12:06:45 <RRSAgent> RRSAgent has joined #ai-gen-stds
12:06:49 <RRSAgent> logging to https://www.w3.org/2026/03/25-ai-gen-stds-irc
12:06:49 <breakout-bot> RRSAgent, do not leave
12:06:50 <breakout-bot> RRSAgent, this meeting spans midnight
12:06:50 <breakout-bot> RRSAgent, make logs public
12:06:51 <breakout-bot> Meeting: AI-generated software and Web standardization
12:06:51 <breakout-bot> Chair: Dominique Hazaël-Massieux
12:06:51 <breakout-bot> Agenda: https://github.com/w3c/breakouts-day-2026/issues/8
12:06:51 <Zakim> Zakim has joined #ai-gen-stds
12:06:52 <breakout-bot> Zakim, clear agenda
12:06:52 <Zakim> agenda cleared
12:06:52 <breakout-bot> Zakim, agenda+ Pick a scribe
12:06:53 <Zakim> agendum 1 added
12:06:53 <breakout-bot> Zakim, agenda+ Reminders: code of conduct, health policies, recorded session policy
12:06:53 <Zakim> agendum 2 added
12:06:53 <breakout-bot> Zakim, agenda+ Goal of this session
12:06:54 <Zakim> agendum 3 added
12:06:54 <breakout-bot> Zakim, agenda+ Discussion
12:06:55 <Zakim> agendum 4 added
12:06:55 <breakout-bot> Zakim, agenda+ Next steps / where discussion continues
12:06:55 <Zakim> agendum 5 added
12:06:55 <breakout-bot> Zakim, agenda+ Adjourn / Use IRC command: Zakim, end meeting
12:06:57 <Zakim> agendum 6 added
12:06:57 <breakout-bot> breakout-bot has left #ai-gen-stds
12:17:06 <dom> dom has joined #ai-gen-stds
12:17:30 <dom> Slideset: https://www.w3.org/2026/Talks/dhm-ai-software/
13:48:23 <hyojin> hyojin has joined #ai-gen-stds
13:56:10 <dom> Present+ Dom, MikeGifford, PaolaDm
13:56:32 <dom> Agenda: https://www.w3.org/events/meetings/d3eeea4f-e4dc-470d-9d4a-6a33a8fdcf8f/#agenda
13:56:37 <dom> Chair: Dom
14:00:51 <astearns> astearns has joined #ai-gen-stds
14:01:24 <fscholz> fscholz has joined #ai-gen-stds
14:01:43 <Roy_Ruoxi> Roy_Ruoxi has joined #ai-gen-stds
14:01:56 <xueyuan> xueyuan has joined #ai-gen-stds
14:02:06 <mgifford2> mgifford2 has joined #ai-gen-stds
14:02:16 <mgifford2> present+
14:02:21 <pchampin> pchampin has joined #ai-gen-stds
14:02:29 <anssik> anssik has joined #ai-gen-stds
14:02:32 <xiaoqian> xiaoqian has joined #ai-gen-stds
14:02:42 <xiaoqian> present+
14:02:43 <Ian> Ian has joined #ai-gen-stds
14:03:06 <Ota> Ota has joined #ai-gen-stds
14:04:19 <denis> denis has joined #ai-gen-stds
14:05:14 <TiffanyBurtin> TiffanyBurtin has joined #ai-gen-stds
14:05:49 <anssik> Present+ Anssi_Kostiainen
14:05:50 <TiffanyBurtin> present+
14:05:58 <fscholz> present+
14:06:26 <mgifford2> Copilot is also integrated into the GitHub environment with https://github.com/codespaces and https://docs.github.com/en/copilot/get-started/quickstart
14:08:18 <mgifford2> Even just GitHub's Accessibility team Accessibility Scanner - https://github.com/github/accessibility-scanner
14:12:43 <anssik> -> One Human + One Agent = One Browser From Scratch https://github.com/embedding-shapes/one-agent-one-browser
14:13:26 <Curve> Curve has joined #ai-gen-stds
14:13:51 <mgifford2> Specs are the new code https://www.linkedin.com/pulse/specs-new-code-tyler-dane-lbz1c/  &.  https://medium.com/@aiforhuman/specs-are-the-new-code-2f3ede491324
14:14:06 <anssik> re One Browser that's another example, "Over three days they drove a single Codex CLI agent to build 20,000 lines of Rust that successfully renders HTML+CSS with no Rust crate dependencies at all"
14:14:16 <anssik> -> https://simonwillison.net/2026/Jan/27/one-human-one-agent-one-browser/
14:14:47 <noah> noah has joined #ai-gen-stds
14:17:28 <mgifford2> https://conesible.de/wab/whitepaper_webaccessbench.pdf note that I don't think there was any advanced prompting with this study to ensure that there is anything more instructive about standards.
14:18:25 <fscholz> fscholz has joined #ai-gen-stds
14:19:59 <Curve> Curve has joined #ai-gen-stds
14:21:09 <pchampin> present+
14:22:50 <Shu> Shu has joined #ai-gen-stds
14:24:19 <martinA> martinA has joined #ai-gen-stds
14:24:21 <dom> q?
14:24:24 <martinA> present+
14:26:39 <dom> Anssi: we had an AI generated implementation of WebNN based on spec, test suites and code samples
14:26:45 <mgifford2> There was also the idea of including RDFa into all of our specifications so that all of our existing specifications can be more semantically linked
14:26:47 <dom> ... provided lots of new useful insights
14:26:55 <dom> scribe+
14:27:27 <mgifford2> q
14:27:33 <mgifford2> q+
14:28:01 <mgifford2> https://mgifford.github.io/ACCESSIBILITY.md
14:28:14 <dom> anssik: the Chrome implemetantions took 3 years from 20+ engineers; the rust implementation was done in a few weeks with one engineer and 1 agent
14:28:24 <TiffanyBurtin> May we have the link to the presentation?
14:28:32 <dom> anssik: spec-driven development is becomign the main way of driving AI assistants
14:29:33 <dom> anssik: in behavior-driven development, you describe expected behaviors in natural language descriptions as specs
14:29:46 <dom> ... humans write the specs as primary artefacts
14:29:54 <Curve> Curve has joined #ai-gen-stds
14:29:59 <dom> ... agents then implement them; humans verify they meet expectations
14:30:20 <dom> ... humans define what "correct" means, with business cases and other considerations (e.g. privacy, etc)
14:30:25 <mgifford2> I've been using AI to build tools like this for accessibility like https://mgifford.github.io/open-scans/reports.html
14:30:40 <dom> qck mgifford2
14:30:42 <dom> ack mgifford2
14:31:08 <dom> Mike: the changes of the past 3 months about what's possible is striking
14:31:55 <paola> paola has joined #ai-gen-stds
14:32:08 <Curve> Curve has joined #ai-gen-stds
14:32:24 <dom> ... gave pointers above to example of making AI-practices replicable en masse
14:32:42 <paola> such an interesting topic, and YES  DOM  we must lern how to work with AI/agents
14:33:17 <dom> anssi: how does the fact that the cost of developing software approaches zero impact our community?
14:34:01 <dom> ... writing a browser engine from scratch didn't seem possible any more; now it may become achievable again
14:34:03 <dom> q+
14:34:42 <dom> ... can this completely change the pace of iteration and power balance on spec-ing/implementing?
14:35:18 <paola> The cost.effort will be allocated differently, but human expertise remains expensive to gather. and remains necessary
14:35:53 <mgifford2> Sorry, wifi turned off.
14:35:57 <mgifford2> q+
14:36:13 <mgifford2> Also how do we do it responsibly? Looking at the W3C's ethics statement and the web sustainability guidelines. I've also created https://mgifford.github.io/SUSTAINABILITY.md to try to help.
14:36:18 <paola> yes of course, everything is completely changing, but humans will have to continue to train, and human expertise costs a lot of effort
14:36:57 <tidoust> tidoust has joined #ai-gen-stds
14:37:18 <paola> so when we go back to the code or documentation next morning we still have a lot of very high skilled work to do
14:37:46 <paola> evaluation, integration, training humans on what AI is doing and how, is going to be expensive
14:37:58 <dom> dom: story of the continued value of interoperability in a world of marginal cost of software development
14:38:09 <dom> anssi: broad adoption still needs wide consensus from humans
14:39:16 <dom> francois: specs are not written today as describing the outcome of code, but now they're almost code - with algorithmic steps
14:39:29 <dom> ... to ensure interop
14:39:47 <dom> ... these algorithms are actually hard to review, tale a lot of WG participants time
14:40:08 <dom> ... is this an opportunity to get back to a less algorithmic approach to spec writing?
14:40:16 <dom> ... and the tools may deal with detect interop issue
14:40:25 <dom> ack mgifford2
14:41:15 <paola> the role of humans is going to change, I think, from programmers to systems engineers of very complex and sophisticated tools, reviewing deploying and using the code is what humans are going to require a lot of effort.  there is a shift.  but AI can and must neveer operate without the humans in the loop
14:42:11 <dom> mikeG: to use AI responsibly (incl. from sustainability perspective), we need to find ways to share the infrastructure responsibly
14:42:49 <dom> ... I'm hoping we can feed web sustainability guidelines in how these AI tools build apps & pages - e.g. making it easier to build dark mode
14:43:40 <mgifford2> the keynote I mentioned https://events.drupal.org/chicago2026/session/keynote-alexandra-bell-security-implications-ai
14:44:31 <paola> so for example, we need to continue to develop tests and make sure the resutls of the testing and evaluations are valid.,  testing can also be automated, but only up to a point.  @dom  let me evaluate if the existing accessibility guidelines written for humans are OK also for machines,  thats an important question- I ll look into it
14:45:16 <dom> dom: does AI tool change the way we relate to complexity of Web technologies due to their long legacy
14:45:18 <mgifford2> q
14:45:19 <mgifford2> https://almanac.httparchive.org/en/2025/accessibility
14:45:28 <dom> ack mgifford2
14:45:32 <dom> ack dom
14:45:33 <dom> ack mgifford
14:45:37 <paola> thank you @mgifford2 and @dom I ll make a note of these points. look forward to the transcript
14:46:22 <dom> mgifford2: comparing to self-driving cars - leaving control to a machine creates anxiety initially
14:46:46 <dom> ... Web almanach is a good source of data about how well our standards are adopted - incl in terms of quality
14:47:22 <dom> ... most accessible TLDs are .gov, .edu, and ... .ai - which can be assumed to have a lot of AI-generated content
14:47:47 <dom> ... it's only matter of time before AI agents create better code following specs than humans
14:47:56 <Roy_Ruoxi> q+
14:48:07 <dom> ... we need to get better measures on this
14:48:44 <dom> Roy: I tried to use AI agents to generate Web sites & apps - the paradigm of creating software is shifting
14:49:15 <dom> ... there is another layer between developers and their apps, the agent who will read the spec
14:49:51 <dom> ... maybe our specs need to consider how to make AI agents create better code, e.g. via skills
14:50:11 <dom> ... @@@ engineering
14:51:40 <paola> Roy_Ruoxi in my experience, we canot generalize, because every coding agent is really different and good at different things,  as a human engineers we must learn how to write the code spec so that the agent can just follow them  basically, the human must write the prompt so that the AI coding agent knows exactly what to do.  humans must learn how
14:51:40 <paola> to write the system specification so that the coding agent undertand how to execute them
14:52:18 <dom>  Anssi: we already see that natural language is a universal programming language
14:52:26 <Roy_Ruoxi> s/@@@/harness
14:52:49 <dom> ... we can define our specs in English, or in algorithmic prose, or likely both
14:53:05 <iris> iris has joined #ai-gen-stds
14:53:21 <dom> ... with agents, we could start in natural language (e.g. explainers), and ask an agent to propose more algorithmic versions for review
14:54:43 <paola> the human must develop the skill to build precise briefs
14:55:08 <dom> ... or the reverse - although natural->algo feels more natural
14:55:17 <mgifford2> q+
14:55:35 <tidoust> ack Roy_Ruoxi
14:55:36 <Roy_Ruoxi> ack me
14:55:43 <tidoust> ack mgifford
14:55:47 <dom> mgifford: there is a lot of anxiety about AI - incl in the sustainability IG, the accessibility community
14:56:19 <dom> ... it's also unsettling  open source communities
14:56:40 <dom> ... it impacts people's jobs and livelihoods
14:57:37 <dom> ... we need to be careful in how we approach these questions, both avoiding over-enthusiasm and over-avoidance
14:58:15 <RRSAgent> I have made the request to generate https://www.w3.org/2026/03/25-ai-gen-stds-minutes.html Ian
14:58:22 <Ian> rrsagent, set logs public
14:58:27 <Ian> present+
14:58:28 <RRSAgent> I have made the request to generate https://www.w3.org/2026/03/25-ai-gen-stds-minutes.html Ian
14:58:34 <paola> thank you everyone,
14:58:37 <mgifford2> https://www.youtube.com/live/EU6n7D91m_M
14:58:57 <mgifford2> That is this year's DriesNote from the Drupal community
16:09:57 <tidoust> RRSAgent, bye
16:09:57 <RRSAgent> I see no action items