21:53:35 RRSAgent has joined #semantics-for-agentic 21:53:39 logging to https://www.w3.org/2025/11/11-semantics-for-agentic-irc 21:53:39 RRSAgent, do not leave 21:53:40 RRSAgent, this meeting spans midnight 21:53:40 RRSAgent, make logs public 21:53:42 Meeting: Semantics for the Agentic Web 21:53:42 Chair: Penelope McLachlan 21:53:42 Agenda: https://github.com/w3c/tpac2025-breakouts/issues/48 21:53:42 Zakim has joined #semantics-for-agentic 21:53:43 Zakim, clear agenda 21:53:43 agenda cleared 21:53:43 Zakim, agenda+ Pick a scribe 21:53:44 agendum 1 added 21:53:44 Zakim, agenda+ Reminders: code of conduct, health policies, recorded session policy 21:53:44 agendum 2 added 21:53:44 Zakim, agenda+ Goal of this session 21:53:45 agendum 3 added 21:53:45 Zakim, agenda+ Discussion 21:53:45 agendum 4 added 21:53:46 Zakim, agenda+ Next steps / where discussion continues 21:53:47 agendum 5 added 21:53:47 Zakim, agenda+ Adjourn / Use IRC command: Zakim, end meeting 21:53:47 agendum 6 added 21:53:47 breakout-bot has left #semantics-for-agentic 23:27:10 ohmata has joined #semantics-for-agentic 23:30:10 noamr has joined #semantics-for-agentic 23:30:11 Siri has joined #semantics-for-agentic 23:30:22 Present+ 23:30:27 johannhof has joined #semantics-for-agentic 23:30:35 present+ 23:30:40 smcgruer_[EST] has joined #semantics-for-agentic 23:31:09 dom has joined #semantics-for-agentic 23:31:26 alisonmaher has joined #semantics-for-agentic 23:31:29 Penny has joined #semantics-for-agentic 23:31:48 present+ Stephen_McGruer 23:31:50 tantek-projector has joined #semantics-for-agentic 23:31:58 Penny has joined #semantics-for-agentic 23:32:03 q? 23:32:30 scribenick: smcgruer_[EST] 23:32:34 Present+ 23:32:41 present+ 23:32:45 present+ 23:32:54 q+ 23:33:06 ack johannhof 23:33:21 whsieh has joined #semantics-for-agentic 23:33:23 ari has joined #semantics-for-agentic 23:33:30 mehm8128 has joined #semantics-for-agentic 23:33:32 csarven has joined #semantics-for-agentic 23:33:34 Jemma has joined #semantics-for-agentic 23:33:35 mattreynolds has joined #semantics-for-agentic 23:33:42 present+ 23:33:44 present+ 23:33:45 present+ 23:33:52 present+ 23:33:52 kgovind has joined #semantics-for-agentic 23:33:55 ktoumura has joined #semantics-for-agentic 23:33:56 ErikAnderson has joined #semantics-for-agentic 23:33:57 present+ 23:34:01 present+ 23:34:05 nournabil has joined #semantics-for-agentic 23:34:11 Topic: Semantics for the Agentic Web 23:34:12 jamesn has joined #semantics-for-agentic 23:34:14 benvds has joined #semantics-for-agentic 23:34:17 Present+Siri, ValdimirLevin, HisayyuiOhma,YuichiMorioka, gerchiko, WensonHsieh, GabrieBrtio, KaustubhaGov, MattGarish, NourNabil, HirokiEndo, AshwanyRayu, ShivaniSharm 23:34:24 lei_zhao has joined #semantics-for-agentic 23:34:25 lionel has joined #semantics-for-agentic 23:34:28 Penny: I will present a few slides, but aiming for an open discussion afterwards 23:34:32 present+ 23:34:32 DP has joined #semantics-for-agentic 23:34:39 cfredric has joined #semantics-for-agentic 23:34:40 LeoLee has joined #semantics-for-agentic 23:34:45 present+ 23:34:46 vmpstr has joined #semantics-for-agentic 23:34:48 [We look at Pennys slides] 23:34:51 present+ 23:34:51 present+ 23:34:54 present+ 23:34:59 Penny: The agentic web is already here 23:35:02 [slide 2] 23:35:07 present+ 23:35:16 ... I expect us to see a trough of disillusionment, people will learn about the brittleness of agents 23:35:36 RobKochman has joined #semantics-for-agentic 23:36:08 kohei has joined #semantics-for-agentic 23:36:08 ... However eventually users will start to delegate full complex and high-risk transactions to agents 23:36:18 ... And getting 99.9% is not acceptable, given bad actors 23:36:18 can someone share the slide url? 23:36:27 Shivani has joined #semantics-for-agentic 23:36:44 pchampin has joined #semantics-for-agentic 23:36:54 [slide 3] 23:37:07 i|[slide 2]|Slideset: TBD 23:37:08 [The slides will be available after, with apologies] 23:37:19 kbx has joined #semantics-for-agentic 23:37:26 Yuichi has joined #semantics-for-agentic 23:37:32 Penny: Today, agents use various fragile parsing mechanisms, often visual 23:37:44 hiroki has joined #semantics-for-agentic 23:37:49 Haruki has joined #semantics-for-agentic 23:38:18 q? 23:38:19 mbgower has joined #semantics-for-agentic 23:38:30 Penny: On top of this, models are constantly changing, which also introduces uncertainty 23:38:47 Pavan has joined #semantics-for-agentic 23:38:53 chrishtr has joined #semantics-for-agentic 23:39:14 Penny: I suspect we need to go beyond DOM or reading pixels for acceptable success rate 23:39:24 hagio has joined #semantics-for-agentic 23:39:27 Penny: So the question here is - is ARIA the answer? 23:39:32 ... ChatGPT Atlas already uses it 23:39:42 [slide 4] 23:39:43 ... It seems nice to tell developers that they need to get ARIA right 23:39:54 ... But the risk is that we may need to tell robots different things to humans 23:40:03 ... And we risk degrading the ARIA-driven experience for real humans 23:40:11 ... E.g. if we want to lie to the robot, or give more information than a human needs, etc 23:40:16 [slide 5] 23:40:32 Penny: For example, trashcan icon on application. Does this mean archive, or delete? 23:40:51 ... Without clear signals, the agent may wipe data accidentally 23:40:56 Matt_King has joined #semantics-for-agentic 23:41:32 Penny: And there's a performance and sustainability problem, pixel scraping is probably a lot more expensive than just looking at the accessibility tree 23:41:46 [slide 6] 23:41:51 Penny: Onto the discussion. Some question seeds: 23:42:02 ... Do we accept the premise? Can we achieve 100% reliability just using inference? 23:42:11 Tatsuya has joined #semantics-for-agentic 23:42:11 ... Are we comfortable conflating human accessibility with machine capability? 23:42:14 q+ to say I launched a small trial balloon in this direction https://github.com/w3c/aria/issues/2046 23:42:23 ... Do we need dedicated agent augmentation for HTML markup 23:42:23 hiro has joined #semantics-for-agentic 23:42:26 ack mbgower 23:42:26 mbgower, you wanted to say I launched a small trial balloon in this direction https://github.com/w3c/aria/issues/2046 23:42:49 mbgower: A few years ago at IBM, we had the visual cues for AI material and no way to do it programmatically 23:42:52 dholbert has joined #semantics-for-agentic 23:43:02 ... At the time, we wanted some kind of semantic structure to deal with non-accessible related information 23:43:07 tomayac has joined #semantics-for-agentic 23:43:11 wendyreid has joined #semantics-for-agentic 23:43:16 present+ 23:43:16 Present+ Thomas_Steiner 23:43:18 ... ARIA seemed like an interesting area to bring it up, but at the time ARIA group said it wasn't an accessibility issue, which was somewhat valid 23:43:21 q+ 23:43:25 q+ 23:43:40 mbgower: I don't believe we will get it 100% by inference. I am concerned with conflation, but I accept it may be necessary. 23:43:41 q? 23:43:43 q+ 23:43:49 ack ErikAnderson 23:43:57 q+ 23:44:03 kush has joined #semantics-for-agentic 23:44:06 ErikAnderson: I as a human cannot get 100% reliability, but I take responsibility and/or blame the site for bad design 23:44:10 front-endian-jane has joined #semantics-for-agentic 23:44:17 ... I think its important we do something here to help agents get closer to 100% 23:44:24 mjwilson has joined #semantics-for-agentic 23:44:29 Penny: I think there's an analogy to self-driving cars - the bar is higher than for humans 23:44:34 ... I think its similar for agents on the web 23:44:38 ack benvds 23:45:09 q+ Isaac 23:45:19 vasilii has joined #semantics-for-agentic 23:45:20 benvds: Very interesting + important. On representing things differently between agents and users is possibly dangerous. Lying to agents becomes an attack, can lead to differences in serving. 23:45:29 q- 23:45:30 ... I also like labelling semantics as destructive vs mutating but reversible 23:45:35 ack wendyreid 23:45:50 Dingwei5 has joined #semantics-for-agentic 23:45:53 christianliebel has joined #semantics-for-agentic 23:46:31 wendyreid: One big problem we already is how context drives interpretation, e.g. edit icon on one site could be highlight on another. Accessibility when done well helps with that (e.g. via labels) but we cannot rely on that today for humans 23:46:38 ... I do love the point of us having a higher bar, however 23:46:43 present+ to mention the W3C RDFa recommendation can express any significant unit of information at any level of granularity in a document, which can be used in any markup language 23:46:48 ack 23:46:55 q+ mjwilson 23:47:01 ack Isaac 23:47:03 reillyg has joined #semantics-for-agentic 23:47:04 q+ 23:47:16 q+ 23:47:47 kbx has joined #semantics-for-agentic 23:47:51 Isaac: Is the question here reliability for b2c or b2b? There's a trade off where we might accept less reliability to be able to achieve things that humans cannot 23:47:59 Present+ Kenji_Baheux 23:48:27 q+ to mention the W3C RDFa recommendation can express any significant unit of information at any level of granularity in a document, which can be used in any markup language 23:48:29 q+ 23:48:36 Penny: Agree. Hopefully nobody is trusting agentic systems in a life or death situation. But in terms of money, there's a lot more flowing in b2b or b2c scenarios. So imagine an agentic based clearing house, what happens there if we don't have 100% certainty. 23:48:41 q+ 23:48:47 GabrielBrito has joined #semantics-for-agentic 23:48:51 q+ 23:49:05 Isaac: Definitely, not suggesting b2b means less reliability. I think there may still be value tradeoffs that are more common in b2b 23:49:15 q? 23:49:18 COGA did a presentation at CSUn back in 2022 about how ARIA could be levered in a way similar to what I infer is an approach here. I can't find the presentation, but here's the original doc https://github.com/w3c/aria/issues/2046 23:49:37 ack mjwilson 23:50:00 mjwilson: I think there's a risk of overloading ARIA. I think there's a similar risk with dedicated tags, because developers may implement those in place of ARIA 23:50:02 ack jamesn 23:50:02 q+ 23:50:25 jamesn: the ARIA technique, agentic browsers are going to use it whether we like it or not. I think we should be giving an alternative, and avoid ARIA being misused 23:50:41 ... if we rely on that being the only way for robots, I think it will make the web worse for humans 23:50:42 q+ 23:50:51 ... ARIA should be a fallback, not the main approach 23:50:56 +1 jamesn 23:51:28 reillyg: Agree with last two comments, I am concerned that sites will try to block bots and they will use ARIA as a way to do that, and end up blocking users who use assistive technologies 23:51:36 q+ 23:51:45 ack reillyg 23:51:46 ... It might be prudent for browser developers to say "on my browser, ARIA will be for humans" 23:51:47 ack reillyg 23:51:48 ack csarven 23:51:48 csarven, you wanted to mention the W3C RDFa recommendation can express any significant unit of information at any level of granularity in a document, which can be used in any 23:51:52 ... markup language 23:52:07 dmurph has joined #semantics-for-agentic 23:52:07 csarven: I think this problem is addressed by existing w3c recommendation (ARIA [something]) 23:52:18 ... Basically orthogonal attributes that don't interfere with ARIA 23:52:22 ... Attributes in HTML elements 23:52:30 s/ARIA [something]/RDFa 23:52:36 ... Can express the content that the human is reading, or something that is only for machines 23:52:52 ... So no duplication of labels since it can be for both, or it can just be for machines 23:52:54 +q 23:53:00 ... I strongly recommend reviewing RFDa 23:53:07 ... I would also like to see more concrete examples of use cases 23:53:11 q+ 23:53:12 q+ to mention the first rule of ARIA, NLWeb 23:53:24 ack kbx 23:53:28 RDFa: https://www.w3.org/TR/rdfa-core/ , https://www.w3.org/TR/html-rdfa/ , https://www.w3.org/TR/xhtml-rdfa-scenarios/ 23:53:30 Penny: Look forward to exploring that! 23:53:43 lionel has joined #semantics-for-agentic 23:53:59 Victor has joined #semantics-for-agentic 23:54:04 kbx: I am worried that developers won't know what they have to explain to models. 23:54:11 q+ to mention "AX" (agentic experience) 23:54:14 ... Agents are not regular users, this might be hard 23:54:17 mvono has joined #semantics-for-agentic 23:54:23 q+ 23:54:35 Penny: So that I understand, its that the model may need very different information than humans? 23:54:56 kbx: Yes. So things like "these icons make sense to my users, because of wider context, but a model has no idea" 23:54:56 s/RFDa/RDFa 23:54:58 dotproto has joined #semantics-for-agentic 23:55:11 ack vasilii 23:55:14 vasilii has joined #semantics-for-agentic 23:55:35 vasilii: For agentic systems to have utility, there needs to be scale. So I want to explore how we can do that today 23:55:57 ... I really like the point of building in a way that works for both humans and agents, without building two separate things 23:56:01 One of your examples seemed to be a challenge with inferring the meaning of a button function based on a crappy label. I did a little exploration of how this could potentially be done just by using an established vocabulary. So it's something of an inferred approach, but it provides some underlying support. I'll see if I can dig it up 23:56:07 ... to avoid expense and also avoid excluding systems that don't adapt 23:56:11 atai has joined #semantics-for-agentic 23:56:22 q+ 23:56:25 ... I also agree that there's still place for deterministic systems, not everything should be agentic 23:56:32 ack benvds 23:56:51 benvds: Conflating accessibility and machine capability - to me the answer is no, based on what has been discussed 23:56:59 ... However, you can still use the technology without conflating it 23:57:20 ... There was/is a Better Together breakout Monday, looking at expanding ARIA 23:57:26 q+ to discuss ARIA and the accessibility tree - and how we can't rely on aria labelling and roles unless the tree is created 23:57:36 ... The goal is to give benefit to assistive technologies without having it be taken over by robot use cases 23:57:37 ack Penny 23:58:02 Penny: Feels like temperature of the room is trending to don't conflate, can augment using the accessibility tree but don't rely on it 23:58:20 ... One question that comes to mind is that any solution we add here needs to be very low effort to developer and also needs to avoid staleness 23:58:29 raginpirate has joined #semantics-for-agentic 23:58:40 dmurph has joined #semantics-for-agentic 23:58:45 ... Having a single location for the information helps avoid staleness 23:58:47 q+ to mention the role of AI in maintenance 23:58:47 q? 23:58:55 q+ 23:59:22 Penny: I also want us to consider that browsers/agents will have to treat webpages as at best neutral if not hostile entities, so encourage folks to think about how this could be used to attack users via their agents 23:59:59 q+ to mention TAG Finding, Self-Describing Web: https://www.w3.org/2001/tag/doc/selfDescribingDocuments which discusses use of RDFa for self-describing HTML 00:00:10 ... Its a bit like having a trusted employee that gets tricked to be malicious, I would be more likely to trust that employee and be vulnerable, versus my own relationship with a malicious website 00:00:55 noamr: Two things. I am struggling to find the distinction of where human ends and machine starts. Is an agent that helps me buy a car a different machine vs something that helps me read a website. ARIA is still a machine, its just not agentic. 00:00:58 q+ to note sometimes the machines make the same mistakes humans do 00:00:59 ack noamr 00:01:25 Yoav has joined #semantics-for-agentic 00:01:25 noamr: I am also missing examples of where ARIA is harmful. I'm sure there are, but I'm not aware 00:01:30 q+ 00:01:36 dom: Not an ARIA expert, but one rule I've heard is don't use ARIA unless you have to 00:02:02 ... I think shifting the conversation towards the accessibility tree (rather than ARIA) is important 00:02:15 ... A similar proposal to RDFa is there's a new community group for ??? 00:02:22 scheib has joined #semantics-for-agentic 00:02:28 ... Looking at web semantics for agents 00:02:38 dom: Also, are agents a new class of users for webpages 00:02:43 q- dmurph 00:02:54 ... Some people are now using ?? to refer to the experience of a website for agent 'users' 00:03:05 ack dom 00:03:05 dom, you wanted to mention the first rule of ARIA, NLWeb and to mention "AX" (agentic experience) and to mention the role of AI in maintenance 00:03:27 hyojin has joined #semantics-for-agentic 00:03:28 Penny: I'm starting to see developer agencies marketing "is your website AX ready", and this seems terrible 00:03:32 s/terrible/concerning 00:03:35 Dingwei has joined #semantics-for-agentic 00:03:50 Mark_Foltz has joined #semantics-for-agentic 00:03:50 dom: Yep, risk of two class, where rich people have their agents doing things for them, and the rest deal with a terrible web 00:04:13 ... That reminds me, AI can also be used to help maintain things like the two ways of dealing with communicating info 00:04:22 tomayac: First rule of ARIA is don't use ARIA 00:04:28 ... Before we reach to that, the solution is actually semantic web 00:04:39 ... So many things are just form operations 00:04:41 +1 tomayac 00:04:52 +1 00:05:01 ... If your website is not recognizable to a machine, then its because you're building div based UI and then its not machine readable 00:05:06 Ugur has joined #semantics-for-agentic 00:05:21 ... I am concerned that people will start using ARIA wrong, making the web worse for human users 00:05:29 ... So we should be careful about telling users to use ARIA 00:05:34 q? 00:05:36 ack tomayac 00:06:12 johannhof: I want to take us one level up. Who are we serving, what's the outcome that we want? 00:06:23 ... To me its that websites should be able to be optimized for browser agents 00:06:24 q+ 00:06:29 q+ to speak to browser automation via WebExtensions 00:06:52 ... This takes us back to the problem of interoperability, but now its no longer between browser engines but between different models/system prompts/etc! 00:07:10 johannhof: I want to unite us on this common goal, that websites should be interoperable 00:07:14 ... Maybe we need to make it easier to test and optimize generally 00:07:16 q? 00:07:22 q+ 00:07:22 ack johannhof 00:07:45 ack jamesn 00:07:45 jamesn, you wanted to discuss ARIA and the accessibility tree - and how we can't rely on aria labelling and roles unless the tree is created 00:07:59 jamesn: I wanted to talk about technical feasibility of using ARIA, if we were going to do that 00:08:12 ... For example, the label for ARIA, you can't get it until you spin up the accessibility tree 00:08:16 q- 00:08:26 ... Thats extra compute cost, are browsers actually going to do that for all agent users 00:08:32 ???: Is it cheaper to screenshot? 00:08:50 s/???/tomayac/ 00:08:52 jamesn: Good question. But they'll be screenshotting anyway, they will be doing multiple thibngs 00:09:13 johannhof: Agree, there's this concept of no point in building complex systems too early, when it gets bypassed shortly just by processing the raw data 00:09:19 ... because the systems get so good 00:09:24 q++ 00:09:30 jamesn: The accessibility tree is also fragile and causes crashes. Would be great if it didn't 00:09:35 ... but that's the reality 00:09:43 q-+ 00:09:47 q+ tomayac 00:10:14 Penny: On the topic of accessibility tree vs not. There's some speculation of using a cheap on device model to lower the cost, and then select between models if the problem isn't solved by the cheap one. 00:10:32 ... I do think it should be as inexpensive as possible to be able to get the answer quickly, yes. 00:10:34 q? 00:10:39 q+ 00:10:53 ugur has joined #semantics-for-agentic 00:11:04 ErikAnderson has joined #semantics-for-agentic 00:11:13 q+ 00:11:17 raginpirate: Wanted to return to question #1 (do we accept the premise). I do agree that we cannot do 100% by inference alone, but want to extend the question 00:11:25 ... Will using 'hints' like ARIA even reach 100%? 00:11:50 q? 00:12:03 front-endian-jane has joined #semantics-for-agentic 00:12:06 ... I am skeptical that the fundamental problem might not be feasible 00:12:29 ... Want to challenge is this the right path, or is the right path that we need agents to NOT operate on the normal web, if we want to achieve actual 100% 00:12:41 Penny: As in, parallel MCP service or similar? 00:12:47 I wanted to comment on no content duplication and enabling agents to "follow their nose". Some of this is not at all distinct from good / useful information for humans. The Self-Describing Web document which is a TAG Finding talks about this in depth and again RDFa is mentioned there for use in markup languages like HTML. 00:12:47 There is a distinction to be made about semantic HTML and reuse of independently developed vocabularies to express some information. Take schemaorg as an example or countless other vocabularies on the web. They are able to describe different kinds of knowledge on the web as well as operations for applications. 00:12:47 hadleybeeman has joined #semantics-for-agentic 00:12:47 raginpirate: Exactly 00:12:49 ack raginpirate 00:12:50 ack csarven 00:12:51 csarven, you wanted to mention TAG Finding, Self-Describing Web: https://www.w3.org/2001/tag/doc/selfDescribingDocuments which discusses use of RDFa for self-describing HTML 00:12:52 present+ 00:13:01 present+ 00:13:07 csarven: Wanted to talk about content duplication that you mentioned earlier. 00:13:15 ... A lot of thats not distinct from useful information for humans 00:13:25 ... Can have a single source of truth thats not hidden information 00:13:32 ... Thats accessible, visible, and available to machines 00:13:36 https://www.w3.org/2001/tag/doc/selfDescribingDocuments 00:13:38 csarven: There's a TAG finding about self describing web 00:14:39 ... Really talks to the question of what is the purpose, the relationship between pieces of information 00:14:43 ... Concrete example, a citation 00:14:57 ... This sentence/argument/thought maybe refutes or augments this other argument somewhere else on the web 00:15:10 ... Actually real linkage between those things 00:15:36 csarven: Second point, there's a distinction to be made about semantic HTML 00:16:12 ... Worth noting that there are other vocabulary than HTML 00:16:14 q? 00:16:35 csarven: we need to take into account different ways of expressing knowledge by different communities 00:16:41 ... No central authority says what a trashcan should be 00:16:48 ... We have bodies that can talk about icons 00:16:54 ... We have bodies that can talk about genes or health data 00:17:21 csarven: Are we describing all of human knowledge at W3C? 00:17:33 ... think about why schema.org exists in the first place 00:17:37 q? 00:17:56 Zakim, please close the queue 00:17:56 ok, smcgruer_[EST], the speaker queue is closed 00:18:07 q+ 00:18:41 In response to the COGA group advocating using ARIA for adding meaning sematically, I did a very limited exploration of a way of using a controlled vocabulary to prompt users for unconventional uses of inputs and buttons. Based on the button example, it seems somewhat valid. 00:18:41 https://drive.google.com/file/d/1Nkpl3GFlGOuc_PdzxH6GJH1_zc6ZOdAH/view?usp=sharing 00:18:48 ack wendyreid 00:18:48 wendyreid, you wanted to note sometimes the machines make the same mistakes humans do 00:19:08 wendyreid: I ran into a real use case for this. Developer tool for developers to recreate accessibility bugs before they try to fix them 00:19:18 ... QA writes really good test steps, so its useful to give to an AI to do stuff 00:19:40 ... But when I asked it to write tests, the AI failed because the test says something like 'click on the buy now button' and the AI failed because it was a link not a button 00:19:55 ... Humans say things like "that looks like a button" vs the underlying technological semantics 00:20:02 q? 00:20:11 +1 to Wendy on the lack of correlation between semantic role and visual presentation 00:20:16 ack Yoav 00:20:42 Yoav: Responding to noamr , we want our pages to talk to machines yes, but there's a significant difference between deterministic and non-deterministic machines 00:20:54 ... Also very unclear how developers are meant to test any of this (+1 to johannhof ) 00:21:04 ... Beyond just the massive combinatorial problem, there's non-determinism 00:21:19 Yoav: I think one reason developers don't test ARIA is that its hard and not enough market share 00:21:30 ... Agents will probably pass the market test side, but the testing is maybe even harder 00:21:41 ... Maybe imperative solution like WebMCP is better than markup 00:21:50 ... Does page markup really need to be machine readable 00:21:56 ... Better if agent calls function rather than click a button 00:22:13 q? 00:22:18 Penny: So is that giving up on functional declarative experiences that don't have JavaScript 00:22:22 Yoav: Maybe, good point 00:22:33 johannhof: I think there's a balance 00:22:35 ack dom 00:22:37 ack dotproto 00:22:37 dotproto, you wanted to speak to browser automation via WebExtensions 00:22:50 dotproto: Browser automation has been a common use-case for extensions for a long time 00:22:59 ... Unsurprisingly we have seen people push into agentic space 00:23:08 ... Had challenges in better serving automation use case 00:23:18 ... Hesitant about exposing all data to a given extension 00:23:18 dom yes please 00:23:28 ... Things like "all URL" permission is terrifying, but is often used 00:23:54 The queue has been closed, but I wanted to add that It's interesting to note that we (humans) also learn from our mistakes. So it would be interesting to think about similar mechanisms for agents. For example, If the agent deletes a cloud file by mistake, how can it identify that this happened and recognize that it should restore the file from the 00:23:54 junk? 00:24:11 dotproto: I've heard from extension developers that there is a clear need to have a better way to interact with and identify with what content is available for interaction in a given webpage without requiring site author to do something 00:24:13 q? 00:24:42 ... looked at accessibility tree for this historically, but all the concerns we've heard today applied then 00:24:44 q? 00:24:52 ack kush 00:25:04 q- 00:25:06 mbgower has joined #semantics-for-agentic 00:25:14 present+ 00:25:17 kush: This issue came up in WebMCP discussion. 00:25:24 ... We are talking about having a declarative version of WebMCP 00:25:33 ... We don't want to have a world where we're forcing developers to have script 00:25:45 ... ARIA folks are reviewing the proposal 00:25:57 kush: We likely need new attributes to let sites opt into which parts of their page are relevant for agents 00:26:15