Skill issue: Lessons from skilling up coding agents to use Langfuse

Talk transcript

1. 0:00

   [upbeat music] Okay, that was quick.

2. 0:16

   Hi, everyone. Super excited to be here. I'm Marc, one of the founders of Langfuse. Uh, when we started Langfuse, like, three years ago when everything was-- felt quite early, uh, building agents that didn't work, uh, and then realized, okay, there needs to be, like, some evaluation tracing built.

3. 0:30

   Like, uh, Langfuse is the open source project in the space by now. I mean, by the metrics that we track, uh, we seem to be the largest one in the space.

4. 0:36

   We do all product engineering out of Europe. Thus, I'm very excited that this conference is coming to Europe because there are so many great people here. Uh, and we always need to resist the urge to, like, ship the whole team to another continent because actually being here is very nice, um, and you can just travel and hang

5. 0:49

   with people on Discord and Zoom. So yeah, very excited to be here in person. And, um, what I wanna talk about today is how, um, well, the lessons we made from scaling up coding agents and actually adding Langfuse to an application because, like, back in the days when you added, like, observability or evals, you needed to read,

6. 1:04

   like, hundreds of pages of docs, like, figure out your own mental model, and now you kind of expect, like, to be hand-holded by a custom agent to do this.

7. 1:11

   And, uh, we've come a long way, uh, in achieving this kind of vision, and I just wanna explain what we learned on the way. Um, first of all, like, I'll start very conceptual, very easy, then a bit more conceptual deeper, and then I'll do-- go to the learnings.

8. 1:22

   So, uh, my mental model for skills is just like, I get this Rubik's Cube when I was a kid. I d- had no idea what to actually do, but I mean, I basically have a bash tool.

9. 1:29

   I can do whatever I want with this Rubik's Cube, but it just looks colorful in different ways and no idea how to solve it. So skills was, like, a great way.

10. 1:35

    Uh, once you get the manual, it's easy. You just need to follow the manual and you can, you can solve a Rubik's Cube. And I feel like the same thing now applies to, to agents, where, uh, like there was this whole debate of, like, workflow versus, uh, fully autonomous agent, and was, like, huge fight on X of,

11. 1:49

    like, what is the best way to build an application? Everyone was kind of like, "Yes, you kind of like need both." And I think Marta this morning had, like, a good note where, like, the surface area of deploying agents is so broad that, um, like for some, you don't need, like, a coding agent, even if it's, like,

12. 2:02

    the best way you can build agents at the frontier today. Like, you don't need this for every application. It makes things slow, expensive. And, um, like, there's this kind of, like, balance between workflow being very reliable and agent now having kind of, like, unlimited capabilities.

13. 2:14

    And I think what's very exciting is that then skills are kind of like a formalized shortcut to make things more reliable, where you historically would've built a workflow. Like, I don't know, you have a customer res-- uh, like, customer support agent, and someone asks for, like, a password reset, then historically you would've built, like, a workflow that's

14. 2:29

    very reliable of, like, ah, you have a router that routes to, like, an agent that can only do password resets, and that agent has, like, the context to do password resets well.

15. 2:36

    Um, that was great, but also if the user then wants to do password reset but also change the email address at the same time, then, like, the router is kind of like, "Okay, I have this email router, I have this password reset router.

16. 2:46

    What, what do I even do?" And I think what's very exciting now is that an agent can just, like, progressively get the context needed to then solve, like, a problem that's multi-domain.

17. 2:54

    Like, that would've historically been in multiple workflows. So that's very exciting. However, uh, it has always been hard to kind of like how to, how to build an agent, what kind of use cases even exist because you have, like, this open-ended text input box often or, like, open world context, so you didn't really know.

18. 3:09

    So what we now see is what many teams do is they have the agent runtime, they trace everything. I mean, I'm building Langfuse. Thus, I put my logo here.

19. 3:16

    But you can also use whatever you want. In the end, it's more about, like, the concept of tracing is what you need to, uh, like, identify what happens at runtime when a trace-- when, like, an agent's executed.

20. 3:25

    Because then it helps you learn two things. One, uh, like, new kind of, like, use cases that you didn't expect, uh, users to do because you might have expected that nobody ever wants to change a password, but now they want to change a password.

21. 3:36

    Thus, you need to see the execution trace of someone being upset via, like, a production eval to then derive that you need to add a skill, um, for handling password resets well.

22. 3:44

    Uh, and two, once you have these, uh, skills, they can get out of date, or you can realize they're not the most efficient way of actually solving for this use case.

23. 3:51

    So it's, like, the second thing of, like, helping you improve skills that you already have in your agent. So this is very conceptual, but this is what we, what we now see, see most teams do, uh, that use Langfuse.

24. 4:01

    Now more towards, like, the learnings, uh, we made, uh, when building a skill to help, uh, customers add Langfuse to their project. So what was the lay of the land before we got started with this?

25. 4:11

    Um, like, four hundred and seventy-eight pages of documentation. Whenever I see, like, the thing deploying, I'm like, "Who wrote all of this?" So apparently, if you build a project over three years, it just grows in complexity.

26. 4:21

    People can do all of these different things, but then you need to read all of these things, and nobody has the time. Across, like, five different f- uh, feature areas and a lot of implementation flexibility because there have always been, like, projects in, like, the eval space that were very, I'd say, opinionated.

27. 4:35

    So they were like, "Oh, you have a chatbot? Then add this project, and it'll just solve this for you opinionated end to end." We always were like, "No, no, we are infrastructure.

28. 4:43

    We do, like, tracing well. Like, if you ingest, like, billions of traces, we'll still work. If you wanna customize your evals, it'll still work." So we're always, like, more on the unopinionated side, uh, which was always, I'd say, uh, a weakness compared to, uh, projects that are more opinionated, but now I think it's a strength because in

29. 4:57

    the end, what, what, what, what do you need, uh, when agents do all of this? You only need the infrastructure piece if agents can then customize for different workflows.

30. 5:03

    Um, so but there's the problem if people wanna add Langfuse to a project. How to do it correctly for a project is to be figured out by an agent.

31. 5:12

    Um, and, uh, interestingly, uh, when we first got into model pre-training context as a project, and you can just ask an agent how to add Langfuse, and it'll spit out, like, Langfuse SDK logic.

32. 5:22

    The first time this happens, it's amazing. But then if you're, like, two years in, like, the project evolves, like, interfaces change. Uh, and now being in pre-training context might even be, like, a disadvantage if, uh, you don't fetch up-to-date information.

33. 5:34

    So we're just like, "Oh, we get, like, all of these hallucinations of methods that have been, uh, available in the past, but they are not, uh, available there today."

34. 5:41

    So yeah, we, we felt like when skills launched, this is, like, the exact pattern that we need in order to help teams achieve this. Um, so I- I'll use an example.

35. 5:49

    Um, when you just ask Claude Code to add, uh, Langfuse to a project, it just works, but it was not working in the best way possible. So for example, user asks, "Add tracing to my agent," and, um, then Claude Code kind of, like, implements the instrumentation based on the outdated pre-training context, then tries to verify whether the

36. 6:08

    tracing works, then realize, oh, it doesn't work, and then only in the second step fetches up-to-date information to then correct the issue. Um, at the same time, how you add tracing or evaluation to a project, they can evaluate in, like, like, gazillion different ways, like online evals, offline evals, human-in-the-loop.

37. 6:24

    There are so many different things, and often the question is like, what even is relevant for your application? Uh, so human agent kind of need to figure it out on the way, but agent is not tasked to help you, uh, like, figure out what's the best thing for your, for your, um, application.

38. 6:37

    So main problems, outdated training data, the non-optimal setup because the agent wasn't really, like, primed to help you discover what to do for your, for your app, and it's very slow because you first add instrumentation of wrong way, then you figure out it's wrong, and then you need to fetch more documentation to, to fix the, the issues.

39. 6:53

    So what did we do? Uh, yeah, this is how the, uh, this is how, uh, just mental, mental note. This is how the trace looked like when we just tried it with Claude Code.

40. 7:00

    So like just tracks two LLM calls in an agent, but like you still don't know what the agent's actually doing. So, um, what was the goal of our skill?

41. 7:07

    Uh, like, give every, uh, Langfuse user, like, they have thousands of teams in the community, thousands of customers on our cloud product, give them all, like, a Langfuse expert to help them quickly set up observability, prom management, evals in line with, like, best practices and up-to-date docs and rep- uh, like, references.

42. 7:21

    Because, uh, all of you, I mean, uh, there's Anabel from the team here as well. Like, if you have questions regarding observability evals, you can talk to us, but in the end, that doesn't scale to thousands of people, uh, to basically talk through your problems and figure out what the best strategy is for you.

43. 7:33

    So we are like, "Okay, skills is the, is the way to go." And this is, like, very conceptual how our skill works, where, um, like user comes in, asks coding engine to do something, and then the skill kind of has like a reference of like ...

44. 7:45

    The skill MD is more like, well, what kind of style do we want, uh, in order to implement, um, Langfuse? So for example, ask follow-up questions before making a decision because there's so much you can be doing, and then references for the different product modules to kind of like progressively disclose additional hints that the agent need, might

45. 8:00

    need to have. And then, uh, it can call, um, the documentation. And interesting enough, as we start open source and tr- like, saw ourselves as, like, independent infrastructure, we always had like APIs for everything because teams built their own, I don't know, own labeling UIs on top of it, own evaluation, uh, execution logic on top of our,

46. 8:16

    our, our backend. And we'd had APIs for everything. Now we've wrapped it in a CLI, and now an agent can just do everything humans needed to do in the UI in the past, um, which is very cool, uh, because, uh, like so many teams spend so many hours every week in clicking around in our UI to evaluate

47. 8:30

    and improve the application. And in the end, how will this look like end of year? It'll probably just be like, uh, like connect repository to Langfuse, and then agent just does the whole thing, uh, like autoregressively.

48. 8:39

    I mean, that's what we are building towards. That's what everyone is building towards, and I think that's a cool step in the right direction. So, um, to, to shortcut to the end result, okay, after, like, conversing with the agent now for the, uh, for a similar thing, it looks, uh, way more detailed, like, um, like detailed evals

49. 8:55

    that are relevant and detailed steps, uh, regarding to execution. Um, so, so just like there, there's a stack difference. And yeah, what did we learn on the way? Um, six main things.

50. 9:04

    Uh, I'll go through every single one of them, um, of, uh, what were basically our realizations, uh, when we're building the skill. Um, and, uh, like one, looking at traces still, uh, like gets you to like 80% of the detail.

51. 9:15

    This was always what we kind of like tried to preach regarding like evals, where many people try to complicate things right away while they haven't digged through, um, through like, just like what did the agent actually do at runtime themselves a couple of times.

52. 9:26

    So, so, uh, what do we do? We, uh, have instrumentation for Claude Code and just ourselves interactively, um, like tried to use Langfuse with Claude Code and then looked through, uh, traces in Langfuse to, uh, understand where did the, um, agent error, how can we improve the skill to make it, like, uh, straight, uh, shooting at the

53. 9:45

    goal instead of like wandering in, in different ways to the, to the target. And that was really helpful, too. Um, there were like some of the, uh, interesting learnings here.

54. 9:53

    Um, for example, for humans, we tried to cut down on the number of environment variables that you need to set in order to set up Langfuse. So for example, we just auto assumed, um, a data region.

55. 10:01

    Like Langfuse is available in Europe, it's available in the US. Uh, fun anecdote, we assumed that only Europeans, uh, like care about data regions. Thus, we made Europe the default.

56. 10:09

    Then we learned some US enterprises also care about data regionality. Now we have a US data region, so many different other data regions. Um, so we always defaulted to Europe, and now we kind of like for an agent, like adding another environment variable, like they don't care.

57. 10:20

    Like it's not effort for them. Thus, uh, we always prompt for figure out what data region the user's actually in and don't assume Europe, for example. Uh, two, hallucinated hal- uh, CLI parameters, uh, because it just, oh, uh, this like includes the word trace.

58. 10:34

    I've seen tracing CLIs before. I just assume what, what we could be doing here, and we just advertise the, the help flag more aggressively. Uh, it takes another turn, but it's fast, and thereby, uh, like it directly knows, uh, what the CLI can do.

59. 10:46

    Um, two, um, we try to help the agent to understand how to navigate available information because, like, I mean, five front documentation pages, how to find the right one instead of like looping through, I don't know, fetching one, then learning something, then fetching another one, always with like thought process in the meantime.

60. 11:00

    Um, so what did we do? We always had this llms.txt, which was very hyped when it launched but never actually used. I think what's now cool is we have this agent sitemap kind of that we just expose to, um, like a coding agent via the skill of like, go there first in order to learn what kind of

61. 11:14

    like documentation is available. And, uh, two, I mean, there's like this whole content negotiation that if you send like an, a request header that you want Markdown, that you get Markdown back from the docs, but some coding agents don't do this by default.

62. 11:26

    So we just advertise this because otherwise some coding agents might try to pass the HTML, which just adds additional tokens. So, uh, like for Langfuse for example, you can just add like a .md to any documentation page, or you can request Markdown, and you'll get a Markdown page.

63. 11:38

    Uh, three, um, I think that was, uh, like one of the things I was most excited about. Uh, we always had this, uh, like docs Q&A agent that was able to answer questions about Langfuse more like interactively.

64. 11:48

    Therefore, we built like a RAG stack, and now we just surface this, uh, RAG stack again via like a search endpoint. So like a coding agent can just ask whatever natural language query about Langfuse, and we'll get back documentation chunks for this query.

65. 12:02

    Why is this exciting? One, you don't need to fetch five different docs page where you can just ask a question, get something back that's relevant directly, uh, solve for the problem.

66. 12:09

    And two, we get to track these search parameters, um, because, uh, like if a coding agent fetches documentation, it's very difficult to understand what did like Claude Code and our user laptop do.

67. 12:18

    But if they, um, ask questions about Langfuse to our search endpoint, we can track the searches and thereby understand what problems do they run into, where do we need to add more documentation pages, because maybe we didn't expect this kind of problem to happen.

68. 12:29

    So yeah, so adding a search endpoint was like, was really cool to, um, capture more data. Um, then, uh, basic eval setup is better than none because we initially struggled to get this done because it's like so broad.

69. 12:41

    Some Langfuse users build like chat applications, real-time voice, video generation, like batch pro- Processing of like invoices in the background of some kind of like text, uh, software. Like so many different use cases where then the question is, what's even like a good evaluation, uh, like setup?

70. 12:55

    And we just created like five, uh, different ones, and this was already helpful, um, because otherwise, uh, it's really hard to, uh, to kind of like measure, measure anything.

71. 13:04

    And what we did here, um... I don't know, can I zoom in? No. Um, so basically we have this, uh, like just like a prompt instrument application with Langfuse and then like a sample, a repository folder.

72. 13:13

    So for example, like an OpenAI custom function RAG, whatever, uh, application. And our checks are just natural language statements that we then, by LLM-as-a-Judge try to, um, like evaluate on top of the file system and diff state before and after, um, running the skill.

73. 13:27

    Uh, so for example, we ex- uh, we expect that our OpenAI instrumentation was added because it's like an OpenAI example, and we, because it's RAG, we, um, expect like some retrieval spans to show up in our, uh, trace because if they, if there are no retrieval spans, then probably we only capture, for example, LLM, um, calls.

74. 13:43

    This was already helpful because that way we were able to make changes and see that we didn't break anything. The whole thing that, uh, why, why we even built Langfuse for like building AI agents now also applies here.

75. 13:52

    Five, uh, dynamic content should be referenced because there's a huge, I'd say, incentive for like developers on the team, but also for users in the community to just contribute a lot of context to the skill because then you're like, "Ah, it's kind of like a local cache of the documentation that's immediately available."

76. 14:06

    However, then the same thing applies, uh, that applies to pre-training context. It's kind of, uh, like it goes out of date, and now we have the documentation, and now we have yet another representation of what Langfuse is.

77. 14:17

    So we'd rather try to point just to straight to the reference of documentation and, um, because otherwise you just duplicate all, all content. And six, uh, we applied, um, of like auto research, uh, to, uh, the scale of, okay, if we have a target function, how can like agents help us improve the agent?

78. 14:33

    Um, because there are so many like different patterns that we can explore. Um, so we set up a target function mostly get towards, uh, like, uh, our experiment really here was help teams move prompts from their local Git repository into Langfuse prompt management, which is used by like larger teams to collaborate on prompts with their non-engineering counterparts

79. 14:48

    because then like PMs can make changes to prompts, iterate on a playground, like all of this kind of like collaborative stuff. Um, and the task was, okay, how do we improve the skill to migrate prompts out of any kind of like code base into our managed prompt system?

80. 15:00

    And, um, in the end, we accepted three out of the six, uh, improvements that were suggested, which I think is a success. Uh, uh, but, uh, it allowed us to experiment much more than we could have explored manually, uh, with the time that we have as we are like a very small team.

81. 15:13

    Uh, learnings, uh, like the target function really matters. Uh, like, uh, I think it's-- it sounds obvious, but, um, for us, defining like the right target function was very hard for this, um, because we, we assumed like a prompt migration should be fast.

82. 15:25

    Fast, we measured in like the number of turns, but if we, uh, basically asked to minimize the number of turns, then like our, uh, like the agent that tried to optimize the skill just took out all of the nodes that we had to, um, like fetch documentation because it was like, "I know how to-- how Langfuse prompt

83. 15:39

    management works. I don't need this. I'll just try it myself." Uh, which then negates the whole thing of we wanna fetch up-to-date context, because otherwise, if you use the skill-- install the skill once, wait three month, then you'll have like wrong context, um, because, uh, we duplicate information.

84. 15:53

    Two, um, like we had like an approval gate usually, uh, where we wanna suggest a plan or ask follow-up questions, suggest plan to a user before doing anything because we kind of like push their prompt to like a central repository, and it's kind of like their data leaving their laptop somewhere else.

85. 16:07

    Um, but the sandbox didn't have this, so we didn't really-- we weren't really able to, uh, to, to try for this. And, um, like Langfuse, like the, the sole feature, like usually we try to make it easy to get going with something, but then it's very deep of how to do it in a good way, and we

86. 16:19

    want agents to directly go for the good way. Like figure out with the user what they wanna achieve and then have like a very full implementation, not start with something and then like two month later go deeper.

87. 16:29

    However, if the target function does not include, uh, like we want like, uh, linking prompt versions to prod traces so then you can see how like different prompt versions impact like, for example, production results.

88. 16:39

    Um, like we didn't have this in the target function. This, uh, like e- everything that y- like nudged towards this was kind of like removed because it's kind of like, it's just like, like garbage on the way that we don't need to achieve the goal.

89. 16:49

    So again, the target function really matters. High level, these were like the, the six, uh, main takeaways. Looking at traces gets you eighty percent of the way. The, uh, production signals really help.

90. 16:58

    So the search endpoint, uh, was really helpful for our documentation. Um, help agent to navigate, uh, the information because otherwise it just searches with like Google, Brave, whatever search and finds all sorts of different things on the internet.

91. 17:11

    Um, even a basic s- eval setup helped. It wasn't that hard to set up. The dynamic content should be referenced, otherwise you have just duplicates, and the auto research was, uh, very helpful to explore things, but, um, it's bound by the target function.

92. 17:23

    Um, topics basically on our minds here are, uh, it's so powerful, but at the same time, you kind of then duplicate stuff into like user space, kind of like somewhere on, on like a machine.

93. 17:33

    Um, like there's no like package management for this, which then like a year-- like tells the user this is outdated. Um, like we could, uh, we, we, we thought about just adding like a timestamp of, uh, this is the current date where it was fetched, uh, the skill, uh, and then just, uh, oh, if this like older

94. 17:48

    than a month, then try to update. But then we go to second problem of skill distribution and like installing into like the agent environment. Usually this is kind of like gated or not possible for the agent, like depending on, uh, what you use.

95. 18:00

    Um, this user needs to do something to install the skill. This also upgrading doesn't, uh, auto-upgrading doesn't really, uh, really work, but it really depends on the coding agent that you use.

96. 18:08

    And, uh, like the target function is interesting for us because like we can either go for user needs to get to like an initial aha of like, oh, this works, or do we want to directly straight shoot for this is the perfect setup of how you would do evals for this use case?

97. 18:22

    But this is-- I mean, without a skill, it takes like usually like an AI engineering team, it takes like month to get to a perfect setup. Do we now aim for an agent to do this in a single shot and overload the user with lots and lots of questions, or do we just try to get to something

98. 18:34

    and then you can still invoke it again of like improve my setup, ask, and then it can ask questions to improve it? So it's kind of like, well, what's, what's the target for the skill?

99. 18:42

    That was, uh, very interesting for us. Uh, yep. I would invite you to try it and give us feedback, um, because that'd be really interesting. Um, and, uh, like, uh, we do lots of, lots of calls with people from the community every week and, uh, like, like Uh, I think it's not a surprise that I think nobody

100. 18:59

     reads documentation themselves, and everyone is just like, "Yeah, just add this to my-- Like, I just want this to work. Like, just add it." Um, so yeah, the skill is the primary way of how things get done.

101. 19:07

     This is also, like, now the advertised way across all of our documentation, um, that you just should ask your coding agent to do whatever you try to do right now.

102. 19:15

     Um, I'm very excited that it works really well, uh, but also I'm, I'm excited to see what, what comes next. Um, for us as a project roadmap-wise, we see the skill right now, like our users use this when getting started with the, our project, but also, um, to drive a lot of automation around the, like, evaluation life

103. 19:30

     cycle of, oh, I now wanna create like an LLM-as-a-Judge that's aligned with user preferences, or, um, like I got user feedback on a hundred different ex-executions. What do they have in common?

104. 19:40

     Um, do you just then fetch this via the CLIs? So many of these workflows that people needed to do manually now coding agents do for them. Um, we'll bring this in product via, like-- We, we'll help automate this via skills one, bring this in product two, and then three, I, I feel like we just need this orchestration

105. 19:52

     agent, then that's what the team is doing right now. So yeah, I'm very excited for our roadmap to, to automate all of this. But yeah, if you have any feedback, uh, I'm around, Anibal is around.

106. 20:00

     I would love to talk to you. Uh, and yeah, thanks so much for your time. [audience clapping]

107. 20:09

     I don't know if... Do we have time for a question?

108. 20:11

     Yeah.

109. 20:11

     Okay. Yep.

110. 20:15

     I may have missed it. Uh, you say like when you were tuning the skill, the human was completely out of the loop. So you were like-- It was basically like Claude Code directly trying to implement using the skill, the right implementation and then you measured this or?

111. 20:32

     Uh, yeah, it was kind of, um... Like, I mean, you kind of wanna be out of the loop for the experimentation and then just review the suggested changes. Uh, so it was kind of, um, like experimental things, um, give us like all sorts of different recommendations and then human review the suggestions.

112. 20:47

     Like, we didn't accept all because many didn't make sense because our target function wasn't perfect. It was really difficult to get to like a very perfect target function, but it's good at just creating ideas and then, uh, we human reviewed all of the ideas to make the changes to the skill.

113. 21:00

     But so at the end, the skill might not be optimizing for the human AI interaction. Let's say I work with like Claude Code and to try to set up like the trace for whichever project.

114. 21:13

     Then the skill right now might be optimizing for something different and not necessarily what you needed or I don't know.

115. 21:19

     Yeah, that's what we try to, uh, kind of like you need to try it yourself to just get a sense of how it feels to use the skill to then, um, like, like add Langfuse to an application.

116. 21:27

     So we just use it ourselves to get a sense for the feeling because it should... Like, where, where you wanna go is it should feel like, um, like an expert user trying to guide you through what you need for your problem, where usually someone comes in with just like, "I need evals because I read about it online,"

117. 21:43

     uh, but I don't know what actually I need for my application and it kind of like it needs guidance of where you wanna go. Uh, like, "Well, what is your problem?

118. 21:50

     I don't know. What do you, what do you worry about? You probably don't need like a, I don't know, hallucination eval, but probably you need something that's very specific to your application."

119. 21:56

     And, um, yeah, that's what we wanna achieve with the skill that you get like some, some like professional guidance.

120. 22:03

     Yep.

121. 22:05

     Time for a question.

122. 22:06

     Uh, I really resonated with your last point about, uh, skills distribution seems practically insane, uh, right now of just like you just install it, do whatever, GitHub repo link is on day one.

123. 22:17

     Uh, what are your thoughts on sort of like the treating skills as packages, like a PC skills kind of approach or like going full all in on, you know, plug-in marketplaces instead?

124. 22:26

     Like what, what do you think is, you know, able to be adopted by the community versus, uh, sustainable long term if you have like problems with stability?

125. 22:35

     Hmm.

126. 22:36

     Sure.

127. 22:36

     Uh, as a small team, I'm not that excited about plug-in marketplaces because then you need, now need to kind of like maintain all of these proprietary integrations, update them in, I don't know, tell Anthropic, tell OpenAI, tell Cursor whenever you're making-

128. 22:47

     Well, Anthropic and OpenAI are now agreeing on sort of like broadly on it.

129. 22:51

     Yeah. S-still, I mean, like for the skill, uh, I think it would be cool if we just had like a well-known skill or something and like whenever someone is like, "Oh, I wanna, for example, use Langfuse," like the agent can just auto discover that it exists.

130. 23:03

     Uh, like we have it across all of our docs, so I think it would be enough if agent can kind of like ask user, "I wanna install skill." Question is, do you even need to ask?

131. 23:10

     Like I think you only need to ask if the skill is kind of like more trusted than the public web. If it's like same trust level, then why even bother asking?

132. 23:17

     Um, and then two, it's kind of like if I have this installed, it's kind of like a cache of something that was up to date when it, when I installed it, but then the question is how do I know whether it's out of date?

133. 23:27

     So I think so just like timestamping it is enough. So when you use the skill, that agent can be like, "Oh, this seems old. I should probably like fetch a new one."

134. 23:35

     Um, I think this would already go a long way. Um, but yeah, I'm, I'm excited to see. Like, uh, we are going more the timestamp fetch route or alert user of this might be out of date.

135. 23:44

     Uh, that's at least like what we discussed now. Um, but yeah, I'm excited to see what everyone's stripping in this space. Yeah, I, I'm around. Uh, thanks so much.

136. 23:52

     Bye-bye. [audience clapping] [outro music]