WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=jA_x7F8caHI NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/jA_x7F8caHI.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:12.719 --> 00:00:15.759 Everyone, are we having fun? s2 00:00:15.920 --> 00:00:18.320 Oh, you've got to give me way more than that. s3 00:00:18.320 --> 00:00:21.600 So, let me tell you, um, my name is Kay Malcolm. s4 00:00:21.600 --> 00:00:24.240 I am a retired hiphop instructor. s5 00:00:24.240 --> 00:00:28.400 So, if I don't get more energy than that, we will start. s6 00:00:28.400 --> 00:00:29.279 We'll start. s7 00:00:29.279 --> 00:00:31.440 Are you having fun? s8 00:00:31.440 --> 00:00:32.000 [cheering] s9 00:00:32.000 --> 00:00:32.399 Okay. s10 00:00:32.399 --> 00:00:32.960 All right. s11 00:00:32.960 --> 00:00:36.239 So, here's what we're going to talk about today. s12 00:00:36.239 --> 00:00:39.360 Now, you guys have heard a lot about two letters. s13 00:00:39.360 --> 00:00:44.559 Does anyone want to guess what those two letters are that I'm going to talk about today? s14 00:00:46.239 --> 00:00:46.719 Data. s15 00:00:46.719 --> 00:00:47.520 That was pretty good. s16 00:00:47.520 --> 00:00:48.719 DB. s17 00:00:48.719 --> 00:00:54.559 I'm going to talk about AI, but I'm specifically going to talk about agent harnesses. s18 00:00:54.559 --> 00:00:57.520 But before I do that, I want to introduce you all to a few people. s19 00:00:57.520 --> 00:00:59.600 Is that okay? s20 00:00:59.600 --> 00:01:00.480 Yes or yes. s21 00:01:00.480 --> 00:01:02.559 Is that okay? s22 00:01:02.559 --> 00:01:04.239 I gave choices. s23 00:01:04.239 --> 00:01:04.720 Yes. s24 00:01:04.720 --> 00:01:05.119 Anyway. s25 00:01:05.119 --> 00:01:05.519 All right. s26 00:01:05.519 --> 00:01:05.920 Okay. s27 00:01:05.920 --> 00:01:06.640 All right. s28 00:01:06.640 --> 00:01:09.439 This is my team. s29 00:01:09.439 --> 00:01:14.479 I run an outbound database product management team at Oracle. s30 00:01:14.479 --> 00:01:18.000 I've been at Oracle a really long time, 20 years. s31 00:01:18.000 --> 00:01:20.240 Funny story, I started when I was 12. s32 00:01:20.240 --> 00:01:23.920 So, don't do the math and don't start start adding in your head. s33 00:01:23.920 --> 00:01:27.280 Um, and we've got a problem. s34 00:01:27.280 --> 00:01:40.479 That problem is I've got one group that does platform development and then I have another group that does content development for live labs, a platform that I wrote myself. s35 00:01:40.479 --> 00:01:44.960 So yeah, I'm an engineer but I'm kind of a developer poser too. s36 00:01:44.960 --> 00:01:53.360 And then I've got another group who does QA and then I have another group who does my front-end development with AI. s37 00:01:53.360 --> 00:01:57.280 Here's what I found out as a leader. s38 00:01:57.280 --> 00:02:05.759 Because in the token maxing era of 2025, because you know, we're not token maxing anymore, right? s39 00:02:05.759 --> 00:02:08.239 We are responsible AI now. s40 00:02:08.239 --> 00:02:21.840 But in the token maxing era, the thing that I found out was while AI was making the individuals on my team faster, there was another problem it was creating. s41 00:02:23.520 --> 00:02:28.720 It wasn't making my team more productive. s42 00:02:28.800 --> 00:02:43.760 And the reason was when one team from the Netherlands checked in code at my 4:00 a.m. in the morning because I've got half of my team that's in AMIA and I have half of my team who that are here in the United States. s43 00:02:43.760 --> 00:02:49.680 They checked in the code but they didn't check in their context from Codeex. s44 00:02:49.680 --> 00:02:51.920 We use Codeex at Oracle. s45 00:02:51.920 --> 00:03:01.040 So then when the US team woke up, they got the code but no information about the context. s46 00:03:01.040 --> 00:03:07.120 So we used AI to solve a problem that AI created. s47 00:03:07.120 --> 00:03:08.879 And here's what we did. s48 00:03:08.879 --> 00:03:10.560 Oh well, let me talk about this first. s49 00:03:10.560 --> 00:03:15.599 So some of the issues, the context, like I said, wasn't shared. s50 00:03:15.599 --> 00:03:18.239 GitHub wasn't tracking that. s51 00:03:18.239 --> 00:03:26.480 I had repositories that were diverging and I was asking the managers who work for me, what's happening to your teams? s52 00:03:26.480 --> 00:03:29.519 Why why are we not going faster? s53 00:03:29.519 --> 00:03:31.440 We're spending all of this money on tokens. s54 00:03:31.440 --> 00:03:40.159 We're spending all this money on AI, yet something is missing because we're still spending time doing testing and validation. s55 00:03:40.159 --> 00:03:49.920 So our netn net wasn't really wasn't really working for us because git records the code and not human intent. s56 00:03:49.920 --> 00:04:03.360 So it's a problem and even though code creation was no longer our problem still had a bottleneck. s57 00:04:03.360 --> 00:04:06.480 We needed a collaboration layer. s58 00:04:06.480 --> 00:04:14.080 Now, I do have members of my team in the audience, so don't judge me, and you know who you are. s59 00:04:14.080 --> 00:04:18.000 I'm not saying that you all didn't collaborate. s60 00:04:18.000 --> 00:04:23.520 But now, we've got a new team member, and that new team member is AI. s61 00:04:23.520 --> 00:04:29.520 So, we needed to figure out how to track our progress in our next steps. s62 00:04:29.520 --> 00:04:39.280 how to rationalize decisions that the agent was making. s63 00:04:39.600 --> 00:04:45.759 We needed to figure out how to resolve questions in conflicts. s64 00:04:45.759 --> 00:04:49.280 Okay, hold my problem. s65 00:04:49.280 --> 00:04:51.600 Will you all hold my problem for me right here? s66 00:04:51.600 --> 00:04:54.240 We're going to just tuck that in a little box. s67 00:04:54.240 --> 00:04:59.520 Let me define what a enterprise agent actually is. s68 00:04:59.520 --> 00:05:05.199 Now, most people think that an enterprise agent is the model and workflow. s69 00:05:05.199 --> 00:05:08.320 How many people agree with me? s70 00:05:08.639 --> 00:05:11.039 Man, this tough crowd you. s71 00:05:11.039 --> 00:05:12.960 Okay, one person. s72 00:05:12.960 --> 00:05:15.199 Okay, the rest of you think it's a little bit more. s73 00:05:15.199 --> 00:05:26.800 Okay, let's see what could it be that a real enterprise agent has tools. s74 00:05:26.960 --> 00:05:30.560 Tools are how it does things. s75 00:05:30.880 --> 00:05:32.880 Context. s76 00:05:33.280 --> 00:05:34.000 The context. s77 00:05:34.000 --> 00:05:35.280 That's a context window. s78 00:05:35.280 --> 00:05:38.720 That's what's in the actual prompt. s79 00:05:39.360 --> 00:05:40.880 Memory. s80 00:05:40.880 --> 00:05:42.880 Huh? s81 00:05:43.120 --> 00:05:51.840 And if you're thinking, "But wait, K, memory, you just said that the model is kind of like the brain of the operation." s82 00:05:51.840 --> 00:05:52.560 Hold tight. s83 00:05:52.560 --> 00:05:58.880 We're going to talk a little bit more about memory retrieval because you don't want to get everything back. s84 00:05:58.880 --> 00:06:04.000 So that's being able to to retrieve the right information back. s85 00:06:04.000 --> 00:06:10.800 And then I know that there are a lot of developers here and you all don't care about security. s86 00:06:11.280 --> 00:06:22.319 I care about security because I work for the most secure database company and I used to work for um agency that has no name. s87 00:06:22.319 --> 00:06:25.680 But guard rails is also important. s88 00:06:25.680 --> 00:06:28.000 This is the harness. s89 00:06:28.000 --> 00:06:38.319 I speak in analogies and I and I speak in stories because if I tell you this and Marvel, you know exactly what I'm talking about. s90 00:06:38.319 --> 00:06:49.360 So the agent, think of it as the model, little brain floating a in a glass jar plus this harness. s91 00:06:49.360 --> 00:06:52.560 This harness is the body. s92 00:06:52.560 --> 00:06:58.240 So it's how the agent can actually do things and get things done. s93 00:06:58.240 --> 00:07:03.520 That memory, that's the part of the central nervous system. s94 00:07:03.520 --> 00:07:10.880 And you remember the central nervous system connects the brain to the rest of the body, legs, arms. s95 00:07:10.880 --> 00:07:16.479 That's the part of the central nervous system that carries context. s96 00:07:16.479 --> 00:07:20.319 So you remember my problem with Git. s97 00:07:20.319 --> 00:07:22.000 What I needed was memory. s98 00:07:22.000 --> 00:07:26.479 Okay, so there are a number of memory types. s99 00:07:26.479 --> 00:07:32.400 I chose five, the five most common ones that people talk about and these are the ones that I want you to remember. s100 00:07:32.400 --> 00:07:34.400 The first one is short-term memory. s101 00:07:34.400 --> 00:07:36.960 That's the session, right? s102 00:07:36.960 --> 00:07:47.599 And so if you're storing memory of an AI uh process, that is the short-term memory is if you're with chat, cloud code, right? s103 00:07:47.599 --> 00:07:49.680 Codeex, pick your poison. s104 00:07:49.680 --> 00:07:55.039 The long-term memory is what persists across sessions. s105 00:07:55.520 --> 00:07:56.879 Episodic memory. s106 00:07:56.879 --> 00:07:58.479 Hm. s107 00:07:58.479 --> 00:08:06.800 What happened the last time I interacted with fill in the blank? s108 00:08:06.800 --> 00:08:09.199 That's your episodic memory. s109 00:08:09.199 --> 00:08:15.759 Procedural memory tools, steps that were taken. s110 00:08:15.759 --> 00:08:17.840 And then finally, semantic memory. s111 00:08:17.840 --> 00:08:21.199 And semantic memory because we're talking enterprise agents. s112 00:08:21.199 --> 00:08:28.800 We're not talking the agent that I built, Sasha Fierce, because remember I told you guys that I was a I'm a a dancer. s113 00:08:28.800 --> 00:08:34.640 So, of course, my my chief of staff is going to be called Sasha Fierce because that was Beyonce. s114 00:08:34.640 --> 00:08:37.519 Any Beyonce fans? s115 00:08:37.519 --> 00:08:38.240 Okay, I'm sorry. s116 00:08:38.240 --> 00:08:40.320 All right, we got to focus. s117 00:08:40.320 --> 00:08:43.519 Okay, so these are the memory types. s118 00:08:43.519 --> 00:08:53.200 Now, when you're defining this real enterprise agent in this memory, there's something you need to consider where to store it. s119 00:08:53.200 --> 00:08:55.600 And so, I'm going to tell you guys a story. s120 00:08:55.600 --> 00:09:00.720 But when I tell you the story, you have to promise me that you're not going to judge me. s121 00:09:00.720 --> 00:09:03.120 Do you promise? s122 00:09:06.320 --> 00:09:08.720 Do you promise s123 00:09:08.720 --> 00:09:10.240 you're not recording me, right? s124 00:09:10.240 --> 00:09:12.320 Because this doesn't paint me in a good light. s125 00:09:12.320 --> 00:09:12.720 Okay. s126 00:09:12.720 --> 00:09:13.040 All right. s127 00:09:13.040 --> 00:09:14.959 The world of data was one simple. s128 00:09:14.959 --> 00:09:18.000 I've been at Oracle a long time, but I came from a customer. s129 00:09:18.000 --> 00:09:19.760 That customer's name was Southern Company. s130 00:09:19.760 --> 00:09:20.800 It was a power company. s131 00:09:20.800 --> 00:09:22.800 I'm based out of Atlanta. s132 00:09:22.800 --> 00:09:28.399 And I was hired at Southern Company because I was a rockar performance tuner. s133 00:09:28.399 --> 00:09:29.760 You had a SQL query. s134 00:09:29.760 --> 00:09:31.519 I mean, I'm dating myself, but whatever. s135 00:09:31.519 --> 00:09:32.800 You had a SQL query. s136 00:09:32.800 --> 00:09:34.959 I knew all of the innit.org parameters. s137 00:09:34.959 --> 00:09:37.760 Even the ones when you called support and they said, "Don't remember these. s138 00:09:37.760 --> 00:09:38.560 Don't write them down." s139 00:09:38.560 --> 00:09:40.399 I wrote them down in my little notebook. s140 00:09:40.399 --> 00:09:44.000 I could tune a query within one inch of its life. s141 00:09:44.000 --> 00:09:51.120 Then one of you came to my desk because I mean the world the world was rows and columns. s142 00:09:51.120 --> 00:10:01.360 It was a great time back in my Albundy days um and said hey I need to store data unstructured. s143 00:10:01.360 --> 00:10:04.320 Why I need to do that? s144 00:10:04.320 --> 00:10:13.839 And so me being K the the diligent DBA, I was like, "Let me figure it out and get back to you." s145 00:10:14.000 --> 00:10:16.720 Did I get back to him? s146 00:10:16.959 --> 00:10:18.880 I didn't get back to him. s147 00:10:18.880 --> 00:10:26.720 Now, the thing you have to know about Southern Company was for every database system that a DBA managed, I had to attend two meetings. s148 00:10:26.720 --> 00:10:31.600 Today, when I hear Sarbain Oxley, I still throw up a little bit in the back of my throat. s149 00:10:31.600 --> 00:10:36.000 So I had to attend a security meeting and a patching meeting every week. s150 00:10:36.000 --> 00:10:37.360 Never failed. s151 00:10:37.360 --> 00:10:45.040 Now because this developer installed a database that was specialized for unstructured. s152 00:10:45.040 --> 00:10:47.120 Okay, there are really smart people in the room. s153 00:10:47.120 --> 00:10:50.399 How many meetings am I going to now? s154 00:10:54.320 --> 00:10:55.279 Four. s155 00:10:55.279 --> 00:10:59.920 Okay, I'm a little annoyed, but I'm like, okay, we can we can we can do this. s156 00:10:59.920 --> 00:11:06.560 Then they said, "Okay, since you're such a good tuner, I need you to figure out this relationship." s157 00:11:06.560 --> 00:11:18.320 Now, the way that Southern Company worked, there was this people could die application, um, and it was a it was like a Nokia phone that people who were climbing the towers, right? s158 00:11:18.320 --> 00:11:23.040 So, you guys have have been in a storm and the power goes out, right? s159 00:11:23.040 --> 00:11:27.680 And then you're pretty sure that within maybe an hour or two the power will go on. s160 00:11:27.680 --> 00:11:39.120 Well, that system that would tell the people who were climbing those trees and risking their lives to turn the power back on sometimes would have false positives or false negatives. s161 00:11:39.120 --> 00:11:50.160 So, they wanted to look at all of the other um uh polls in the area to try to get away from the false positive or the false negative. s162 00:11:50.160 --> 00:11:54.160 And so I did that in a SQL query and it was it was amazing. s163 00:11:54.160 --> 00:11:58.079 It was a five nested union all statement. s164 00:11:58.079 --> 00:11:59.519 It was some of my best work. s165 00:11:59.519 --> 00:12:06.320 Now it might have taken like 20 minutes to work but it was like a predecessor to graph. s166 00:12:06.399 --> 00:12:10.320 Yeah, they they install Neo4j. s167 00:12:10.800 --> 00:12:13.200 So now how many meetings am I going to? s168 00:12:13.200 --> 00:12:14.480 Six. s169 00:12:14.480 --> 00:12:16.959 That's a problem. s170 00:12:17.040 --> 00:12:19.600 So I um Oh, let me I got ahead of myself. s171 00:12:19.600 --> 00:12:20.720 So you know what I did? s172 00:12:20.720 --> 00:12:22.959 I quit. s173 00:12:23.200 --> 00:12:27.760 I left and I came to Oracle because I was like this is a problem and maybe I can go to Oracle to help solve it. s174 00:12:27.760 --> 00:12:33.440 So then Joe Mundy called me and he said hey um we are installing Reddus. s175 00:12:33.440 --> 00:12:35.120 Oracle is late to the game. s176 00:12:35.120 --> 00:12:38.079 We've got a vector database. s177 00:12:38.160 --> 00:12:40.160 Okay. s178 00:12:40.800 --> 00:12:43.839 But here's your problem, Joe. s179 00:12:45.920 --> 00:12:49.040 Agents now need access to all of this data. s180 00:12:49.040 --> 00:13:03.120 So if data is in an Oracle database, if then it's also in an unstructured JSON database, if it's in a graph database and it's in a vector database, where is your single source of the truth? s181 00:13:03.120 --> 00:13:05.040 The agent has to figure that out. s182 00:13:05.040 --> 00:13:08.079 Sometimes it'll get it right. s183 00:13:08.480 --> 00:13:12.399 Most times it'll get it wrong and it's going to burn up a whole bunch of tokens. s184 00:13:12.399 --> 00:13:20.399 And so now if you want to store your memory somewhere, you can store it in a file system. s185 00:13:20.399 --> 00:13:29.040 You can store it in clawed or chatgpt because we all know about the memory.md file. s186 00:13:29.519 --> 00:13:31.120 But that's going to be a problem. s187 00:13:31.120 --> 00:13:32.240 Now I want to illustrate this. s188 00:13:32.240 --> 00:13:33.760 I need four volunteers. s189 00:13:33.760 --> 00:13:35.600 I can see you raise your hand. s190 00:13:35.600 --> 00:13:37.920 One, two. s191 00:13:37.920 --> 00:13:38.560 Okay, I can't. s192 00:13:38.560 --> 00:13:39.200 Maybe I can. s193 00:13:39.200 --> 00:13:40.720 Three. s194 00:13:40.720 --> 00:13:41.600 I need a fourth. s195 00:13:41.600 --> 00:13:43.040 Ah, fourth in the back. s196 00:13:43.040 --> 00:13:43.440 Okay. s197 00:13:43.440 --> 00:13:44.639 Fourth in the back. s198 00:13:44.639 --> 00:13:47.440 You are going to be our old reliable. s199 00:13:47.440 --> 00:13:49.040 You're going to be a relational database. s200 00:13:49.040 --> 00:13:50.560 Yes or yes. s201 00:13:50.560 --> 00:13:52.880 You got your So, you have your assignment. s202 00:13:52.880 --> 00:13:53.200 Okay. s203 00:13:53.200 --> 00:13:54.639 And then there was someone here. s204 00:13:54.639 --> 00:13:57.600 You're going to be my unstructured database. s205 00:13:57.600 --> 00:13:58.000 Okay. s206 00:13:58.000 --> 00:13:59.360 And then where was my other? s207 00:13:59.360 --> 00:14:00.880 Ah, very good. s208 00:14:00.880 --> 00:14:02.320 You're going to be my graph database. s209 00:14:02.320 --> 00:14:03.040 You good? s210 00:14:03.040 --> 00:14:03.920 Relationship guy. s211 00:14:03.920 --> 00:14:05.680 You look like a relationship guy. s212 00:14:05.680 --> 00:14:06.079 All right. s213 00:14:06.079 --> 00:14:06.560 Very good. s214 00:14:06.560 --> 00:14:06.959 Fourth. s215 00:14:06.959 --> 00:14:09.279 Where was my fourth? s216 00:14:09.279 --> 00:14:10.800 Was it you? s217 00:14:10.800 --> 00:14:12.079 Yes. s218 00:14:12.079 --> 00:14:12.800 Yeah. s219 00:14:12.800 --> 00:14:15.440 You are my vector database. s220 00:14:15.440 --> 00:14:16.160 Okay. s221 00:14:16.160 --> 00:14:20.880 Now, everybody be really, really quiet. s222 00:14:20.880 --> 00:14:24.560 For my four volunteers, I need you all. s223 00:14:24.560 --> 00:14:27.920 I'm going to say something to you. s224 00:14:27.920 --> 00:14:34.000 And I need you all to decide how you're going to store it and who's going to have the single source of the truth. s225 00:14:34.000 --> 00:14:40.240 You can't get up from your seats and you have to whisper because if you talk loud that's 5x the tokens for you. s226 00:14:40.240 --> 00:14:41.839 Yes. s227 00:14:41.839 --> 00:14:42.079 Okay. s228 00:14:42.079 --> 00:14:43.040 Are we ready? s229 00:14:43.040 --> 00:14:44.720 All right. s230 00:14:44.720 --> 00:14:48.399 The cow jumped over the moon. s231 00:14:49.040 --> 00:14:51.040 Go. s232 00:14:53.519 --> 00:14:55.519 Doesn't really work, does it? s233 00:14:55.519 --> 00:14:56.639 That's a problem. s234 00:14:56.639 --> 00:14:59.920 Okay, Oracle. s235 00:14:59.920 --> 00:15:12.000 And if you don't forget one, if you forget everything I say and you remember one thing, Oracle is not the Oracle that you think that is why I am here today. s236 00:15:12.000 --> 00:15:26.639 How many of you knew that Oracle could natively in the same table down to the same partition store JSON graph vector my vector friend over there my JSON friend s237 00:15:26.639 --> 00:15:44.160 spatial you want your memory to be immutable blockchain in the same database raise your hand yeah we have a marketing problem so any data type can be stored in a 26AI s238 00:15:44.160 --> 00:16:01.920 database any workload anywhere AWS GCP Azure OCI on prem choice and flexibility so now when we take this and we talk about the agent I want to be able to store my long-term and procedural memory in relational s239 00:16:01.920 --> 00:16:11.680 on in JSON I want to store my short-term and my long-term memory graph I want to store procedural because procedural that's how I figure out the relationships right the steps s240 00:16:11.680 --> 00:16:14.079 my episodic and semantic memory. s241 00:16:14.079 --> 00:16:18.720 I need to do some vector and then store it also as text. s242 00:16:18.720 --> 00:16:25.759 Now, if I have four different databases, you all saw they can't talk to each other. s243 00:16:25.759 --> 00:16:27.839 It's going to be a problem. s244 00:16:27.839 --> 00:16:38.320 And so, what I'm saying to you today is the Oracle AI database is the best place to store this agent memory that's going to power your harness. s245 00:16:38.320 --> 00:16:40.196 Remember your harness is your body s246 00:16:40.196 --> 00:16:40.639 [music] s247 00:16:40.639 --> 00:16:44.320 and that memory is your central nervous system. s248 00:16:44.320 --> 00:16:46.399 Okay, back to PY. s249 00:16:46.399 --> 00:16:51.920 So the problem that I had, we solved it with a memory broker named Py. s250 00:16:51.920 --> 00:16:54.160 We used agent memory. s251 00:16:54.160 --> 00:16:58.160 We got out of that automatic continuity. s252 00:16:58.160 --> 00:17:07.839 So with my team, they were able to share not just their code, but PY also kept track of the context. s253 00:17:07.839 --> 00:17:20.079 So if one context window was had procedural memory, episodic memory, information about the long-term memory, that was then shared with the other folks on the team. s254 00:17:20.079 --> 00:17:21.760 You could call them agents, if you will. s255 00:17:21.760 --> 00:17:26.480 They're just human agents shared across forks. s256 00:17:26.480 --> 00:17:41.600 The developers on the team remained in control while Polly was able to create the context, figure out which fork and branch it belonged to and which commit it belonged to. s257 00:17:41.600 --> 00:17:49.600 Now this is a very simplistic example but when you take this to the enterprise here's what happens. s258 00:17:49.600 --> 00:17:54.080 Memory is the thing that becomes non-negotiable in an agent's harness. s259 00:17:54.080 --> 00:17:59.440 Now these are three papers that that I read um on the airplane. s260 00:17:59.440 --> 00:18:11.200 This first one is from open AI and it's about its in-house data agent and the thing that it says is it is saying that its in-house data agent actually needs memory. s261 00:18:11.200 --> 00:18:20.080 Memory was crucially important to ensure that its agent was able to filter correctly instead of trying to string match. s262 00:18:20.080 --> 00:18:23.440 Harrison Chase said, "Your harness, your memory. s263 00:18:23.440 --> 00:18:26.640 And if you don't own your harness, you don't own your memory." s264 00:18:26.640 --> 00:18:28.080 Which is key. s265 00:18:28.080 --> 00:18:31.120 And then I'm sure you all are wondering, "Well, Claude has memory. s266 00:18:31.120 --> 00:18:32.240 Why can't I use that?" s267 00:18:32.240 --> 00:18:45.360 Well, it's kind of like file system memory and it works with one, but just like in my example, when you scale past one, and you're going to scale past one in the enterprise, it creates a problem. s268 00:18:45.360 --> 00:18:51.919 So Oracle has a Oracle agent memory pap package. s269 00:18:51.919 --> 00:18:54.240 PIP install Oracle agent memory. s270 00:18:54.240 --> 00:18:56.000 You get access to it. s271 00:18:56.000 --> 00:19:08.320 And this memory is this SDK that we have is the thing that will hold your live conversations, your memories, your facts and figure out what is worth keeping. s272 00:19:08.320 --> 00:19:16.400 So if we look at Py now, Kevin can share his context with Py, our memory broker. s273 00:19:16.400 --> 00:19:19.200 We use the Oracle agent memory SDK. s274 00:19:19.200 --> 00:19:23.120 It's stored in an Oracle autonomous database. s275 00:19:23.120 --> 00:19:31.120 We can use the LLM of our choice or we can use a local model through the Oracle private AI services container. s276 00:19:31.200 --> 00:19:38.559 And then Linda, who's actually sitting right here, can interact and work with Kevin, no issues. s277 00:19:38.559 --> 00:19:42.320 So yes, AI makes individuals faster. s278 00:19:42.320 --> 00:19:47.039 Shared memory on an Oracle AI database makes teams faster. s279 00:19:47.039 --> 00:19:49.280 So I don't want you all to compromise. s280 00:19:49.280 --> 00:20:03.280 In the age of AI, what 26AI does is you can choose and pick what's best for agent memory file system stored in a database file system or in the database. s281 00:20:03.280 --> 00:20:07.039 If you need to do data modeling, you've got JSON, you've got relational. s282 00:20:07.039 --> 00:20:10.000 We've got choice. s283 00:20:10.080 --> 00:20:12.720 Okay, I've got some goodies for you. s284 00:20:12.720 --> 00:20:14.720 The Oracle AI developer hub. s285 00:20:14.720 --> 00:20:19.919 That's where you can guys you guys can get coding materials, the applications, what I talked about today. s286 00:20:19.919 --> 00:20:21.440 Livelabs.oracle.com. s287 00:20:21.440 --> 00:20:30.159 If you've done any of our workshops today, that happens to be something that I wrote myself about six years ago and 40 million users um ago. s288 00:20:30.159 --> 00:20:33.600 Spend my OCI tenency money. s289 00:20:33.600 --> 00:20:40.720 kick the tires on any Oracle technology um for six hours, 12 hours, however long you need. s290 00:20:40.720 --> 00:20:44.480 Um and then I'm giving you all all a Mac Mini. s291 00:20:44.480 --> 00:20:45.360 No, I'm just kidding. s292 00:20:45.360 --> 00:20:47.360 I'm giving you an OCI Mini. s293 00:20:47.360 --> 00:20:50.559 So, I don't know if you knew, but there is an always free OCI. s294 00:20:50.559 --> 00:21:02.400 It is the most generous of any of the hyperscalers where you can get a free Oracle database, free compute, you can send 3,000 emails a month, 200 gig in storage, s295 00:21:02.400 --> 00:21:05.440 and if you click on that, you can get access to it. s296 00:21:05.440 --> 00:21:10.320 Or just search Google for Oracle Cloud, always free. s297 00:21:10.320 --> 00:21:11.919 Connect with me. s298 00:21:11.919 --> 00:21:15.840 If you build something, will you all message me and let me know? s299 00:21:15.840 --> 00:21:18.240 Yes or yes? s300 00:21:18.240 --> 00:21:20.480 Thank you. s301 00:21:34.294 --> 00:21:36.294 [music]