[{"author": "Nick Doty", "text": "

I'm confused, is this a tool to facilitate and accelerate the existential threat to the IETF?

", "time": "2026-07-23T14:42:46.000Z"}, {"author": "Ryo Yanagida", "text": "

'let's do some damage control' ? maybe?

", "time": "2026-07-23T14:43:26.000Z"}, {"author": "Susan Hares", "text": "

@mark are the iesg ballots and the iesg minutes both inputed.into the tool.

", "time": "2026-07-23T14:44:12.000Z"}, {"author": "Priyanka Sinha", "text": "

Very positive talk for me .. I think the policy considerations would be addressed by better and larger LLMs mostly --
\nIt would be awesome .. any further experimentation with a local LLM and say a open source LLM coding platform would be wonderful to try .. or even hosted by IESG/IETF/IRTF

", "time": "2026-07-23T14:49:07.000Z"}, {"author": "Susan Hares", "text": "

@mark - do AI tools aid sealioning?

", "time": "2026-07-23T14:50:00.000Z"}, {"author": "Dhruv Dhody", "text": "

Like any tools it is used in all sorts of ways

", "time": "2026-07-23T14:51:07.000Z"}, {"author": "Dan York", "text": "

Thanks for putting this all together, Mark.

", "time": "2026-07-23T14:51:42.000Z"}, {"author": "Susan Hares", "text": "

+thanks for creating the tools.

", "time": "2026-07-23T14:52:08.000Z"}, {"author": "avri doria", "text": "

What a fine effort. thanks.

", "time": "2026-07-23T14:52:15.000Z"}, {"author": "Kaliya Young", "text": "

You could insist on proof of personhood - the Linux Foundation is moving this direction - https://lfms26.sched.com/event/2ETT5/facilitated-discussion-the-first-person-linux-kernel-maintainer-verification-project-greg-kroah-hartman-linux-kernel-maintainer-daniela-barbosa-hart-montgomery-lf-decentralized-trust-glenn-gore-affinidi

", "time": "2026-07-23T14:52:27.000Z"}, {"author": "Alexandr Railean", "text": "

Maybe another thing - proof of thought: if you use AI, include your prompt and the input you gave it.

", "time": "2026-07-23T14:53:31.000Z"}, {"author": "Tom Newton", "text": "

If someone goes as far as getting Mark's tool, there's a degree to which that demonstrates engagement (and wanting to get things right)

", "time": "2026-07-23T14:58:46.000Z"}, {"author": "Chris Box", "text": "

Chairs: I suspect this is a really valuable conversation for the IETF so if you feel able to extend the timeslot a little that might be worthwhile

", "time": "2026-07-23T14:59:34.000Z"}, {"author": "Ignacio Castro", "text": "

yes, we cut on the welcome to be able to do that

", "time": "2026-07-23T14:59:55.000Z"}, {"author": "Ignacio Castro", "text": "

stilll, lots of questions :-)

", "time": "2026-07-23T15:00:13.000Z"}, {"author": "Nick Doty", "text": "

I agree people are using it (and we aren't going to be able to ban it), but I don't see how we are helping them by providing the tools to use it this way

", "time": "2026-07-23T15:00:43.000Z"}, {"author": "Nick Doty", "text": "

if the norm becomes \"don't ask questions, just ask Mark's LLM\", that's bad for the users, and for the community, and gives a lot of power to the author of the tool

", "time": "2026-07-23T15:01:55.000Z"}, {"author": "Priyanka Sinha", "text": "

@nick the most visible impact I think with this kind of accuracy with the embeddings would be to accelerate the review backlog for IESG

", "time": "2026-07-23T15:01:56.000Z"}, {"author": "Tommy Jensen", "text": "

@Nick, what would you envision would be the correct way for a participant to use AI? The answer helps directly inform formalized community guidelines

", "time": "2026-07-23T15:02:13.000Z"}, {"author": "Dhruv Dhody", "text": "

Looks like we need blue tick marks in IETF - certified person?

", "time": "2026-07-23T15:02:26.000Z"}, {"author": "Alexandr Railean", "text": "

@nick If you don't provide them, they're easy to replicate. If you provide them - you get control over what virtues those tools are imbued with.

", "time": "2026-07-23T15:02:42.000Z"}, {"author": "Nick Doty", "text": "

@priyanka, in that the IESG should outsource their reviews to an LLM?

", "time": "2026-07-23T15:02:49.000Z"}, {"author": "Nick Doty", "text": "

@alexandr control over the the virtues! interesting. I wonder if any attacker could change the instructions from \"do disclose\" to \"don't disclose\".

", "time": "2026-07-23T15:03:16.000Z"}, {"author": "Priyanka Sinha", "text": "

@nick they can use it as a more effective tool .. like we use calculators in accounting .. I work in AI accounting in my day job :-)

", "time": "2026-07-23T15:03:45.000Z"}, {"author": "Jean Queralt", "text": "

The IETF should explore standardizing network prompting language.
\nI can't see how we will avoid people resolving their networking needs by prompting their swarms and go with it.

", "time": "2026-07-23T15:04:13.000Z"}, {"author": "Carolina Caeiro", "text": "

Am I beyond the lock, Ignacio?

", "time": "2026-07-23T15:04:14.000Z"}, {"author": "Alexandr Railean", "text": "

@nick use remote attestation to ensure they're using the right tools ;)

", "time": "2026-07-23T15:04:18.000Z"}, {"author": "Ignacio Castro", "text": "

in if you can be very succinct ;0

", "time": "2026-07-23T15:04:30.000Z"}, {"author": "Nick Doty", "text": "

@priyanka, I guess I'm waiting for evidence that it improves reviews. my experience so far is that it does the opposite. misleads the reviewer, makes mistakes, encourages the reviewer to stop paying attention

", "time": "2026-07-23T15:04:35.000Z"}, {"author": "Dhruv Dhody", "text": "

Things like Yang reviews for sure tools are good

", "time": "2026-07-23T15:04:56.000Z"}, {"author": "Priyanka Sinha", "text": "

@nick I agree to that

", "time": "2026-07-23T15:05:00.000Z"}, {"author": "Lixia Zhang", "text": "

@jean that's a great idea

", "time": "2026-07-23T15:05:28.000Z"}, {"author": "Andrew Campling", "text": "

On using AI to comment on topics that you have no understanding, there seems to be a long tradition of people doing that anyway

", "time": "2026-07-23T15:05:29.000Z"}, {"author": "Tommy Jensen", "text": "

+1 Andrew, a perfect example of AI only being new in its rate of production.

", "time": "2026-07-23T15:06:24.000Z"}, {"author": "Lixia Zhang", "text": "

@Tommy, people do whatever they want in the absence of accountability

", "time": "2026-07-23T15:07:24.000Z"}, {"author": "Dhruv Dhody", "text": "

But who is actually reviewing all the -00 drafts?

", "time": "2026-07-23T15:07:29.000Z"}, {"author": "Jean Queralt", "text": "

@Lixia I did a HotRFC in 120 about \"What does a post-IETF world look like\" exploring the topic.

", "time": "2026-07-23T15:07:34.000Z"}, {"author": "Dhruv Dhody", "text": "

Unless it is actuallly on the agenda

", "time": "2026-07-23T15:07:41.000Z"}, {"author": "Nick Doty", "text": "

@tommy, I think translation and editing for communication clarity would be good uses.

", "time": "2026-07-23T15:07:44.000Z"}, {"author": "Jay Daley", "text": "

Requiring proof of being human won't really stop much - the most dangerous issue is real humans using AI to engage on topics they know nothing about

", "time": "2026-07-23T15:08:54.000Z"}, {"author": "Lixia Zhang", "text": "

AI can help us analyze communications, expose everything on the table (yes this needs to be regulated / controlled to avoid negative impacts)

", "time": "2026-07-23T15:09:35.000Z"}, {"author": "Nick Doty", "text": "

+1 jay, at least for now, the misuses are often approved by a human

", "time": "2026-07-23T15:09:37.000Z"}, {"author": "Lixia Zhang", "text": "

@Jay, see my comment on accountability

", "time": "2026-07-23T15:10:24.000Z"}, {"author": "Dhruv Dhody", "text": "

Many people also right drafts by borrowing styles from existing drafts and RFCs and modify it for your extensions for example! I wonder if those things would be flagged as AI written even if done manually?

", "time": "2026-07-23T15:12:06.000Z"}, {"author": "Alexandr Railean", "text": "

What does it show for drafts written 10..20 years ago?

", "time": "2026-07-23T15:12:30.000Z"}, {"author": "John Klensin", "text": "

Jay, I agree... but real humans aggressively commenting about comments they know nothing about is already a serious enough problem --and one that have not dealt with well in the last couple of decades-- that I'm not sure how much worse AI involvement would make it.

", "time": "2026-07-23T15:12:48.000Z"}, {"author": "John Klensin", "text": "

Perhaps, at worst, quantitative problem rather than a qualitative one.

", "time": "2026-07-23T15:13:09.000Z"}, {"author": "Lixia Zhang", "text": "

@Dhruv AI can learn to differentiate in anyway one wants

", "time": "2026-07-23T15:13:17.000Z"}, {"author": "Dhruv Dhody", "text": "

Maybe we should tackle the behavior irrespective of it comes with or without Ai

", "time": "2026-07-23T15:13:31.000Z"}, {"author": "Lixia Zhang", "text": "

yes

", "time": "2026-07-23T15:13:54.000Z"}, {"author": "John Klensin", "text": "

That might well be a better approach

", "time": "2026-07-23T15:14:00.000Z"}, {"author": "Tommy Jensen", "text": "

Lixia, I hear you. I don't have succinct thoughts for chat prior to proposal text, but I do hear you that rules without accountability aren't helpful

", "time": "2026-07-23T15:14:05.000Z"}, {"author": "avri doria", "text": "

AI's role in job generation: human in the loop verifiers.

", "time": "2026-07-23T15:14:14.000Z"}, {"author": "Tommy Jensen", "text": "

Yes, exactly what Dhruv is saying. Overloading a list via AI or via unpaid sock puppet interns is the same problem.

", "time": "2026-07-23T15:14:37.000Z"}, {"author": "Andrew Campling", "text": "

+1 to tackling the behaviour. As I said on the mic, reviewing everything to see where we assume good faith engagement and making adjustments would be a great place to start.

", "time": "2026-07-23T15:15:51.000Z"}, {"author": "Jay Daley", "text": "

@Lixia I agree on the need for visibility - see the tool I (Claude) wrote yesterday: https://github.com/JayDaley/mailing-list-ai-check

", "time": "2026-07-23T15:16:03.000Z"}, {"author": "Miracle Gerrald", "text": "

AI should be used responsibly, and used as an accelerator and assistant. We cannot just replace and fully trust the AI to do the in-depth research for us. We have to re-check its validity, ensuring the response is reliable yet accountable.

\n

For newcomers, I think one of the good practice for using AI, is being proactive with them and asking for a reference before sending the prompt/instructions. Therefore we can at least verify if the answers are legit or not.

\n

Setting up an MCP server, with a custom instructions to see the provided resources first before searching the rest on the Internet, is also another good option to try.

", "time": "2026-07-23T15:16:14.000Z"}, {"author": "Tommy Jensen", "text": "

The other thing that comes to mind with this and the previous presentation: we need to avoid talking about AI as acting on behalf of anybody. The human bears the same responsibility for their behavior and content whether they do it, an LLM does it, or a non-LLM text generator does it

", "time": "2026-07-23T15:16:22.000Z"}, {"author": "Tommy Jensen", "text": "

hm, rephrase: the human is the only actor. everything else is a non-participating tool.

", "time": "2026-07-23T15:17:30.000Z"}, {"author": "Lixia Zhang", "text": "

yes human is the only responsible actor, must take responsibility for his/her AI tools actions

", "time": "2026-07-23T15:19:23.000Z"}, {"author": "Priyanka Sinha", "text": "

I wholly agree with you @Tommy .. and I have a granted patent that points towards this ideology. US9591087B2

", "time": "2026-07-23T15:20:47.000Z"}, {"author": "Tommy Jensen", "text": "

Agreed, but I would even go so far as say only actor. What I'm thinking is if anything is acting on my behalf, then by definition I haven't seen the final form of content before it is submitted, which is an automatic failure in responsibility.

", "time": "2026-07-23T15:20:48.000Z"}, {"author": "Tommy Jensen", "text": "

(have to run to conflict, thank you to both speakers and the great discussion)

", "time": "2026-07-23T15:21:10.000Z"}, {"author": "Lixia Zhang", "text": "

agree

", "time": "2026-07-23T15:21:25.000Z"}, {"author": "Chris Box", "text": "

Alvaro I suggest the AI interim would be worth publicising on ietf@ietf.org as it's likely to be of wide interest. (And personally speaking I'm not subscribed to rasprg.)

", "time": "2026-07-23T15:28:57.000Z"}, {"author": "Ignacio Castro", "text": "

makes sense

", "time": "2026-07-23T15:29:27.000Z"}, {"author": "Susan Hares", "text": "

To presenter do you track mortality and the IESG review state?

", "time": "2026-07-23T15:33:03.000Z"}, {"author": "Dhruv Dhody", "text": "

I-D to RFCs takes years so year on year correlations would not be there

", "time": "2026-07-23T15:33:57.000Z"}, {"author": "Ignacio Castro", "text": "

indeed, and increased substantially over time (3 years median from 00 to RFC in recent years)

", "time": "2026-07-23T15:34:37.000Z"}, {"author": "Carolina Caeiro", "text": "

wonder how they estimated associated region for the proposals

", "time": "2026-07-23T15:38:04.000Z"}, {"author": "John Klensin", "text": "

Be interesting to see variations on some of the graphs scaled by participants

", "time": "2026-07-23T15:38:13.000Z"}, {"author": "Susan Hares", "text": "

RfC publication rated are impacted by IESG review cycle.

", "time": "2026-07-23T15:39:51.000Z"}, {"author": "MICHEL SALES BONFIM", "text": "

Ow, this tool really exists

", "time": "2026-07-23T15:41:00.000Z"}, {"author": "John Klensin", "text": "

Susan, of course. And my comment about scaling applies there too -- many of these graphs and statistics might be more useful if scaled by various external factors -- IESG review cycles, regional participation, etc.

", "time": "2026-07-23T15:42:24.000Z"}, {"author": "Dhruv Dhody", "text": "

I don\u2019t think here IESG state was not considered at all

", "time": "2026-07-23T15:43:54.000Z"}, {"author": "Dhruv Dhody", "text": "

Not all drafts are meant to be RFcs

", "time": "2026-07-23T15:45:24.000Z"}, {"author": "Ignacio Castro", "text": "

https://dl.acm.org/doi/10.1145/3744200.3744757 on affiliations

", "time": "2026-07-23T15:47:58.000Z"}, {"author": "Dhruv Dhody", "text": "

Maybe a new state in datatracker

", "time": "2026-07-23T15:49:12.000Z"}, {"author": "Chris Box", "text": "

If the charts are measuring the same thing (no. of emails in time interval) then IETF has about 30x more volume than RIPE

", "time": "2026-07-23T15:52:45.000Z"}, {"author": "Ignacio Castro", "text": "

Beyond the population size in each org, the need for email communication varies, so would look at trends rather than absolute value

", "time": "2026-07-23T15:53:44.000Z"}, {"author": "Chris Box", "text": "

I was offering no opinion on which one was better. There are arguments for both.

", "time": "2026-07-23T15:54:33.000Z"}, {"author": "Dhruv Dhody", "text": "

Older generation or just longer being on the list

", "time": "2026-07-23T15:56:25.000Z"}, {"author": "Dhruv Dhody", "text": "

Need some correlation to understand at what age does one join a community

", "time": "2026-07-23T15:57:29.000Z"}, {"author": "Nick Doty", "text": "

I'm curious about the shift to other communication tools/fora

", "time": "2026-07-23T16:07:28.000Z"}, {"author": "Tommy Jensen", "text": "

+1 literally was just about to ask about GitHub stealing from email

", "time": "2026-07-23T16:08:02.000Z"}, {"author": "Nick Doty", "text": "

we did some work on trying to track multiple email addresses in BigBang, around From addresses and names like that

", "time": "2026-07-23T16:10:06.000Z"}, {"author": "Dhruv Dhody", "text": "

But if you made it to IETF you are likely used to mailing list already even if not preferred

", "time": "2026-07-23T16:11:11.000Z"}, {"author": "Ignacio Castro", "text": "

Also did some mapping to unique identifiershttps://sodestream.github.io/paper-the-web-we-weave-untangling-the-social-graph-of-the-ietf.html

", "time": "2026-07-23T16:11:34.000Z"}, {"author": "Tom Newton", "text": "

I remember posting \"mailing lists are important and fun\" was the first task we were given as undergrads (late 90s Cs course)

", "time": "2026-07-23T16:12:04.000Z"}, {"author": "Chris Box", "text": "

The way slide 3 was drawn tends to imply (to me at least) that implementations occur after the standard is developed. As we know, running code is frequently an integral part of the development process. I haven't read the draft so I don't know if the draft is clearer on that point?

", "time": "2026-07-23T16:18:05.000Z"}, {"author": "Nick Doty", "text": "

I would love to see that Huawei slide

", "time": "2026-07-23T16:22:51.000Z"}, {"author": "Nick Doty", "text": "

I was surprised by how confusing it was even with things like Cisco, and different segments or representations of Cisco

", "time": "2026-07-23T16:23:10.000Z"}, {"author": "Ryo Yanagida", "text": "

@Nick Doty https://datatracker.ietf.org/meeting/interim-2026-rasprg-01/materials/slides-interim-2026-rasprg-01-sessa-analysing-internet-standards-development-organisation-data-00
\nSlide 7

", "time": "2026-07-23T16:25:11.000Z"}, {"author": "Nick Doty", "text": "

:clap:

", "time": "2026-07-23T16:25:43.000Z"}, {"author": "Ryo Yanagida", "text": "

loads of fun

", "time": "2026-07-23T16:25:49.000Z"}, {"author": "Chris Box", "text": "

Sounds like there is value in starting to routinely collect affiliation data. Would also allow proof of humanity.

", "time": "2026-07-23T16:30:07.000Z"}, {"author": "Nick Doty", "text": "

I think if the authors are willing to work with rasprg, I think it would be great for rasprg to adopt and eventually publish this

", "time": "2026-07-23T16:30:39.000Z"}, {"author": "Nick Doty", "text": "

lack of datatracker documentation is a really important finding and concrete recommendation

", "time": "2026-07-23T16:33:20.000Z"}, {"author": "Nick Doty", "text": "

thanks all for excellent presentations!

", "time": "2026-07-23T16:35:36.000Z"}]