling who wrote a message, and that gets worse with every model release. Pangram's error rates in the deck are the vendor's own figures. Spend the $450 a month on review, not on scoring senders.</p>", "time": "2026-11-18T17:29:50.000Z"}, {"author": "Beno\u00eet Claise", "text": "<p>@Eric I have no problem with people using AI to improve emails (translation) versus generated ideas from LLMs. Leadership members should have their own point of views.</p>", "time": "2026-11-18T17:29:59.000Z"}, {"author": "Selena Deckelmann", "text": "<p>Another use case for consideration regarding Pangram: <a href=\"https://wikiedu.org/blog/2026/01/29/generative-ai-and-wikipedia-editing-what-we-learned-in-2025/\">https://wikiedu.org/blog/2026/01/29/generative-ai-and-wikipedia-editing-what-we-learned-in-2025/</a></p>", "time": "2026-11-18T17:30:27.000Z"}, {"author": "Glenn Deen", "text": "<p>I think perhaps our worry is less about AI use and perhaps more about Autonomous AI use</p>", "time": "2026-11-18T17:30:31.000Z"}, {"author": "Steve Traugott", "text": "<p>Mark +1</p>", "time": "2026-11-18T17:30:40.000Z"}, {"author": "Cullen Jennings", "text": "<p>I am pretty skeptical that filtering on data generated the way Jay was describing would result in a metric that you could filter on.</p>", "time": "2026-11-18T17:30:46.000Z"}, {"author": "Eric Rescorla", "text": "<p><span class=\"user-mention silent\" data-user-id=\"8001\">Glenn Deen</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245081\">said</a>:</p>\n<blockquote>\n<p>I think perhaps our worry is less about AI use and perhaps more about Autonomous AI use</p>\n</blockquote>\n<p>My concern is that but also just the very spammy and unhelpful and clogs the list</p>", "time": "2026-11-18T17:31:12.000Z"}, {"author": "Hesham Moussa", "text": "<p>@pete Resnick at the moment this sounds like we are just assuming that the pangram is 100% correct, but what if it is hallucinating? How do we know if it actually working?</p>", "time": "2026-11-18T17:31:41.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>@Hesham even if accurate at 50%, it shows a trend since early 2026</p>", "time": "2026-11-18T17:32:43.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"7990\">Hesham Moussa</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245094\">said</a>: The ground truth is the work.</p>", "time": "2026-11-18T17:32:53.000Z"}, {"author": "Cullen Jennings", "text": "<p>What pangram detects is \"was AI used\", it does not detect is this a valid opinion of a real human that is useful input to the WG.</p>", "time": "2026-11-18T17:32:54.000Z"}, {"author": "Jaime Jimenez", "text": "<p>FYI: You can also check drafts for AI signs on <a href=\"https://ietf.jaime.win/is-it-ai\">https://ietf.jaime.win/is-it-ai</a></p>", "time": "2026-11-18T17:32:55.000Z"}, {"author": "Hesham Moussa", "text": "<p>I am thinking of creating a fake mailing list that is a controlled environment where we control the portion of AI generated versus human generated emails. We know the split. Then we test the tool on it and see if it is able to detect the split accurately. Basically bench mark the tool</p>", "time": "2026-11-18T17:33:05.000Z"}, {"author": "Jon Geater", "text": "<blockquote>\n<p>The answer to AI volume is AI review.</p>\n</blockquote>\n<p>Yes I'm in favour of raising the bar for what can be added as an ID. Automating the exisintg rules (did they even run ID-Nits?) and add more quality control</p>", "time": "2026-11-18T17:33:46.000Z"}, {"author": "Glenn Deen", "text": "<p>@EKR - I have been thinking that having some agent on lists that recognizes spammy and unhelpful stuff might be useful.   </p>\n<p>For the lists I casually follow with daily summaries it would be nice to have an included contextual summary - yes my own agent can do it, but it might be useful to have a shared summary with context added.</p>", "time": "2026-11-18T17:33:53.000Z"}, {"author": "Jaime Jimenez", "text": "<p>Adding a checker on idnits submission, showing a warning to submitter would be helpful</p>", "time": "2026-11-18T17:34:20.000Z"}, {"author": "zhang jun", "text": "<p>@Hesham Moussa, then in such a case, the controlled data can be biased. Because there can be different type of ai-generated emails, and human-generation emails.</p>", "time": "2026-11-18T17:34:24.000Z"}, {"author": "Eric Rescorla", "text": "<p>How can the processing time or storage be a meaningful number here.</p>", "time": "2026-11-18T17:34:26.000Z"}, {"author": "Andrew Campling", "text": "<p>@Hesham Creating a control list makes sense, but the pace of evolution of the tools means that it will only be helpful as an indicator at a point in time.</p>", "time": "2026-11-18T17:34:43.000Z"}, {"author": "Colin Perkins", "text": "<p>Time for people to read the drafts I can believe is an issue; storage and processing has to be trivial</p>", "time": "2026-11-18T17:34:54.000Z"}, {"author": "Eric Rescorla", "text": "<p>I strongly suspect that the entire storage could be stored on a free AWS instance</p>", "time": "2026-11-18T17:35:24.000Z"}, {"author": "Brian Trammell", "text": "<p>as long as humans have methods to determine which drafts they should actually spend time on</p>", "time": "2026-11-18T17:35:28.000Z"}, {"author": "Hesham Moussa", "text": "<p>@zhang, the point of a controlled data is exactly that, we can test various aspects</p>", "time": "2026-11-18T17:35:36.000Z"}, {"author": "zhang jun", "text": "<p>thinking about that, if you have a system to score the ai-likeliness. then it is always possible to train a LLM to have a low score for a text inputted for such a system.</p>", "time": "2026-11-18T17:36:14.000Z"}, {"author": "Colin Perkins", "text": "<p>The entire IETF email archive is 40G. It's not  big.</p>", "time": "2026-11-18T17:36:28.000Z"}, {"author": "Eric Rescorla", "text": "<p>If processing and storage of these drafts is somehow a significant cost, that is reflective of very severe issues in the architecture of our infrastructure</p>", "time": "2026-11-18T17:36:29.000Z"}, {"author": "Jaime Jimenez", "text": "<p>storage is free and draft verification checks for STYLE deviation is very cheap. You would flag the majority of autogenerated drafts</p>", "time": "2026-11-18T17:36:45.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>The issue is indeed probably not the IT cost but the human cost, e.g., polluting the bottom of a WG list of related drafts</p>", "time": "2026-11-18T17:36:57.000Z"}, {"author": "Glenn Deen", "text": "<p>I feel part of the -00 issue is that there is has always been a bit of the barrier for non-regular-IETF folks getting through the mechanical process of writing a -00.   AI tools take awat that barrier, meaning we have an injest scaling problem</p>", "time": "2026-11-18T17:37:18.000Z"}, {"author": "Brian Carpenter", "text": "<p>43 drafts in half an hour is fairly diagnostic</p>", "time": "2026-11-18T17:37:22.000Z"}, {"author": "Andrew Campling", "text": "<p>I take it that at a key part of the underlying problem the volume increase (and possibly quality decrease) rather than the use of AI per se?</p>", "time": "2026-11-18T17:37:48.000Z"}, {"author": "Steve Traugott", "text": "<p>Glenn +1</p>", "time": "2026-11-18T17:37:52.000Z"}, {"author": "Selena Deckelmann", "text": "<p>For another very rough data point: Wikipedians have established a few policies to restrict AI use and the % of new articles and drafts flagged as AI generated dropped by 5% from late 2025 to 2026.  (From 20 to 15)</p>", "time": "2026-11-18T17:38:36.000Z"}, {"author": "Pete Resnick", "text": "<p>(Token philosopher) I hope phenomenology is not happening at IETF.</p>", "time": "2026-11-18T17:38:37.000Z"}, {"author": "Nick Doty", "text": "<p>\"like everyone else\"?</p>", "time": "2026-11-18T17:38:41.000Z"}, {"author": "Ionu\u021b Mihalcea", "text": "<p>@Glenn: I wonder if the initial surge will see a drop-off because of that, if people notice that most of these new -00s don't get any attention, so it's not worth doing. Don't know if there's any data from other fields.</p>", "time": "2026-11-18T17:39:01.000Z"}, {"author": "Muhammad Jadoon", "text": "<p>Automated meeting notes would be helpful actually</p>", "time": "2026-11-18T17:39:22.000Z"}, {"author": "Pete Resnick", "text": "<p>A good deal of WGs take seconds instead of minutes.</p>", "time": "2026-11-18T17:39:49.000Z"}, {"author": "Chris Box", "text": "<p>Pleased to see the Yes Prime Minister quote</p>", "time": "2026-11-18T17:40:09.000Z"}, {"author": "Steve Traugott", "text": "<p>I'd have a hard time taking minutes at Eric's speaking speed ;-)</p>", "time": "2026-11-18T17:40:14.000Z"}, {"author": "Suresh Krishnan", "text": "<p>@Steve, the AI minutes of ekr speaking are great</p>", "time": "2026-11-18T17:40:36.000Z"}, {"author": "Joel Halpern", "text": "<p>I was initially skeptical of this tool.  EKR has turned it into something quite useful.</p>", "time": "2026-11-18T17:40:56.000Z"}, {"author": "Brian Trammell", "text": "<p>+1</p>", "time": "2026-11-18T17:41:10.000Z"}, {"author": "Glenn Deen", "text": "<p>@Ionu\u021b Mihalcea - maybe it will drop, but things will likely get worse before.</p>", "time": "2026-11-18T17:41:22.000Z"}, {"author": "Pete Resnick", "text": "<p>EKR might not judge you. I will.</p>", "time": "2026-11-18T17:41:28.000Z"}, {"author": "Karen O'Donoghue", "text": "<p>I really like this tool. I do read and make any corrections (usually the identification of individuals), before using them, but they are really helpful.</p>", "time": "2026-11-18T17:42:34.000Z"}, {"author": "Ole Jacobsen", "text": "<p>Does this tool have the ability to detect speech at 100 Milli-EKR?</p>", "time": "2026-11-18T17:42:43.000Z"}, {"author": "Brian Trammell", "text": "<p>yep. autominutes are awesome. you still need note takers. </p>\n<p>i've had success with having antigravity merge the notes with the autominutes and flag anything that's inconsistent. do this quickly enough after a meeting as a chair and your recollection is enough to resolve said conflicts.</p>", "time": "2026-11-18T17:43:04.000Z"}, {"author": "Nick Doty", "text": "<p>I suspect it will get even harder and harder to get volunteers to take notes or minutes as people see the availability of automated tools</p>", "time": "2026-11-18T17:44:20.000Z"}, {"author": "Karen O'Donoghue", "text": "<p>In small working groups, it is already impossible to get note takers.</p>", "time": "2026-11-18T17:44:45.000Z"}, {"author": "Glenn Deen", "text": "<p>I really like the tool, the classic role of scribe can really be action-item collector - until the tool does it.</p>", "time": "2026-11-18T17:44:47.000Z"}, {"author": "Suresh Krishnan", "text": "<p>@Brian: the downside of having a notetaker is that it sidelines who is potentially an active participant in discussions.</p>", "time": "2026-11-18T17:44:51.000Z"}, {"author": "Brian Trammell", "text": "<p>at the same time if you mainly just need key points to keep the autominutes on track, you can almost do that from the chair</p>", "time": "2026-11-18T17:44:51.000Z"}, {"author": "Robert Sparks", "text": "<p>short response to the assertion that storing the drafts is free. Storing the increasing number of bits all by itself is small cost. Allowing things like search over the content of all the new bits, producting lists of the things, making the things rapidly available, and the other additional load that the extra bits drive is likely still not outrageous, but it _is_ an increase in cost. But the better question is how hard the existance of the extra pile of bits is on the regular participant.</p>", "time": "2026-11-18T17:45:00.000Z"}, {"author": "Brian Trammell", "text": "<p>(I've done that for interims FWIW, though there the edit log to the WG github during the meeting is usually sufficient to capture the meeting's output)</p>", "time": "2026-11-18T17:45:44.000Z"}, {"author": "Pete Resnick", "text": "<p>How much did he say this was costing him?</p>", "time": "2026-11-18T17:45:53.000Z"}, {"author": "Brian Trammell", "text": "<p>200 a meeting</p>", "time": "2026-11-18T17:46:08.000Z"}, {"author": "Brian Trammell", "text": "<p>eight drinks at the hotel bar</p>", "time": "2026-11-18T17:46:22.000Z"}, {"author": "Dhruv Dhody", "text": "<p>EKR minutes generally capture what the presenter presented, whereas the longstanding wisdom within WGs has been to focus on the Q&amp;A. I wanted to get a sense of whether folks find AI-generated summaries of presentations useful?</p>", "time": "2026-11-18T17:46:36.000Z"}, {"author": "Brian Trammell", "text": "<p>dhruv: depends on the speaker and slides. \"read a wall of text\" presentations less so</p>", "time": "2026-11-18T17:47:10.000Z"}, {"author": "Joel Halpern", "text": "<p>What I generally look for is the Q&amp;A and the tool seems to capture that too.</p>", "time": "2026-11-18T17:47:18.000Z"}, {"author": "Cullen Jennings", "text": "<p>If we recorded the meeting in VCON format we could have a bunch of the stuff to make this easier</p>", "time": "2026-11-18T17:47:35.000Z"}, {"author": "Pete Resnick", "text": "<p>Making this an official tool with the ability to add these features seems pretty reasonable.</p>", "time": "2026-11-18T17:47:59.000Z"}, {"author": "Mark Nottingham", "text": "<p>I will enjoy reading the automatic minutes generated from this.</p>", "time": "2026-11-18T17:48:21.000Z"}, {"author": "Andrew Campling", "text": "<p>@Dhruv I think summaries can be useful,.  It's easy enough to generate those for yourself but then of course we end up with multiple different summaries.  I think that having an \"official\" summary is a good idea.</p>", "time": "2026-11-18T17:48:31.000Z"}, {"author": "Jay Daley", "text": "<p>VCON is now available  - I will announce shortly</p>", "time": "2026-11-18T17:48:37.000Z"}, {"author": "Glenn Deen", "text": "<p>the video and audio are also published on youtube</p>", "time": "2026-11-18T17:52:52.000Z"}, {"author": "Jim Fenton", "text": "<p>Doesn't the Note Well already require speakers to give permission for use like this?</p>", "time": "2026-11-18T17:53:33.000Z"}, {"author": "Joel Halpern", "text": "<p>In fact our process requires that participants be identifiable.</p>", "time": "2026-11-18T17:53:54.000Z"}, {"author": "\u00c9ric Vyncke", "text": "<p>Just wondering whether the minutes of this session will be correct for Ekr's part ;-)</p>", "time": "2026-11-18T17:54:07.000Z"}, {"author": "Andrew Campling", "text": "<p>If we standardise on this tool, hopefully it will encourage chairs to ensure participants use the Meetecho tool to join the queue and to name themselves when they start their intervention.  :-)</p>", "time": "2026-11-18T17:54:22.000Z"}, {"author": "Joel Halpern", "text": "<p>I believe the PII rules recognize an exception for information needed for the organizations work.  Which seems to be the case for things like session minutes.  Redacting the minutes wihtout redacting the recording seems very odd.</p>", "time": "2026-11-18T17:54:57.000Z"}, {"author": "Jaime Jimenez", "text": "<p>Live Datatracker fetch is a bit brutal. For non-live tools, rsync on ietf is a better approach.</p>", "time": "2026-11-18T17:55:03.000Z"}, {"author": "Mark Atwood", "text": "<p>The Meetecho mic queue already knows who's speaking. A feed of it would give every minutes tool, EKR's included, names without guessing. I asked on tools-discuss on Monday.</p>", "time": "2026-11-18T17:55:30.000Z"}, {"author": "Colin Perkins", "text": "<p>The Datatracker API is great for somethings, horrible for others. Agree that there are some small changes that would help.</p>", "time": "2026-11-18T17:55:51.000Z"}, {"author": "Dan York", "text": "<p>Yes, +1 for Mark's \"ietf-llm\" tool... VERY powerful for catching up on what is happening in different working groups.</p>", "time": "2026-11-18T17:55:51.000Z"}, {"author": "Mark Nottingham", "text": "<p>(this is the accompanying ietf-skill -- I'll be talking about it in wit area :)</p>", "time": "2026-11-18T17:56:44.000Z"}, {"author": "Robert Sparks", "text": "<p>/api/v1 is unsustainable - we really need to get on with defining apis that provide what people actually want instead of a weak approximation of writing sql queries.</p>", "time": "2026-11-18T17:57:05.000Z"}, {"author": "Mark Atwood", "text": "<p>(just provide graphql of all the underlying data)</p>", "time": "2026-11-18T17:57:43.000Z"}, {"author": "Chris Lemmons", "text": "<blockquote>\n<p>The Meetecho mic queue already knows who's speaking.</p>\n</blockquote>\n<p>This is only true in the average case. Almost certainly worth optimizing for, but there are definitely rough edges. The MOPS meeting the other day comes to mind, with 5 speakers up front passing 2 mics around interchangeably without identifying themselves each time. (It was a panel discussion.) That's always going to be !!fun!! to process. :)</p>", "time": "2026-11-18T17:57:52.000Z"}, {"author": "Mark Nottingham", "text": "<p>Robert I wonder if it makes sense to expose the dataset publicly on S3, R2, etc.</p>", "time": "2026-11-18T17:58:24.000Z"}, {"author": "Mark Nottingham", "text": "<p>(much like Common Crawl does)</p>", "time": "2026-11-18T17:58:33.000Z"}, {"author": "Jay Daley", "text": "<p>For reference, the problem is that extracting anything that resembles a \"voiceprint\" is different PII and potentially covered by privacy laws in a different way from the recordings themselves.</p>", "time": "2026-11-18T17:59:04.000Z"}, {"author": "Behcet Sarikaya", "text": "<p><span class=\"user-mention\" data-user-id=\"30\">@Roman Danyliw</span> No HedgeDoc record on this meeting</p>", "time": "2026-11-18T17:59:24.000Z"}, {"author": "Jen Linkova", "text": "<p>+100 to Suresh...I tried being a minute taker a few times, and I ended up re-watching recording that evening and filling minutes based on the recording.</p>", "time": "2026-11-18T17:59:41.000Z"}, {"author": "Joel Halpern", "text": "<p>@Jay Daley if you were trying to raise that in your microphone comment, I think it got lost.</p>", "time": "2026-11-18T17:59:46.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"326\">Chris Lemmons</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245366\">said</a>: Agreed: queue first, voice cues as the fallback. The average case is most of a meeting.</p>", "time": "2026-11-18T17:59:48.000Z"}, {"author": "Andrew Dolan", "text": "<p>+1 to all these uses. I think it makes reviewing materials much more approachable. ekr noted in the presentation for corrections, but I want to highlight again - as workflows move more toward the use of AI to consume transcripts, materials, slides, drafts, etc, treating those as untrusted input for hidden prompt injection should be watched for...</p>", "time": "2026-11-18T17:59:48.000Z"}, {"author": "Colin Perkins", "text": "<p>We need both dumps of the data and also APIs that make it easy to get changes</p>", "time": "2026-11-18T18:00:08.000Z"}, {"author": "Glenn Deen", "text": "<p>I've literally used EKR\"s notes to feed into a AI along with an I-D rev to make sure the requested edits were all properly made</p>", "time": "2026-11-18T18:00:08.000Z"}, {"author": "Dan York", "text": "<p>The minutes are useful in that they can be ingested by Mark's ietf-llm so that you can then query against them. <span aria-label=\"smiley\" class=\"emoji emoji-1f603\" role=\"img\" title=\"smiley\">:smiley:</span></p>", "time": "2026-11-18T18:00:09.000Z"}, {"author": "Ionu\u021b Mihalcea", "text": "<p>@Jay: I was just going to suggest (/s) building a voiceprint database since we have such a rich dataset across the years (again, /s)</p>", "time": "2026-11-18T18:00:17.000Z"}, {"author": "Robert Sparks", "text": "<p>we already make almost all of  the database available. But yes, something like what you describe that populates buckets with _things of a shape that we can maintain_ is a good direction to aim for.</p>", "time": "2026-11-18T18:00:19.000Z"}, {"author": "Dhruv Dhody", "text": "<p>it was a good way for new folks to volunteer to be helpful in a WG -- i was one of those <span aria-label=\"slight smile\" class=\"emoji emoji-1f642\" role=\"img\" title=\"slight smile\">:slight_smile:</span></p>", "time": "2026-11-18T18:00:21.000Z"}, {"author": "Mark Nottingham", "text": "<p>For the record, one of ietf-llm's major data sources is EKR's transcripts</p>", "time": "2026-11-18T18:00:22.000Z"}, {"author": "Brian Carpenter", "text": "<p>Sir Humphrey was right. We need an agreed record of a meeting.</p>", "time": "2026-11-18T18:00:36.000Z"}, {"author": "Chris Lemmons", "text": "<p>I used the ietf-llm skill plus manually providing transcripts to have an agent locate all the people who had publically provided feedback on my draft so I could properly thank them in my slides. :)</p>", "time": "2026-11-18T18:00:49.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"14\">Jay Daley</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245382\">said</a>: Telling which of the 12 people in a room is talking is a closed-set problem, and it works. Picking one person out of 8 billion from a voice doesn't, whatever the movies show. Telling speakers apart in a meeting isn't a voiceprint database.</p>", "time": "2026-11-18T18:00:58.000Z"}, {"author": "Eric Rescorla", "text": "<p><span class=\"user-mention silent\" data-user-id=\"8361\">Mark Atwood</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245406\">said</a>:</p>\n<blockquote>\n<p><span class=\"user-mention silent\" data-user-id=\"14\">Jay Daley</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245382\">said</a>: Telling which of the 12 people in a room is talking is a closed-set problem, and it works. Picking one person out of 8 billion from a voice doesn't, whatever the movies show. Telling speakers apart in a meeting isn't a voiceprint database.</p>\n</blockquote>\n<p>Correct. This is just doing in-world diarization.</p>", "time": "2026-11-18T18:01:34.000Z"}, {"author": "Jaime Jimenez", "text": "<p>For those into STT, Parakeet TDT v2 is very good. It is for free\\</p>", "time": "2026-11-18T18:01:55.000Z"}, {"author": "Dan York", "text": "<p><span class=\"user-mention silent\" data-user-id=\"114\">Dhruv Dhody</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245398\">said</a>:</p>\n<blockquote>\n<p>it was a good way for new folks to volunteer to be helpful in a WG -- i was one of those :)</p>\n</blockquote>\n<p>Yes, back in the 2000s volunteering to do minutes as a newcomer was how I learned the names of all the people involved in the groups.</p>", "time": "2026-11-18T18:02:05.000Z"}, {"author": "David Schinazi", "text": "<p>Does anyone have an umbrella that the chairs could borrow?</p>", "time": "2026-11-18T18:02:37.000Z"}, {"author": "Jen Linkova", "text": "<p>Dan: you must be on of those people who can listen and write at the same time? I can't do it even in my native language...</p>", "time": "2026-11-18T18:02:50.000Z"}, {"author": "Mark Atwood", "text": "<p>I spent years explaining to police that AWS Rekognition can't do what the movies show. Voice is the same: matching against a known set works, and searching for anyone doesn't.</p>", "time": "2026-11-18T18:03:31.000Z"}, {"author": "Steve Traugott", "text": "<p>Possibly unpopular opinion -- submitter is responsible for what they submit, regardless of how it was written.  Disclosure risks punting responsibility.</p>", "time": "2026-11-18T18:04:15.000Z"}, {"author": "Dan York", "text": "<p><span class=\"user-mention silent\" data-user-id=\"720\">Jen Linkova</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245425\">said</a>:</p>\n<blockquote>\n<p>Dan: you must be on of those people who can listen and write at the same time? I can't do it even in my native language...</p>\n</blockquote>\n<p>Yes, I am. Largely by virtue of having taking a typing class in my last semester of high school <span aria-label=\"rolling on the floor laughing\" class=\"emoji emoji-1f923\" role=\"img\" title=\"rolling on the floor laughing\">:rolling_on_the_floor_laughing:</span></p>", "time": "2026-11-18T18:04:18.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"5020\">Steve Traugott</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245441\">said</a>:</p>\n<blockquote>\n<p>Possibly unpopular opinion -- submitter is responsible for what they submit, regardless of how it was written.  Disclosure risks punting responsibility.</p>\n</blockquote>\n<p>Steven thats my position.  My tools are my business, my posts are my responsibilty</p>", "time": "2026-11-18T18:04:55.000Z"}, {"author": "Behcet Sarikaya", "text": "<p>I think we should be able ack AI in the drafts we write <br>\nbut also state that the responsibility belongs to the authors</p>", "time": "2026-11-18T18:05:23.000Z"}, {"author": "Eric Rescorla", "text": "<p>To Cullen's question, I have spent about $65 on this meeting so far. $55 on transcription and &lt;$10 on various text-based tasks</p>", "time": "2026-11-18T18:05:39.000Z"}, {"author": "David Schinazi", "text": "<p>Good thing the IESG has insurance for what happens in their official capacity</p>", "time": "2026-11-18T18:05:41.000Z"}, {"author": "Mark Atwood", "text": "<p>do we have to ack using a grammer checker?</p>", "time": "2026-11-18T18:05:47.000Z"}, {"author": "Mark Atwood", "text": "<p>what is a \"fake\" internet draft?</p>", "time": "2026-11-18T18:06:12.000Z"}, {"author": "S Moonesamy", "text": "<p>@Mark Atwood No</p>", "time": "2026-11-18T18:06:38.000Z"}, {"author": "Robert Wilton", "text": "<p>Yes, it makes sense to acknowledge use of AI in drafts.  It is pretty simple to add.</p>", "time": "2026-11-18T18:06:52.000Z"}, {"author": "Behcet Sarikaya", "text": "<p>Maybe companies should send AI to IETF meetings</p>", "time": "2026-11-18T18:07:03.000Z"}, {"author": "Mark Atwood", "text": "<p><span class=\"user-mention silent\" data-user-id=\"7352\">S Moonesamy</span> <a href=\"#narrow/channel/334-genarea/topic/ietf-127/near/245458\">said</a>:</p>\n<blockquote>\n<p>@Mark Atwood No</p>\n</blockquote>\n<p>all commercial online grammer checkers now are using inference models, not regexs.</p>", "time": "2026-11-18T18:07:26.000Z"}, {"author": "Joel Halpern", "text": "<p>I got pushed to use a grammar checker.  It has helped me.  I do explictly review each suggestion it makes.</p>", "time": "2026-11-18T18:07:27.000Z"}, {"author": "Robert Wilton", "text": "<p>+1 to Bron.</p>", "time": "2026-11-18T18:07:29.000Z"}, {"author": "Mirja K\u00fchlewind", "text": "<p>Isn\u2019t that meaningless if basically everybody uses AI to some extend?</p>", "time": "2026-11-18T18:07:34.000Z"}, {"author": "Robert Carolina", "text": "<p>+1 Bron</p>", "time": "2026-11-18T18:07:38.000Z"}, {"author": "Mark Nottingham", "text": "<p>+1 Bron</p>", "time": "2026-11-18T18:07:39.000Z"}, {"author": "Jaime Jimenez", "text": "<p>+1 to bron</p>", "time": "2026-11-18T18:07:40.000Z"}, {"author": "Glenn Deen", "text": "<p>+! to Bron</p>", "time": "2026-11-18T18:07:47.000Z"}, {"author": "Suresh Krishnan", "text": "<p>+1 Bron</p>", "time": "2026-11-18T18:07:50.000Z"}, {"author": "Steve Traugott", "text": "<p>In practice, accurate disclosure means somehow separating generated vs tweaked vs handwritten -- it's typically all interleaved in the same git commit.  I don't see how enforcing it would work.</p>", "time": "2026-11-18T18:07:56.000Z"}, {"author": "S Moonesamy", "text": "<p>@Mark Atwood I submitted a draft.  There are references to past RFC Editor policies in there.  You can build on that.</p>", "time": "2026-11-18T18:08:15.000Z"}, {"author": "Andrew Campling", "text": "<p>+1 to Bron</p>", "time": "2026-11-18T18:08:17.000Z"}, {"author": "Nick Doty", "text": "<p>knowing who's responsible isn't helpful to me -- people are happy to take responsibility for the long dismissive writing that they didn't bother to write or read</p>", "time": "2026-11-18T18:08:28.000Z"}, {"author": "Dhruv Dhody", "text": "<p>Support for AI-use disclosure kinda goes hand in hand with a non-judgmental attitude toward AI use....</p>", "time": "2026-11-18T18:09:04.000Z"}, {"author": "David Schinazi", "text": "<p>There's no point having a requirement if it's not paired with enforcement</p>", "time": "2026-11-18T18:09:13.000Z"}, {"author": "Brian Trammell", "text": "<p>there's a difference between \"I used an LLM to translate my thoughts\" and \"I don't understand the base concepts in this area and I let a flash model slop up a double-zero\"</p>", "time": "2026-11-18T18:09:14.000Z"}, {"author": "Jen Linkova", "text": "<p>People can write very usefu drafts using AI and 100% organic drafts can be useless. Do we really care if a paragraph of a text was polished using AI, written by AI completely or just spell-checked?</p>", "time": "2026-11-18T18:09:20.000Z"}, {"author": "Jaime Jimenez", "text": "<p>Flagging drafts based on style deviation, vocabulary, etc costs nothing. Using model ensembles in addition to that costs few cents per draft (if you want).</p>", "time": "2026-11-18T18:09:20.000Z"}, {"author": "Mark Nottingham", "text": "<p>+1 MT. A disclosure requirement puts the incentives in the wrong direction.</p>", "time": "2026-11-18T18:09:26.000Z"}, {"author": "Phillip Hallam-Baker", "text": "<p>[repeat as it glitched] EKR, are you using a model could run on one of my idle hosts?</p>", "time": "2026-11-18T18:09:34.000Z"}, {"author": "Cullen Jennings", "text": "<p>The key thing is that when I send text to IETF, IETF has certain rights. That I used AI should not change that</p>", "time": "2026-11-18T18:09:36.000Z"}, {"author": "Brian Carpenter", "text": "<p>@David, it is indeed uneneforceable</p>", "time": "2026-11-18T18:09:44.000Z"}, {"author": "Andrew Campling", "text": "<p>I'm less concerned about AI-assisted authoring of documents than by the use of agents to autonomously monitor and respond to mailing lists</p>", "time": "2026-11-18T18:10:19.000Z"}, {"author": "Andrew Campling", "text": "<p>The day the roof fell in on the IETF....</p>", "time": "2026-11-18T18:10:58.000Z"}, {"author": "Steve Traugott", "text": "<p>Hey, us remote folks could keep going...  ;-)</p>", "time": "2026-11-18T18:11:04.000Z"}, {"author": "zhang jun", "text": "<p>in nature, ai is an information to information function.</p>", "time": "2026-11-18T18:11:39.000Z"}, {"author": "Brian Carpenter", "text": "<p>We'll see what EKR's AI does with this</p>", "time": "2026-11-18T18:11:40.000Z"}, {"author": "Brian Carpenter", "text": "<p>I got the message that requiring disclosure is a non-starter</p>", "time": "2026-11-18T18:12:20.000Z"}, {"author": "Monika Ermert", "text": "<p>is it a problem for the IETF IPR management if people create with AI (with or without disclosing)</p>", "time": "2026-11-18T18:12:42.000Z"}, {"author": "Nick Doty", "text": "<p>for those of us inundated with reading all of your LLM-generated text, it would be very helpful for me to have a disclosure, so I can allocate my time efficiently</p>", "time": "2026-11-18T18:12:48.000Z"}, {"author": "Steve Traugott", "text": "<p>for the record -- physical room had to be evacuated at this point because el nino and a roof leak</p>", "time": "2026-11-18T18:13:34.000Z"}, {"author": "Nick Doty", "text": "<p>perhaps it would turn out that almost everyone discloses and the signal would eventually mean less. in the meantime, it would be useful to readers.</p>", "time": "2026-11-18T18:13:36.000Z"}, {"author": "Colin Perkins", "text": "<p>We can watch the room gradually fill with water :)</p>", "time": "2026-11-18T18:13:40.000Z"}, {"author": "Brian Carpenter", "text": "<p>@Nick, what if the content is useful?</p>", "time": "2026-11-18T18:13:56.000Z"}, {"author": "Nick Doty", "text": "<p>@Brian, if the content is useful, great! but if I have too much to read, I know where to spend my time</p>", "time": "2026-11-18T18:14:22.000Z"}, {"author": "Nick Doty", "text": "<p>for example, sometimes people write something that seems nonsensical. if I know it's AI-generated, I can dismiss it. if I know it's not AI-generated, I know it's worth spending the time to understand and clarify with them</p>", "time": "2026-11-18T18:14:55.000Z"}]