WEBVTT 00:00.440 --> 00:05.520 And so with this, we now turn to the other category, the second category of agents. 00:05.720 --> 00:13.320 So by contrast with the workflow patterns with agents, the process is more open ended. 00:13.440 --> 00:15.200 It's something that can keep going. 00:15.680 --> 00:17.320 It has feedback loops. 00:17.360 --> 00:24.240 It's typically a design that allows for information to come back and be processed multiple times. 00:24.520 --> 00:29.720 And importantly, there's no fixed path through the design pattern. 00:29.760 --> 00:35.440 It's not like there's a series of steps as we had in those prior patterns, but rather it's something 00:35.440 --> 00:37.400 that's fluid and dynamic. 00:37.520 --> 00:41.160 And as as a result, it's something which can be much more powerful. 00:41.360 --> 00:42.480 There's many more. 00:42.680 --> 00:48.960 There's a much greater kind of problem that could be taken on by this sort of agent pattern, but it's 00:48.960 --> 00:50.280 less predictable. 00:50.320 --> 00:56.920 And so there are absolutely concerns about robustness, about guardrails, about how do you get the 00:56.920 --> 01:03.320 best out of this kind of flexibility, but still have a system that can run in production and be guaranteed 01:03.360 --> 01:05.800 to complete in a certain time and so on. 01:06.600 --> 01:12.730 And this is the diagram that I would bring up, which is fairly generic kind of diagram that's just 01:12.730 --> 01:17.010 meant to show that you have the inputs and outputs now shown as a human. 01:17.010 --> 01:18.410 But but we don't really have output anymore. 01:18.410 --> 01:24.490 We have environment that's meant to reflect some, some sort of outside world that can be interacted 01:24.490 --> 01:30.210 with, such as the lights that I wore around me, that the human can make a request that goes to an 01:30.210 --> 01:30.730 LM. 01:30.850 --> 01:37.050 The LM is able to take actions on the environment, as you can see, and the LM is able to get back 01:37.090 --> 01:39.690 information from the environment in some way. 01:39.890 --> 01:42.730 And then this is simply a repeating loop. 01:42.770 --> 01:50.170 The LM can continually make other actions and get back feedback, and at such time as it desires, it 01:50.170 --> 01:51.330 can choose to stop. 01:51.570 --> 01:54.130 And so this is the the the diagram. 01:54.290 --> 02:01.970 And there aren't more specific design patterns because it is in itself this sort of open ended design 02:01.970 --> 02:02.650 pattern. 02:02.650 --> 02:05.730 It's saying this is a it's almost a sort of meta design. 02:05.730 --> 02:12.330 It's saying the the LM gets to choose its own design for how it's going to solve the problem. 02:12.570 --> 02:17.460 And that's really it's it's quite hand-wavy, but that's the way to distinguish it. 02:17.500 --> 02:23.420 If we look back at this pattern, I mean, this pattern arguably also could keep going forever. 02:23.420 --> 02:28.940 It's not as if this has a fixed stop, but there's clearly more certainty about what's happening for 02:28.940 --> 02:35.300 what reason than with the agentic patterns, where it is more fluid by its nature. 02:35.620 --> 02:43.500 So that's really the core idea behind the agent patterns, that there is much more flexibility that 02:43.500 --> 02:51.140 allows us to to be able to solve much more complex problems where Llms can really plot their own paths. 02:51.140 --> 02:56.940 But it does introduce some new issues, such as not knowing how long it will take before it will will 02:56.980 --> 03:01.700 complete its task, not knowing whether it will complete its task at all, not knowing what kind of 03:01.700 --> 03:05.260 quality outputs we will get, and not knowing how much it will cost. 03:05.300 --> 03:11.740 These are all the kinds of challenges that agent developers like ourselves need to face, and they are 03:11.740 --> 03:15.020 the sorts of things that we will be tackling in the coming weeks. 03:15.740 --> 03:20.240 And so if this feels a bit unsatisfying that the design patterns for workflows are so clear. 03:20.240 --> 03:22.760 And now this looks just so loose. 03:23.000 --> 03:29.200 Then all I can say is, as we put this into practice, as we code different real agent environments, 03:29.200 --> 03:33.400 this is going to become more and more concrete, and you're going to be able to you'll see this as second 03:33.400 --> 03:34.560 nature before too long. 03:34.880 --> 03:40.680 Now let's just quickly say a few more words about some of the drawbacks that I alluded to a second ago 03:40.720 --> 03:45.480 with using agent frameworks and agentic AI and design patterns. 03:45.800 --> 03:49.680 So obviously there is this sense that you have an unpredictable path. 03:49.720 --> 03:55.320 You don't actually know what order tasks will take place in or even what tasks will happen. 03:55.920 --> 03:58.120 You don't know what the output will be. 03:58.200 --> 04:00.840 There's no guarantees that you're going to get a great output. 04:01.000 --> 04:07.800 Now, this this whole Agentic architecture allows us to take on much bigger, harder problems than we've 04:07.800 --> 04:09.280 ever been able to take on before. 04:09.600 --> 04:16.200 But it does have this inherent uncertainty associated with giving Llms autonomy over how they tackle 04:16.200 --> 04:16.880 problems. 04:17.360 --> 04:19.160 There's also unpredictable costs. 04:19.160 --> 04:23.240 I alluded to that as well a moment ago, but because you don't know how long it's going to take, you 04:23.240 --> 04:29.810 don't know how much it's going to cost in terms of running the APIs and potentially building up a big 04:29.810 --> 04:30.330 bill. 04:31.370 --> 04:36.610 And so as a result, there's a couple of mitigations that are incredibly important, both of which we 04:36.610 --> 04:39.170 will spend plenty of time on in the coming weeks. 04:39.410 --> 04:45.290 One of them is, of course, monitoring, being sure that you have the kinds of visibility into what's 04:45.290 --> 04:51.650 going on behind the scenes so you can understand what's going on with your models, with their interactions, 04:51.650 --> 04:55.690 particularly when you have potentially many agents all interacting. 04:55.730 --> 04:57.930 And we're going to see that with open AI SDK. 04:57.970 --> 05:02.730 We're going to see that the trace, the ability to to watch different agents interacting. 05:02.730 --> 05:06.010 When we look at a graph, we're going to see Lang Smith their tooling. 05:06.010 --> 05:11.690 So we're going to see a lot of the ways that you can have this visibility into what is going on under 05:11.690 --> 05:13.370 the hood in your agent systems. 05:13.690 --> 05:15.410 And then guardrails. 05:15.450 --> 05:22.290 Guardrails is the name for the kinds of protections that you can write in software that makes sure that 05:22.290 --> 05:27.610 your models are doing what they should be doing, or that they're not sort of leaving some, some constraints, 05:27.650 --> 05:29.690 some, some rails that you put in place. 05:29.930 --> 05:35.860 And in fact, that quote that I put here comes straight from the OpenAI agents SDK, which has a ton 05:35.900 --> 05:42.300 of functionality dedicated to guardrails, and they say that it ensures they behave safely, consistently 05:42.300 --> 05:44.820 and within the boundaries that you wish. 05:45.060 --> 05:49.860 And so we will be building those guardrails ourselves in week two. 05:50.540 --> 05:54.300 And with that, that does actually conclude the day. 05:54.340 --> 05:58.580 Two lectures, as I promised you, it was shorter than day one. 05:58.620 --> 06:00.620 And I hope you didn't mind that we covered a lot of theory. 06:00.620 --> 06:03.500 It's really interesting stuff, these gigantic design patterns. 06:03.500 --> 06:09.420 I hope now you've got some clarity on what agents are and that there is ambiguity on this, but but 06:09.460 --> 06:14.580 ways that you can talk about it and think about what it means to be autonomous and what it means to 06:14.580 --> 06:16.980 be able to plot your own path. 06:17.140 --> 06:21.180 Let let an LLM decide the path that things will take. 06:21.540 --> 06:25.940 And so with that, congratulations on completing day two and day three. 06:25.940 --> 06:28.300 We're going to talk about orchestrating Llms. 06:28.300 --> 06:31.180 It's going to be more and more coding, more hands on, which will be fun. 06:31.180 --> 06:34.980 We're going to build with a lot of APIs and get ready for that. 06:34.980 --> 06:37.940 And after that we're going to move on to tools. 06:38.140 --> 06:39.700 So I will see you next time.