WEBVTT
00:00:02.160 --> 00:00:16.079
In the past 24 hours, two of the biggest players in the AI space, OpenAI and Google, both launched updated versions of their simple drag and drop agents that work for you with your context around the clock.
00:00:16.239 --> 00:00:21.679
And that got me thinking of the common features versus benefits methodology when it comes to marketing.
00:00:21.839 --> 00:00:23.679
If you've never heard of it, it's pretty simple.
00:00:23.920 --> 00:00:28.000
Features describe the technical facts or the specs of what a product does.
00:00:28.079 --> 00:00:34.479
And well, the benefits explain the personal value the human gets from using said products.
00:00:34.640 --> 00:00:41.039
And in AI, we've seen a similar features versus benefits narrative take shape over the past few years.
00:00:41.359 --> 00:00:46.719
The feature, large language models are smarter and faster than humans when used correctly.
00:00:46.960 --> 00:00:49.840
The benefit, humans can be more productive.
00:00:50.159 --> 00:00:54.880
But the feature side has completely exploded over the past two months.
00:00:55.039 --> 00:00:58.240
And the benefit side is still being written.
00:00:58.560 --> 00:00:59.679
Stick with me here.
00:00:59.920 --> 00:01:14.400
So as large language models have become overly agentic by default overnight and as capable as humans, there's a new benefit paradigm for AI agents that has flown completely under the radar.
00:01:14.719 --> 00:01:17.840
And I know that this is the next big trend coming.
00:01:17.920 --> 00:01:22.560
It doesn't have a name, but I'm going to go ahead and name it now and explain the concept.
00:01:22.640 --> 00:01:28.640
I'm calling this Scheduled Agentic Context Carry or SAC, S-A-C-C.
00:01:28.959 --> 00:01:39.920
And I think the companies taking the time to understand and iterate on this new concept now are going to be the ones crushing their year-end goals and KPIs in quarter four.
00:01:40.159 --> 00:01:51.680
So let's unwind this kind of new concept together because I think understanding this now is one of the most important investments you can make on your AI journey this year.
00:01:51.920 --> 00:01:55.040
So let's start there with our Start Here series.
00:01:55.200 --> 00:01:57.519
If you're new here, welcome to Everyday AI.
00:01:57.599 --> 00:02:00.319
And this is our here series.
00:02:00.560 --> 00:02:03.599
But let's first start with the big picture of what's going on.
00:02:03.840 --> 00:02:11.120
Anthropic, OpenAI, Microsoft, and Perplexity have all shipped scheduled agents just this spring.
00:02:11.439 --> 00:02:15.280
And most business leaders are still using AI just like a chatbot.
00:02:15.360 --> 00:02:24.800
I'm going to go in and I'm going to technically reactively ask an AI chatbot for something and probably have to re-explain a lot to waste a lot of time.
00:02:24.960 --> 00:02:31.120
But there's been a quiet workflow pattern emerging beneath every one of these recent product launches.
00:02:31.280 --> 00:02:37.120
And that's why we're talking about this new concept of scheduled agentic context carry.
00:02:37.919 --> 00:02:42.240
So on today's show, that's exactly what we're going to over, kind of go over.
00:02:42.319 --> 00:02:43.520
And here's what you're going to learn.
00:02:43.680 --> 00:02:51.280
You're going to learn what agentic context carry actually means and why you've never heard of it, and but why it absolutely matters.
00:02:51.439 --> 00:02:57.439
You're going to know how this hidden workflow bridges, chatbots, and the fully autonomous AI future.
00:02:57.599 --> 00:03:08.000
You're going to understand the timing of all these things coming together, and specifically these now huge 1 million token context windows that have quietly changed everything this spring.
00:03:08.159 --> 00:03:13.360
And you're going to know the exact three steps to deploy this pattern inside of your business today.
00:03:13.599 --> 00:03:16.080
All right, let's get started.
00:03:16.400 --> 00:03:17.759
Welcome to Everyday AI.
00:03:17.840 --> 00:03:20.639
My name is Jordan Wilson, and this is the Start Here series.
00:03:20.800 --> 00:03:26.319
So after 750 plus podcasts, I've never had an answer when someone was like, I'm new.
00:03:26.400 --> 00:03:27.199
Where do I start?
00:03:27.360 --> 00:03:28.560
Well, you start here.
00:03:28.879 --> 00:03:39.599
Our Start Here series is an ongoing effort to help business leaders both better understand trending and emerging concepts, but also for those who are brand new to get caught up.
00:03:39.759 --> 00:03:47.439
So I'd recommend you start with episode one of the Start Here Series and listen in order, but go ahead and listen to this one and then you can go backtracked.
00:03:47.520 --> 00:03:52.000
But make sure you go to starthere series.com because, well, it's going to make it much easier to do that.
00:03:52.159 --> 00:03:55.599
That's going to give you free access to our inner circle community.
00:03:55.759 --> 00:04:01.599
Right now, there's no other way for the general public to sign up except starthearseries.com.
00:04:01.840 --> 00:04:12.560
And then inside the Start Here Series space, there will be an updated Spotify playlist where you can go listen to all of the Start Here series very easily in order in a dedicated playlist.
00:04:12.879 --> 00:04:13.120
All right.
00:04:13.199 --> 00:04:25.120
And if you miss our last Start Here Series episode, we uh in volume 20, so this is volume 21, we talked about AI change management that works, five moves, the top five percent make.
00:04:25.360 --> 00:04:30.160
All right, but let's get into this concept of a gentic context carry.
00:04:30.319 --> 00:04:38.399
And y'all, every single major AI lab and the big third-party players launched something.
00:04:38.560 --> 00:04:41.680
So from the big four, right?
00:04:41.839 --> 00:04:50.720
So that's Anthropic, Microsoft, Google, OpenAI, and then even Perplexity and you know, open clock technically fall under this category.
00:04:50.879 --> 00:04:55.839
But literally, everyone launched something and it's all very timely.
00:04:55.920 --> 00:05:16.480
So I did mention just the past 24 hours with big announcements uh from Google Gemini at their Cloud Next conference, and then with OpenAI's new agents that we're going to talk about here in a minute, but also clawed code routines, uh, right, that can bring uh automated, scheduled agent runs to your guest stop.
00:05:16.800 --> 00:05:36.639
So it is kind of like the uh maybe this open claw movement that happened, you know, really in February and March of this year, actually kind of forced the hand of all of the big companies to say, okay, it seems like we essentially to oversimplify it, we need an AI agent that can run on a cron, right?
00:05:36.720 --> 00:05:43.360
Run on a schedule, where someone can go in and they say, Hey, AI agent, at this time, every single day, I want this to happen.
00:05:43.519 --> 00:05:48.000
Uh, so we had the Claude Code Routines, which I absolutely love.
00:05:48.079 --> 00:05:49.279
That runs on your desktop.
00:05:49.519 --> 00:05:59.360
Similarly, OpenAI in their codex platform just added schedule work and persistent memory two days after that claw uh code routines announcement in mid-April.
00:05:59.519 --> 00:06:05.040
And then we also just got wind that co-pilot co-work uh officially launched in Frontier.
00:06:05.199 --> 00:06:12.160
So you have essentially all these scheduled uh agent platforms slash co-working platforms, right?
00:06:12.240 --> 00:06:14.800
So, like Claude Co-Work uh is the big one.
00:06:14.879 --> 00:06:21.920
Uh Microsoft Copilot uses essentially the Claud Co-Work technology because they are an investor in Anthropic.
00:06:22.000 --> 00:06:25.040
Uh so you have those kind of two places come together.
00:06:25.120 --> 00:06:30.879
You have these co-working uh kind of elements that allow for scheduling and it brings all your context.
00:06:31.040 --> 00:06:35.759
And then you have these scheduled agents, and it's all literally exploded out of nowhere.
00:06:35.920 --> 00:06:43.519
And although, you know, this may technically be a more timely um episode with all of these things happening now.
00:06:43.680 --> 00:07:02.240
The reason why I'm doing it in the start here series, whether you are listening to this in April or you're listening to it, I don't know, in the year 2027, is because I think that this is going to be a noticeable pivot in how the enterprise starts to interface with AI agents.
00:07:02.399 --> 00:07:03.839
Because here's the reality, right?
00:07:03.920 --> 00:07:08.000
We've been hearing since probably late 2024 that, oh, it's the year of AI agents.
00:07:08.160 --> 00:07:09.279
And it didn't happen.
00:07:09.360 --> 00:07:11.199
And in 2025, it didn't happen.
00:07:11.360 --> 00:07:28.639
But I think we've now come to that realization in 2026 in a certain way, because we've noticed that the fully autonomous uh AI agents, where you just give them a goal and then they go off and run on their own, not as reliable as we'd like, mainly because of safety concerns, guardrails, et cetera.
00:07:28.800 --> 00:07:40.720
So I think that this new kind of co-working scheduled agents uh is the stepping stone to where we will ultimately be when we have, you know, more of like, oh my gosh, this is artificial general intelligence.
00:07:40.800 --> 00:07:47.600
We have AGI because I give uh an agent a goal and it doesn't need me for anything, right?
00:07:47.680 --> 00:07:48.399
We're not there yet.
00:07:48.480 --> 00:07:50.000
So we are in this in-between phase.
00:07:50.079 --> 00:07:54.639
I don't know if this phase is gonna last for a couple of quarters, a couple of years.
00:07:54.800 --> 00:07:55.519
I'm not sure.
00:07:55.680 --> 00:07:59.439
But it has definitely taken shape so quickly over the last few weeks.
00:07:59.600 --> 00:08:02.000
And that's led to kind of, again, this feature benefit.
00:08:02.160 --> 00:08:15.600
Because when I think of traditional large language models, right, uh essentially once once companies understood their utility, right, the immediate benefit was oh, more time, productivity, right?
00:08:15.680 --> 00:08:19.759
We can do more, uh, you know, do do more or save time.
00:08:19.920 --> 00:08:23.839
Um, but what about for the actual agents, right?
00:08:24.240 --> 00:08:33.759
I think when we thought about AI in the future versus benefits kind of paradox, we thought about the benefit on the human, but what about the benefit on the AI system?
00:08:33.919 --> 00:08:40.000
Because as they start to get agentic and more human-like and how they can work, well, they start to benefit as well.
00:08:40.159 --> 00:08:44.799
And that benefit is the agentic context carry that we're talking about today.
00:08:44.879 --> 00:08:46.320
And this is huge.
00:08:46.559 --> 00:08:55.600
And then, like I said, just in the past 24 hours, right, we had Google launch their Gemini Enterprise Agent platform and OpenAI launched their workspace agents inside of ChatGPT.
00:08:55.679 --> 00:09:00.399
So we're gonna be going over that a little bit more on tomorrow's uh show and FYI.
00:09:00.480 --> 00:09:06.960
We did kind of go over some good examples of this context carry uh on yesterday's show on codex.
00:09:07.200 --> 00:09:07.519
All right.
00:09:07.600 --> 00:09:10.799
So uh if you missed that one, yeah, I'm gonna plug both of these shows.
00:09:10.960 --> 00:09:21.360
Uh so make sure to go listen to the codecs uh kind of super app preview 762, and then make sure to join us tomorrow more on these two recent launches.
00:09:21.600 --> 00:09:32.879
But here's a little bit on what the ChatGBT schedule of agents can do, just so we can kind of set the stage for why this context carry is extremely important.
00:09:33.200 --> 00:09:39.759
All right, so how OpenAI says it in their recently released blog posts, they say build once, scale across your team, right?
00:09:39.919 --> 00:09:42.000
Create an agent once, share it with your team.
00:09:42.080 --> 00:09:50.799
Uh, work that runs itself, so you can run agents on schedules to handle recurring tasks, and then keep the work moving across tools.
00:09:50.960 --> 00:10:00.639
All right, so you can uh the agents use your tools to gather information, take action uh without needing step-by-step guidance.
00:10:01.919 --> 00:10:19.840
So now, yeah, I had to do a little wind up here because I wanted everyone to really understand how big this is and how quickly it's happening before I really unwrapped this kind of um concept that I coined, right?
00:10:20.000 --> 00:10:25.279
Um, of agentic uh context carry.
00:10:25.360 --> 00:10:29.120
Yeah, it's it's so new, even even um uh you know, I don't just want to say sack, right?
00:10:29.200 --> 00:10:31.120
But scheduled agentic context carry.
00:10:31.279 --> 00:10:32.559
So this means, right?
00:10:32.639 --> 00:10:37.120
I want to break down each of the four words and how they work together.
00:10:37.279 --> 00:10:43.759
So scheduled, obviously, means that the agent wakes up or runs on a cadence, not only when prompted.
00:10:43.919 --> 00:10:52.799
So that can be both a time cadence, like we just talked about in the ChatGPT agents, or as an example in Claude routines, it can be a trigger, right?
00:10:52.879 --> 00:10:56.240
When you get a certain type of email, then an agent is going to run.
00:10:56.320 --> 00:10:58.240
All right, that's what scheduled means.
00:10:58.399 --> 00:11:00.399
Uh, next, context carry.
00:11:00.559 --> 00:11:13.600
That means your memory, either your personal memory and preferences, your company's data, all of those things, dynamic data pipelines, uh, tool access, all those things that persist between runs.
00:11:13.840 --> 00:11:14.320
All right.
00:11:14.559 --> 00:11:17.120
And that is the big piece there.
00:11:17.279 --> 00:11:20.320
That is the context and the carry.
00:11:20.480 --> 00:11:24.879
And obviously, these agentic models, all by default, are able to do this.
00:11:25.039 --> 00:11:25.679
Uh, right.
00:11:25.840 --> 00:11:29.679
So the models themselves, they can call tools, right?
00:11:29.759 --> 00:11:39.200
They can, you know, call on you know these connected apps on these MCP servers that you can bring in, you know, thousands of different apps that you use.
00:11:40.639 --> 00:11:44.159
But the actual carry, that's what's important.
00:11:44.720 --> 00:11:52.320
Because as we've gotten these new context windows, which I'm going to get into a little bit more, that's what makes this all possible.
00:11:52.559 --> 00:11:59.679
And the ability now for an agent to go out and learn something, right, about you or your company without you having to teach it.
00:11:59.840 --> 00:12:03.039
So let's just say you have a scheduled agent, right?
00:12:03.120 --> 00:12:06.159
You give it information about your company, your company's goals.
00:12:06.320 --> 00:12:11.360
Maybe you're looking to acquire uh, you know, a new client or a new customer.
00:12:11.440 --> 00:12:14.320
Uh, but the industry, whatever industry you're working in is moving fast.
00:12:14.559 --> 00:12:24.080
Let's say you have an agent that goes out, you know, every Sunday night, uh, it pulls up the in the industry's most recent white papers, uh, industry news, etc.
00:12:24.480 --> 00:12:35.519
And oh, all of a sudden, when you didn't know it, it found out that you have a you know a huge new potential um, you know, buyer moving in into your state that wasn't there before.
00:12:35.679 --> 00:12:36.159
Um, right.
00:12:36.480 --> 00:12:40.639
And the reason that this can happen is because it's able to carry the context with you.
00:12:40.799 --> 00:12:51.519
All of those documents that you share, preferences, the memory of your uh you know, recent chats, but also that can run in essentially the same context window over and over.
00:12:51.679 --> 00:13:01.440
So not only can it carry in the context that you give it uh according to your information, uh, but also the persistent memory of that actual conversation.
00:13:01.600 --> 00:13:03.039
So it's going to know, right?
00:13:03.120 --> 00:13:08.639
If you have uh a run that goes every single day, it is going to carry that trend line with it.
00:13:08.799 --> 00:13:21.200
So it does start to turn into, oh, you know, it's like when you hire a junior employee, after a couple, you know, days on the job, they kind of start to get it after a couple of weeks on the job, you're like, okay, it's picking up momentum.
00:13:21.360 --> 00:13:22.159
The same thing.
00:13:22.240 --> 00:13:26.480
And that's uh, I think why this is a very exciting time in AI.
00:13:26.639 --> 00:13:33.519
And this, I think, is that bridge between, you know, uh the simple chatbots to the fully autonomous agent.
00:13:33.679 --> 00:13:47.519
Because as much as every you know, open claw aficionado wants you to believe, we are not yet at the point where uh we have true autonomy in agents, right?
00:13:47.600 --> 00:13:54.720
Where you give them a goal and they can safely go execute that goal without constant human intervention.
00:13:54.879 --> 00:13:56.159
Uh, is it possible?
00:13:56.399 --> 00:13:57.279
Sure, right?
00:13:57.360 --> 00:14:07.600
If you have a very well-defined goal, if you have strict guardrails, um, and if you are using it in a narrow capacity, I don't think we have autonomous general agents.
00:14:07.679 --> 00:14:18.159
Um, I think we have autonomous narrow agents that can do one very, very simple uh task if you give it, or a goal if it's very, very specific, right?
00:14:18.320 --> 00:14:20.799
But what happens if the guardrails are changed?
00:14:20.879 --> 00:14:22.240
What happens if the industry is changed?
00:14:22.320 --> 00:14:23.919
What happens if your data is corrupted, right?
00:14:24.000 --> 00:14:29.120
An autonomous agent would, in theory, be able to uh figure those things out.
00:14:29.279 --> 00:14:30.320
We don't have that right now.
00:14:30.399 --> 00:14:35.919
And I think that this uh agenconte context query is that stepping stone that's gonna help us get there.
00:14:36.240 --> 00:14:40.080
I I've also um you know talked about this a little bit before.
00:14:40.240 --> 00:14:43.519
Previously, I'd called it kind of like, you know, the human AI duct tape.
00:14:43.600 --> 00:14:47.440
It's all those intermediate steps in between that a human had to do, right?
00:14:47.600 --> 00:14:53.279
If you run something in, you know, deep research uh inside ChatGPT, well, now I have to copy that.
00:14:53.440 --> 00:14:58.240
I have to go put it in a doc, and then I have to upload it to this project folder as an example.
00:14:58.559 --> 00:15:12.799
That is where this context carry and the larger context windows starts to erase um all of that manual, you know, human AI duct tape that, you know, those steps that us humans working with multiple AI systems would have to continually make.
00:15:13.039 --> 00:15:17.519
Because now these agents also have write ability, right?
00:15:17.679 --> 00:15:26.720
Whereas before, you know, three to six months ago, they didn't have the ability to write to your Google Docs, they didn't have the ability to send Gmails, right?
00:15:26.960 --> 00:15:28.480
Now they do, right?
00:15:28.559 --> 00:15:33.440
If you give them the permissions and if you're feeling you know spicy and you want to roll the dice.
00:15:33.679 --> 00:15:41.360
Uh, but that agentic context carry layers the schedule and the memory over that, but no one is really talking about this.
00:15:41.440 --> 00:15:44.879
I don't know, maybe maybe I'm too dorky and excited about where we are.
00:15:45.039 --> 00:15:53.759
Um, but the reality is I think that there's been this narrative right now.
00:15:53.919 --> 00:15:59.519
And um, I'm actually gonna call out a recent uh tweet uh I saw or in X.
00:15:59.840 --> 00:16:02.080
What do you call what do you call a tweet on X anymore?
00:16:02.159 --> 00:16:02.399
I don't know.
00:16:02.480 --> 00:16:03.840
It's just why I call it Twitter.
00:16:04.000 --> 00:16:05.840
Uh you can't verb X.
00:16:06.000 --> 00:16:10.080
Uh, but there's a recent viral uh tweet out on Twitter, uh, right?
00:16:10.159 --> 00:16:14.639
So ChatGPT had a recent integration with Starbucks, right?
00:16:14.720 --> 00:16:20.559
And someone said, oh, you know, why are all of these apps, right, why do they exist?
00:16:20.720 --> 00:16:22.080
It doesn't make sense, right?
00:16:22.240 --> 00:16:30.639
Because it's gonna take me, you know, two minutes at the absolute fastest to make an order on this Chat GPT Starbucks app integration.
00:16:30.799 --> 00:16:32.240
I'm just using this as an example.
00:16:32.480 --> 00:16:44.799
Throw in any uh you know business app that you're using inside or you know, connector that you're using inside Gemini Copilot uh Claude or ChatGPT, but this kind of viral incident with Starbucks, right?
00:16:44.879 --> 00:16:55.360
It's like, okay, well, it takes two minutes to order it via the the the app and if uh the chat gpt app, but if I go into the actual Starbucks app, I can do it 20 seconds, right?
00:16:55.440 --> 00:16:57.120
So this is done, right?
00:16:57.360 --> 00:17:07.119
Um, but I think people are missing the point because it's not just about one app, it's not just about chat GPT interfacing with one app.
00:17:07.359 --> 00:17:18.319
Because in the new agent builder, right, as an example, the brand new uh agent builder, you can go to create an agent all by hand.
00:17:18.480 --> 00:17:20.960
I'm literally clicking around as I do this now.
00:17:21.440 --> 00:17:25.200
You can connect 20 apps, right?
00:17:25.440 --> 00:17:33.680
Your Gmail, your Slack, your Notion, your Teams, your Outlook email, your Google Calendar, Google Drive, whatever, right?
00:17:33.920 --> 00:17:36.000
MCP servers, you can do all those things.
00:17:36.079 --> 00:17:41.680
Uh, you can connect agent skills, you can upload files, you can manage the memory, right?
00:17:41.839 --> 00:17:52.960
So it's not just about, oh my gosh, you know, using a single app to do a task is so much slower than it is to just do it individually in that platform or on that website.
00:17:53.119 --> 00:17:54.000
That's not what it is.
00:17:54.079 --> 00:17:56.559
It's about eliminating that human AI duct tape.
00:17:56.720 --> 00:18:01.279
It's about, you know, the the 30 small human steps in between that are required.
00:18:01.440 --> 00:18:03.119
That is the context carrying.
00:18:03.359 --> 00:18:08.880
Us humans have been the one carrying the context because AI agents didn't have the ability.
00:18:08.960 --> 00:18:12.720
They didn't have the um the tools to do that.
00:18:13.279 --> 00:18:14.319
Now they do.
00:18:14.480 --> 00:18:15.200
That's the thing.
00:18:15.359 --> 00:18:28.240
Yes, I can much more quickly go open my Gmail, read an email, and respond to it uh than uh a connection in Gemini, ChatGPT, Quad, et cetera, right?
00:18:28.400 --> 00:18:33.359
But what about when there's a Google Doc that goes with it?
00:18:33.519 --> 00:18:34.960
I have to look in my calendar.
00:18:35.200 --> 00:18:39.279
Oh, there's actually three or four different emails, um, right?
00:18:39.440 --> 00:18:42.400
Oh, there's that file in my drive.
00:18:42.480 --> 00:18:45.119
There's a Slaughter conversation about that, right?
00:18:45.279 --> 00:18:53.519
Now all of a sudden, yes, it might be quicker to do all of those small tasks individually in those apps or on those websites.
00:18:53.680 --> 00:18:59.039
But when you have to carry the context yourself manually as the human, that's where you can start, right?
00:18:59.279 --> 00:19:10.880
This has essentially been the uh mundane nature of knowledge work in front of a computer over the past 20 years, you know, as SaaS and applications have exploded.
00:19:11.039 --> 00:19:12.400
But that's what we do.
00:19:12.640 --> 00:19:29.759
And that is where the true benefit of now agentic context carry because us humans no longer have to do the duct tape and have to remember and have to bring that context from app A to app B to app C to storage D to messaging platform E and F.
00:19:30.720 --> 00:19:34.000
The agent does it all for us in one swoop.
00:19:34.160 --> 00:19:49.119
So, yes, it might take you five times as long to accomplish a goal inside of an AI agent, but that's not counting the human error, the human lookup, the human retrieval that has to happen every single step of the way in between.
00:19:49.359 --> 00:19:52.079
That's the big unlock here, y'all.
00:19:52.319 --> 00:19:56.319
Uh, but also, I don't know, I start to forget things fairly quickly.
00:19:56.480 --> 00:19:57.440
Maybe it's just me.
00:19:57.599 --> 00:20:03.359
Like I literally use the you know this concept of agentic context carry all the time, right?
00:20:03.519 --> 00:20:08.720
I was actually walking, uh, I was actually walking to uh my office, right?
00:20:08.799 --> 00:20:12.480
Sometimes I record um, you know, from my kind of home office.
00:20:12.559 --> 00:20:15.359
Sometimes I uh, you know, record from my actual office.
00:20:15.440 --> 00:20:18.799
And you know, I'm working on a cool partnership here uh with a group Sage.
00:20:19.200 --> 00:20:28.960
And I had a couple different email threads with with travel, there was some Google Docs, there was all these things, and I'm like, ah, which, you know, and I have multiple emails, right?
00:20:29.119 --> 00:20:36.160
I have multiple email accounts, certain forms go different places, and I'm like, my gosh, like this is gonna take me a long time.
00:20:36.400 --> 00:20:40.319
Instead, right, just use in this instance, use Claude.
00:20:40.480 --> 00:20:42.000
It went and carried that context.
00:20:42.079 --> 00:20:44.640
But what about when you can schedule those things, right?
00:20:44.880 --> 00:20:59.200
Uh, and to say, hey, every day at you know, 2 a.m., I want you to go through my my email, my calendar, uh, notion, Slack, all of these things.
00:20:59.279 --> 00:21:06.720
Yeah, it might take the agent longer to do that if than if you were to, but it's gonna do it on its own schedule.
00:21:07.119 --> 00:21:09.680
And it's gonna carry the context from app to app.
00:21:09.759 --> 00:21:11.759
So that's where the new breakthrough comes.
00:21:11.920 --> 00:21:15.200
It's the capabilities that have made the cross-app technology possible.
00:21:15.359 --> 00:21:19.519
So here's where the unlock and the timing all comes into play, right?
00:21:19.680 --> 00:21:22.319
Um, I love Venn diagrams, right?
00:21:22.559 --> 00:21:30.799
This is where it's kind of the capabilities and the technology and the need have all overlapped with this perfect timing.
00:21:30.960 --> 00:21:40.240
So this is, you know, if you think of like cowork or agentic scheduling, the features and then the context window all coming together and exploding at the same time, right?
00:21:40.319 --> 00:21:43.359
So Claude and Frapic has really led the way in this.
00:21:43.519 --> 00:21:48.559
So now they have that 1 million token context window by default, right?
00:21:48.720 --> 00:21:59.759
Codecs a little bit uh, you know, behind, although there is an experimented uh experimental uh 1 million token context window uh in the command line interface, but on the app, I believe it's 250.
00:22:01.680 --> 00:22:02.400
So what does that mean?
00:22:02.480 --> 00:22:02.720
Right.
00:22:02.799 --> 00:22:11.359
If you're not too technical, that just means, right, the free version of Chat GPT, last time I checked, uh I didn't check the free version in a while.
00:22:11.519 --> 00:22:15.839
But let's just say in 2025, it was about 8,000 token context window, right?
00:22:16.000 --> 00:22:17.839
So now you're looking at 1 million.
00:22:17.920 --> 00:22:19.039
So do the math there.
00:22:19.119 --> 00:22:22.559
Or you know, or you know, going to 258,000.
00:22:22.720 --> 00:22:31.839
Essentially, now AI models and AI agents can remember things over a very longer period of time, right?
00:22:32.000 --> 00:22:34.960
Whereas before, they essentially had very short-term memory.
00:22:35.119 --> 00:22:42.480
You would, you know, especially if you were on a free plan or, you know, early in 2024, 2025, AI models forgot things very quickly.
00:22:42.640 --> 00:22:45.200
So especially when it came to handling your data.
00:22:45.359 --> 00:22:48.000
So if you upload a file, you're working with it, right?
00:22:48.079 --> 00:22:55.599
Maybe updating a job description and doing some research on recent law changes to make sure that your you know job description reflects those or something like that, right?
00:22:55.680 --> 00:22:59.039
And it's going well, and all of a sudden it's, oh, wait, it's dumb, right?
00:22:59.200 --> 00:23:00.720
That's because it ran over the context widow.
00:23:00.880 --> 00:23:11.519
Context widows used to be very small, but now as they become bigger and bigger and bigger, right, it's essentially you're working with an AI model that has a bigger brain that's able to carry the conversation for longer.
00:23:12.480 --> 00:23:17.200
So now you know this kind of trending concept that's happening.
00:23:17.440 --> 00:23:23.200
It's not just going in and working with one app or one connector, right?
00:23:23.359 --> 00:23:41.200
It's bringing in all of the different tech stack that you have to use on a daily basis, that your company has to use on a daily basis, eliminating all of those manual steps in between, because the reality is, just like a large language model, us humans, we have a context window as well.
00:23:41.440 --> 00:23:43.039
How much time do you spend, right?
00:23:43.119 --> 00:23:45.359
Even pre-AI, it was obviously way worse.
00:23:45.599 --> 00:24:05.440
But sometimes you spend just as much time trying to either track, remember, or find certain information where it lives within your kind of SaaS database as it actually takes to create that new business value once you do find it, or reply to a certain email, or to finish a certain deck or a project or fill out a spreadsheet, right?
00:24:05.519 --> 00:24:10.240
Sometimes you spend as much time just trying to retrieve that information.
00:24:10.480 --> 00:24:17.920
So that's where the multiple apps, because the big context window and the new agentic capabilities, those three things coming together come to play.
00:24:18.079 --> 00:24:25.839
So now that you know it's here and you know that this I think is the intermediate step stepping stone until we have those fully autonomous agents.
00:24:26.000 --> 00:24:31.920
You need to take advantage of this scheduled agentic context carry sack, right?
00:24:32.319 --> 00:24:35.359
Here's how step one, three steps, ready?
00:24:35.440 --> 00:24:36.480
I'm gonna go quick.
00:24:36.640 --> 00:24:40.799
Uh, connect your live data sources and your preferences first.
00:24:41.200 --> 00:24:42.319
Make sure to do this.
00:24:42.480 --> 00:24:46.160
And you know, I do have to you know put out the normal disclaimer, right?
00:24:46.400 --> 00:24:48.880
The uh responsible AI person I am.
00:24:48.960 --> 00:24:49.359
Uh, right.
00:24:49.599 --> 00:24:50.640
I'm a business owner.
00:24:50.799 --> 00:24:54.480
I decide if this is safe for my organization, but you need to do the same, right?
00:24:54.640 --> 00:24:57.519
You shouldn't be doing this with shadow AI tools.
00:24:57.599 --> 00:25:00.319
Uh, you need to be doing this with approved tools.
00:25:00.480 --> 00:25:02.240
So make sure you go through the proper channels.
00:25:02.319 --> 00:25:07.599
But let's just say you have, you know, Claude approved or you have ChatGPT approved, whatever it is, right?
00:25:07.680 --> 00:25:10.880
Uh, and you have these connectors or apps approved as well.
00:25:11.039 --> 00:25:11.279
All right.
00:25:11.359 --> 00:25:21.119
So you need to authorize those live connectors to your, as an example, your email, your calendar, your Slack, your drive, all of those important things, your CRM, right?
00:25:21.200 --> 00:25:21.839
It's huge.
00:25:22.000 --> 00:25:30.480
Then you need to understand how each system's computer use and access works, and then ensure your custom instructions and memory are updated accordingly.
00:25:30.640 --> 00:25:36.160
So, first, you have to get your data sources, uh, your preferences, and your memory in line.
00:25:36.240 --> 00:25:40.960
Because when you talk about context theory, well, context is the base, right?
00:25:41.119 --> 00:25:45.519
We did an earlier show in the start here series on the importance of context engineering.
00:25:45.599 --> 00:25:47.920
So make sure you go back and listen to that one as well.
00:25:48.079 --> 00:25:51.359
And then also, all these platforms, they support MCPs.
00:25:51.519 --> 00:26:02.720
So even if your uh, you know, your app of choice, whatever you're using, doesn't um have any of the, you know, oh, it doesn't have an app connection to ChatGPT or it doesn't have an app connection to Claude.
00:26:02.799 --> 00:26:06.960
Well, chances are you can just use a an MCP server and get that cooked up right away.
00:26:07.119 --> 00:26:08.240
So that's step one.
00:26:08.319 --> 00:26:13.039
Step two, you need to context stuff in a dedicated memory thread.
00:26:13.200 --> 00:26:17.759
Here's a little, I wouldn't necessarily call this a cheat code per se.
00:26:17.920 --> 00:26:20.079
And this is much different, right?
00:26:20.160 --> 00:26:29.680
I I I've obviously talked taught uh you know this concept of of prime prompt polish, you know, the basics of prompt engineering 101 with the context window.
00:26:29.759 --> 00:26:33.920
It doesn't throw away those best practices, but it does kind of change what can get done.
00:26:34.079 --> 00:26:41.440
So here's a little uh little cheat sheet one for you for listening to this uh episode now for 26 minutes and running.
00:26:41.599 --> 00:26:49.279
Uh these systems now, the context windows are enormous.
00:26:49.440 --> 00:27:01.599
You can work on it in theory for a very depends on what tools you're calling, but you know, in Claude Code as an example, uh, you know, running your routines all in the same thread, a daily schedule, it's gonna hold, right?
00:27:01.680 --> 00:27:11.359
I have some that run every single day that started uh when it first came out, you know, two-ish weeks ago, and they're not even close to hitting uh the context window.
00:27:11.599 --> 00:27:11.839
Right.
00:27:12.000 --> 00:27:18.799
So at a million tokens, the agent can hold just weeks of regular usage working memory at once.
00:27:18.960 --> 00:27:25.039
So here's what I like to do connect everything to one thread, right?
00:27:25.119 --> 00:27:27.440
They these all work a little differently, right?
00:27:27.599 --> 00:27:38.720
Uh you know, codex works a little bit differently than Claude Code works a little bit differently than these brand new, you know, ancient builders essentially that we got from uh from ChatGPT and from Google Gemini.
00:27:38.880 --> 00:27:46.240
But essentially, uh, if you do connect things on a thread by thread or a folder-by-folder basis, have one where you just context stuff, right?
00:27:46.400 --> 00:27:48.400
Put all your context in there at once.
00:27:48.480 --> 00:27:50.240
And then that can be your daily drive.
00:27:50.400 --> 00:27:57.519
Because the good thing is, is then also if you need to take it into a different direction, you can just fork that threat at that point, right?
00:27:57.599 --> 00:28:15.599
So you at least have this unified base where you can every single day, right, have it be the one that brings you your morning triage, the one for your you know, most common day-to-day tasks, but aren't necessarily specifically project-based that require a lot of different uh directional feedback, uh, et cetera.
00:28:15.680 --> 00:28:15.920
Right.
00:28:16.079 --> 00:28:19.599
So from that, you can iterate on the reasoning until it matches your standard.
00:28:19.680 --> 00:28:20.400
That's the big thing.
00:28:20.559 --> 00:28:21.440
You need to context.
00:28:21.599 --> 00:28:25.039
Step two is technically context stuff and iterate as well.
00:28:25.200 --> 00:28:27.759
Well, iterate will actually get a little bit more into step three.
00:28:27.920 --> 00:28:28.160
Sorry.
00:28:28.319 --> 00:28:29.920
I I jumped ahead of myself.
00:28:30.000 --> 00:28:35.119
So step two, context stuff and uh context stuff in a dedicated memory thread.
00:28:35.200 --> 00:28:38.480
And then step three, iterate with chain of thought.
00:28:38.720 --> 00:28:41.359
If you listen to the show at all, you know how important this is.
00:28:41.519 --> 00:28:46.480
You saw this in my little demo uh that I did yesterday on codecs, right?
00:28:46.720 --> 00:28:55.599
You need to understand um how these models work because they are generative, they are not deterministic, they're gonna work slightly different each time.
00:28:55.680 --> 00:28:58.000
So you can't just run something once, right?
00:28:58.160 --> 00:29:04.000
Especially as these uh the capabilities become uh greater and greater, right?
00:29:04.160 --> 00:29:06.240
A lot, I I see a common mistake a lot.
00:29:06.400 --> 00:29:09.440
People will run something once and they're like, oh yeah, this is great.
00:29:09.519 --> 00:29:10.640
Let's put it out in production.
00:29:10.799 --> 00:29:16.319
Okay, well, that's could be dangerous, especially if you're doing something uh public-facing or client-facing.
00:29:16.400 --> 00:29:21.039
You probably wouldn't want to do that just yet, uh, right, because there's always going to be edge cases.
00:29:21.200 --> 00:29:34.799
You can run the same scheduled run every single day for seven days, and it might two of the days it might call the tool that you didn't want it to, or uh maybe it's uh not calling a tool that you are telling it to.
00:29:34.960 --> 00:29:38.480
So you really do have to review the chain of thought and iterate.
00:29:38.559 --> 00:29:46.559
So what that means, most systems, you can kind of have some level of observability, traceability uh as they go by looking at the system.
00:29:46.720 --> 00:29:52.079
So if they're scheduled, you can go, you know, usually you might click on, you know, might say, oh, thought for one hour.
00:29:52.160 --> 00:30:00.960
And you can click that and then see every single tool, every single step, and kind of get how the uh that schedule would run or that co-work session, how it worked.
00:30:01.039 --> 00:30:03.039
And you can kind of trace it the same way, right?
00:30:03.119 --> 00:30:05.839
When you think of like Bath, right, where you had to show your work.
00:30:05.920 --> 00:30:09.039
I don't understand like the new common core bath stuff, right?
00:30:09.200 --> 00:30:14.640
But back in my day, right, we just, I don't know, wrote down the the numbers in a column, right?
00:30:14.720 --> 00:30:15.759
But you had to show your work.
00:30:15.920 --> 00:30:24.160
So you should always be checking the work of your, you know, co-working run of your scheduled agent task and then iterating it.
00:30:24.240 --> 00:30:36.160
You need to uh refine the prompt, make it better, and then once it is kind of quote unquote ready for production and you've built those guardrails in place, that's when you can save it as a routine or a scheduled automation.
00:30:36.240 --> 00:30:39.039
Then what's refined, that is that hidden workflow.
00:30:39.119 --> 00:30:43.599
So it is those three steps that really allow that agentic context carry.
00:30:43.759 --> 00:30:44.559
Again, faster.
00:30:44.799 --> 00:30:48.000
Step one, connect your live data sources and preferences first.
00:30:48.240 --> 00:30:51.200
Step two, context stuff in a dedicated memory thread.
00:30:51.359 --> 00:30:56.799
And then step three, iterate with chain of thought reasoning before you put it out into production.
00:30:56.960 --> 00:31:04.079
But then schedule that thing and take advantage of this stepping stone that I think is going to be huge.
00:31:04.240 --> 00:31:06.160
And the time is now.
00:31:06.480 --> 00:31:09.440
So, like I said before, this is not the final destination.
00:31:09.599 --> 00:31:15.599
I think that truly autonomous agents with persistent memory are the next big deal, but we that could be far off.
00:31:15.680 --> 00:31:16.000
Who knows?
00:31:16.079 --> 00:31:28.880
I mean, maybe we'll have that, you know, next month, but it could still be another year, two years or more until we actually see autonomous agents that you can give them a goal and they don't really require much else.
00:31:29.200 --> 00:31:32.079
This is the now or the next.
00:31:32.240 --> 00:31:37.920
So understand um and really push this agentic context carry.
00:31:38.000 --> 00:31:49.359
The leaders pulling ahead this quarter are building scheduled context on autopilot, not just you know, better one-off prompts, not just you know, sharing skills within your organization.
00:31:49.519 --> 00:31:53.519
That's no longer enough, uh, right, to really be pushing in your space.
00:31:53.680 --> 00:32:08.240
So pick one recurring task this week, take it through those three steps, and then deploy your first kind of hidden workflow inside of it that's taking advantage of scheduled agentic context carry.
00:32:08.400 --> 00:32:10.079
I hope this was helpful, y'all.
00:32:10.240 --> 00:32:13.759
If it was, please go to starthearseries.com.
00:32:13.920 --> 00:32:22.880
That's gonna take you straight to a signup form to get access to our community for free, the everyday AI inner circle.
00:32:22.960 --> 00:32:36.079
And then in the Start Here Series space, you can go find every single Start Here Series podcast, read every single Start Here Series newsletter, all in one space, connect and network with others who are doing the same.
00:32:36.240 --> 00:32:36.559
All right.
00:32:36.720 --> 00:32:37.759
I hope this was helpful.
00:32:37.920 --> 00:32:38.720
Thanks for tuning in.
00:32:38.880 --> 00:32:42.640
Hope to see you back tomorrow and every day for more everyday AI.
00:32:42.720 --> 00:32:43.759
Thanks, y'all.
00:00:02.160 --> 00:00:16.079
In the past 24 hours, two of the biggest players in the AI space, OpenAI and Google, both launched updated versions of their simple drag and drop agents that work for you with your context around the clock.
00:00:16.239 --> 00:00:21.679
And that got me thinking of the common features versus benefits methodology when it comes to marketing.
00:00:21.839 --> 00:00:23.679
If you've never heard of it, it's pretty simple.
00:00:23.920 --> 00:00:28.000
Features describe the technical facts or the specs of what a product does.
00:00:28.079 --> 00:00:34.479
And well, the benefits explain the personal value the human gets from using said products.
00:00:34.640 --> 00:00:41.039
And in AI, we've seen a similar features versus benefits narrative take shape over the past few years.
00:00:41.359 --> 00:00:46.719
The feature, large language models are smarter and faster than humans when used correctly.
00:00:46.960 --> 00:00:49.840
The benefit, humans can be more productive.
00:00:50.159 --> 00:00:54.880
But the feature side has completely exploded over the past two months.
00:00:55.039 --> 00:00:58.240
And the benefit side is still being written.
00:00:58.560 --> 00:00:59.679
Stick with me here.
00:00:59.920 --> 00:01:14.400
So as large language models have become overly agentic by default overnight and as capable as humans, there's a new benefit paradigm for AI agents that has flown completely under the radar.
00:01:14.719 --> 00:01:17.840
And I know that this is the next big trend coming.
00:01:17.920 --> 00:01:22.560
It doesn't have a name, but I'm going to go ahead and name it now and explain the concept.
00:01:22.640 --> 00:01:28.640
I'm calling this Scheduled Agentic Context Carry or SAC, S-A-C-C.
00:01:28.959 --> 00:01:39.920
And I think the companies taking the time to understand and iterate on this new concept now are going to be the ones crushing their year-end goals and KPIs in quarter four.
00:01:40.159 --> 00:01:51.680
So let's unwind this kind of new concept together because I think understanding this now is one of the most important investments you can make on your AI journey this year.
00:01:51.920 --> 00:01:55.040
So let's start there with our Start Here series.
00:01:55.200 --> 00:01:57.519
If you're new here, welcome to Everyday AI.
00:01:57.599 --> 00:02:00.319
And this is our here series.
00:02:00.560 --> 00:02:03.599
But let's first start with the big picture of what's going on.
00:02:03.840 --> 00:02:11.120
Anthropic, OpenAI, Microsoft, and Perplexity have all shipped scheduled agents just this spring.
00:02:11.439 --> 00:02:15.280
And most business leaders are still using AI just like a chatbot.
00:02:15.360 --> 00:02:24.800
I'm going to go in and I'm going to technically reactively ask an AI chatbot for something and probably have to re-explain a lot to waste a lot of time.
00:02:24.960 --> 00:02:31.120
But there's been a quiet workflow pattern emerging beneath every one of these recent product launches.
00:02:31.280 --> 00:02:37.120
And that's why we're talking about this new concept of scheduled agentic context carry.
00:02:37.919 --> 00:02:42.240
So on today's show, that's exactly what we're going to over, kind of go over.
00:02:42.319 --> 00:02:43.520
And here's what you're going to learn.
00:02:43.680 --> 00:02:51.280
You're going to learn what agentic context carry actually means and why you've never heard of it, and but why it absolutely matters.
00:02:51.439 --> 00:02:57.439
You're going to know how this hidden workflow bridges, chatbots, and the fully autonomous AI future.
00:02:57.599 --> 00:03:08.000
You're going to understand the timing of all these things coming together, and specifically these now huge 1 million token context windows that have quietly changed everything this spring.
00:03:08.159 --> 00:03:13.360
And you're going to know the exact three steps to deploy this pattern inside of your business today.
00:03:13.599 --> 00:03:16.080
All right, let's get started.
00:03:16.400 --> 00:03:17.759
Welcome to Everyday AI.
00:03:17.840 --> 00:03:20.639
My name is Jordan Wilson, and this is the Start Here series.
00:03:20.800 --> 00:03:26.319
So after 750 plus podcasts, I've never had an answer when someone was like, I'm new.
00:03:26.400 --> 00:03:27.199
Where do I start?
00:03:27.360 --> 00:03:28.560
Well, you start here.
00:03:28.879 --> 00:03:39.599
Our Start Here series is an ongoing effort to help business leaders both better understand trending and emerging concepts, but also for those who are brand new to get caught up.
00:03:39.759 --> 00:03:47.439
So I'd recommend you start with episode one of the Start Here Series and listen in order, but go ahead and listen to this one and then you can go backtracked.
00:03:47.520 --> 00:03:52.000
But make sure you go to starthere series.com because, well, it's going to make it much easier to do that.
00:03:52.159 --> 00:03:55.599
That's going to give you free access to our inner circle community.
00:03:55.759 --> 00:04:01.599
Right now, there's no other way for the general public to sign up except starthearseries.com.
00:04:01.840 --> 00:04:12.560
And then inside the Start Here Series space, there will be an updated Spotify playlist where you can go listen to all of the Start Here series very easily in order in a dedicated playlist.
00:04:12.879 --> 00:04:13.120
All right.
00:04:13.199 --> 00:04:25.120
And if you miss our last Start Here Series episode, we uh in volume 20, so this is volume 21, we talked about AI change management that works, five moves, the top five percent make.
00:04:25.360 --> 00:04:30.160
All right, but let's get into this concept of a gentic context carry.
00:04:30.319 --> 00:04:38.399
And y'all, every single major AI lab and the big third-party players launched something.
00:04:38.560 --> 00:04:41.680
So from the big four, right?
00:04:41.839 --> 00:04:50.720
So that's Anthropic, Microsoft, Google, OpenAI, and then even Perplexity and you know, open clock technically fall under this category.
00:04:50.879 --> 00:04:55.839
But literally, everyone launched something and it's all very timely.
00:04:55.920 --> 00:05:16.480
So I did mention just the past 24 hours with big announcements uh from Google Gemini at their Cloud Next conference, and then with OpenAI's new agents that we're going to talk about here in a minute, but also clawed code routines, uh, right, that can bring uh automated, scheduled agent runs to your guest stop.
00:05:16.800 --> 00:05:36.639
So it is kind of like the uh maybe this open claw movement that happened, you know, really in February and March of this year, actually kind of forced the hand of all of the big companies to say, okay, it seems like we essentially to oversimplify it, we need an AI agent that can run on a cron, right?
00:05:36.720 --> 00:05:43.360
Run on a schedule, where someone can go in and they say, Hey, AI agent, at this time, every single day, I want this to happen.
00:05:43.519 --> 00:05:48.000
Uh, so we had the Claude Code Routines, which I absolutely love.
00:05:48.079 --> 00:05:49.279
That runs on your desktop.
00:05:49.519 --> 00:05:59.360
Similarly, OpenAI in their codex platform just added schedule work and persistent memory two days after that claw uh code routines announcement in mid-April.
00:05:59.519 --> 00:06:05.040
And then we also just got wind that co-pilot co-work uh officially launched in Frontier.
00:06:05.199 --> 00:06:12.160
So you have essentially all these scheduled uh agent platforms slash co-working platforms, right?
00:06:12.240 --> 00:06:14.800
So, like Claude Co-Work uh is the big one.
00:06:14.879 --> 00:06:21.920
Uh Microsoft Copilot uses essentially the Claud Co-Work technology because they are an investor in Anthropic.
00:06:22.000 --> 00:06:25.040
Uh so you have those kind of two places come together.
00:06:25.120 --> 00:06:30.879
You have these co-working uh kind of elements that allow for scheduling and it brings all your context.
00:06:31.040 --> 00:06:35.759
And then you have these scheduled agents, and it's all literally exploded out of nowhere.
00:06:35.920 --> 00:06:43.519
And although, you know, this may technically be a more timely um episode with all of these things happening now.
00:06:43.680 --> 00:07:02.240
The reason why I'm doing it in the start here series, whether you are listening to this in April or you're listening to it, I don't know, in the year 2027, is because I think that this is going to be a noticeable pivot in how the enterprise starts to interface with AI agents.
00:07:02.399 --> 00:07:03.839
Because here's the reality, right?
00:07:03.920 --> 00:07:08.000
We've been hearing since probably late 2024 that, oh, it's the year of AI agents.
00:07:08.160 --> 00:07:09.279
And it didn't happen.
00:07:09.360 --> 00:07:11.199
And in 2025, it didn't happen.
00:07:11.360 --> 00:07:28.639
But I think we've now come to that realization in 2026 in a certain way, because we've noticed that the fully autonomous uh AI agents, where you just give them a goal and then they go off and run on their own, not as reliable as we'd like, mainly because of safety concerns, guardrails, et cetera.
00:07:28.800 --> 00:07:40.720
So I think that this new kind of co-working scheduled agents uh is the stepping stone to where we will ultimately be when we have, you know, more of like, oh my gosh, this is artificial general intelligence.
00:07:40.800 --> 00:07:47.600
We have AGI because I give uh an agent a goal and it doesn't need me for anything, right?
00:07:47.680 --> 00:07:48.399
We're not there yet.
00:07:48.480 --> 00:07:50.000
So we are in this in-between phase.
00:07:50.079 --> 00:07:54.639
I don't know if this phase is gonna last for a couple of quarters, a couple of years.
00:07:54.800 --> 00:07:55.519
I'm not sure.
00:07:55.680 --> 00:07:59.439
But it has definitely taken shape so quickly over the last few weeks.
00:07:59.600 --> 00:08:02.000
And that's led to kind of, again, this feature benefit.
00:08:02.160 --> 00:08:15.600
Because when I think of traditional large language models, right, uh essentially once once companies understood their utility, right, the immediate benefit was oh, more time, productivity, right?
00:08:15.680 --> 00:08:19.759
We can do more, uh, you know, do do more or save time.
00:08:19.920 --> 00:08:23.839
Um, but what about for the actual agents, right?
00:08:24.240 --> 00:08:33.759
I think when we thought about AI in the future versus benefits kind of paradox, we thought about the benefit on the human, but what about the benefit on the AI system?
00:08:33.919 --> 00:08:40.000
Because as they start to get agentic and more human-like and how they can work, well, they start to benefit as well.
00:08:40.159 --> 00:08:44.799
And that benefit is the agentic context carry that we're talking about today.
00:08:44.879 --> 00:08:46.320
And this is huge.
00:08:46.559 --> 00:08:55.600
And then, like I said, just in the past 24 hours, right, we had Google launch their Gemini Enterprise Agent platform and OpenAI launched their workspace agents inside of ChatGPT.
00:08:55.679 --> 00:09:00.399
So we're gonna be going over that a little bit more on tomorrow's uh show and FYI.
00:09:00.480 --> 00:09:06.960
We did kind of go over some good examples of this context carry uh on yesterday's show on codex.
00:09:07.200 --> 00:09:07.519
All right.
00:09:07.600 --> 00:09:10.799
So uh if you missed that one, yeah, I'm gonna plug both of these shows.
00:09:10.960 --> 00:09:21.360
Uh so make sure to go listen to the codecs uh kind of super app preview 762, and then make sure to join us tomorrow more on these two recent launches.
00:09:21.600 --> 00:09:32.879
But here's a little bit on what the ChatGBT schedule of agents can do, just so we can kind of set the stage for why this context carry is extremely important.
00:09:33.200 --> 00:09:39.759
All right, so how OpenAI says it in their recently released blog posts, they say build once, scale across your team, right?
00:09:39.919 --> 00:09:42.000
Create an agent once, share it with your team.
00:09:42.080 --> 00:09:50.799
Uh, work that runs itself, so you can run agents on schedules to handle recurring tasks, and then keep the work moving across tools.
00:09:50.960 --> 00:10:00.639
All right, so you can uh the agents use your tools to gather information, take action uh without needing step-by-step guidance.
00:10:01.919 --> 00:10:19.840
So now, yeah, I had to do a little wind up here because I wanted everyone to really understand how big this is and how quickly it's happening before I really unwrapped this kind of um concept that I coined, right?
00:10:20.000 --> 00:10:25.279
Um, of agentic uh context carry.
00:10:25.360 --> 00:10:29.120
Yeah, it's it's so new, even even um uh you know, I don't just want to say sack, right?
00:10:29.200 --> 00:10:31.120
But scheduled agentic context carry.
00:10:31.279 --> 00:10:32.559
So this means, right?
00:10:32.639 --> 00:10:37.120
I want to break down each of the four words and how they work together.
00:10:37.279 --> 00:10:43.759
So scheduled, obviously, means that the agent wakes up or runs on a cadence, not only when prompted.
00:10:43.919 --> 00:10:52.799
So that can be both a time cadence, like we just talked about in the ChatGPT agents, or as an example in Claude routines, it can be a trigger, right?
00:10:52.879 --> 00:10:56.240
When you get a certain type of email, then an agent is going to run.
00:10:56.320 --> 00:10:58.240
All right, that's what scheduled means.
00:10:58.399 --> 00:11:00.399
Uh, next, context carry.
00:11:00.559 --> 00:11:13.600
That means your memory, either your personal memory and preferences, your company's data, all of those things, dynamic data pipelines, uh, tool access, all those things that persist between runs.
00:11:13.840 --> 00:11:14.320
All right.
00:11:14.559 --> 00:11:17.120
And that is the big piece there.
00:11:17.279 --> 00:11:20.320
That is the context and the carry.
00:11:20.480 --> 00:11:24.879
And obviously, these agentic models, all by default, are able to do this.
00:11:25.039 --> 00:11:25.679
Uh, right.
00:11:25.840 --> 00:11:29.679
So the models themselves, they can call tools, right?
00:11:29.759 --> 00:11:39.200
They can, you know, call on you know these connected apps on these MCP servers that you can bring in, you know, thousands of different apps that you use.
00:11:40.639 --> 00:11:44.159
But the actual carry, that's what's important.
00:11:44.720 --> 00:11:52.320
Because as we've gotten these new context windows, which I'm going to get into a little bit more, that's what makes this all possible.
00:11:52.559 --> 00:11:59.679
And the ability now for an agent to go out and learn something, right, about you or your company without you having to teach it.
00:11:59.840 --> 00:12:03.039
So let's just say you have a scheduled agent, right?
00:12:03.120 --> 00:12:06.159
You give it information about your company, your company's goals.
00:12:06.320 --> 00:12:11.360
Maybe you're looking to acquire uh, you know, a new client or a new customer.
00:12:11.440 --> 00:12:14.320
Uh, but the industry, whatever industry you're working in is moving fast.
00:12:14.559 --> 00:12:24.080
Let's say you have an agent that goes out, you know, every Sunday night, uh, it pulls up the in the industry's most recent white papers, uh, industry news, etc.
00:12:24.480 --> 00:12:35.519
And oh, all of a sudden, when you didn't know it, it found out that you have a you know a huge new potential um, you know, buyer moving in into your state that wasn't there before.
00:12:35.679 --> 00:12:36.159
Um, right.
00:12:36.480 --> 00:12:40.639
And the reason that this can happen is because it's able to carry the context with you.
00:12:40.799 --> 00:12:51.519
All of those documents that you share, preferences, the memory of your uh you know, recent chats, but also that can run in essentially the same context window over and over.
00:12:51.679 --> 00:13:01.440
So not only can it carry in the context that you give it uh according to your information, uh, but also the persistent memory of that actual conversation.
00:13:01.600 --> 00:13:03.039
So it's going to know, right?
00:13:03.120 --> 00:13:08.639
If you have uh a run that goes every single day, it is going to carry that trend line with it.
00:13:08.799 --> 00:13:21.200
So it does start to turn into, oh, you know, it's like when you hire a junior employee, after a couple, you know, days on the job, they kind of start to get it after a couple of weeks on the job, you're like, okay, it's picking up momentum.
00:13:21.360 --> 00:13:22.159
The same thing.
00:13:22.240 --> 00:13:26.480
And that's uh, I think why this is a very exciting time in AI.
00:13:26.639 --> 00:13:33.519
And this, I think, is that bridge between, you know, uh the simple chatbots to the fully autonomous agent.
00:13:33.679 --> 00:13:47.519
Because as much as every you know, open claw aficionado wants you to believe, we are not yet at the point where uh we have true autonomy in agents, right?
00:13:47.600 --> 00:13:54.720
Where you give them a goal and they can safely go execute that goal without constant human intervention.
00:13:54.879 --> 00:13:56.159
Uh, is it possible?
00:13:56.399 --> 00:13:57.279
Sure, right?
00:13:57.360 --> 00:14:07.600
If you have a very well-defined goal, if you have strict guardrails, um, and if you are using it in a narrow capacity, I don't think we have autonomous general agents.
00:14:07.679 --> 00:14:18.159
Um, I think we have autonomous narrow agents that can do one very, very simple uh task if you give it, or a goal if it's very, very specific, right?
00:14:18.320 --> 00:14:20.799
But what happens if the guardrails are changed?
00:14:20.879 --> 00:14:22.240
What happens if the industry is changed?
00:14:22.320 --> 00:14:23.919
What happens if your data is corrupted, right?
00:14:24.000 --> 00:14:29.120
An autonomous agent would, in theory, be able to uh figure those things out.
00:14:29.279 --> 00:14:30.320
We don't have that right now.
00:14:30.399 --> 00:14:35.919
And I think that this uh agenconte context query is that stepping stone that's gonna help us get there.
00:14:36.240 --> 00:14:40.080
I I've also um you know talked about this a little bit before.
00:14:40.240 --> 00:14:43.519
Previously, I'd called it kind of like, you know, the human AI duct tape.
00:14:43.600 --> 00:14:47.440
It's all those intermediate steps in between that a human had to do, right?
00:14:47.600 --> 00:14:53.279
If you run something in, you know, deep research uh inside ChatGPT, well, now I have to copy that.
00:14:53.440 --> 00:14:58.240
I have to go put it in a doc, and then I have to upload it to this project folder as an example.
00:14:58.559 --> 00:15:12.799
That is where this context carry and the larger context windows starts to erase um all of that manual, you know, human AI duct tape that, you know, those steps that us humans working with multiple AI systems would have to continually make.
00:15:13.039 --> 00:15:17.519
Because now these agents also have write ability, right?
00:15:17.679 --> 00:15:26.720
Whereas before, you know, three to six months ago, they didn't have the ability to write to your Google Docs, they didn't have the ability to send Gmails, right?
00:15:26.960 --> 00:15:28.480
Now they do, right?
00:15:28.559 --> 00:15:33.440
If you give them the permissions and if you're feeling you know spicy and you want to roll the dice.
00:15:33.679 --> 00:15:41.360
Uh, but that agentic context carry layers the schedule and the memory over that, but no one is really talking about this.
00:15:41.440 --> 00:15:44.879
I don't know, maybe maybe I'm too dorky and excited about where we are.
00:15:45.039 --> 00:15:53.759
Um, but the reality is I think that there's been this narrative right now.
00:15:53.919 --> 00:15:59.519
And um, I'm actually gonna call out a recent uh tweet uh I saw or in X.
00:15:59.840 --> 00:16:02.080
What do you call what do you call a tweet on X anymore?
00:16:02.159 --> 00:16:02.399
I don't know.
00:16:02.480 --> 00:16:03.840
It's just why I call it Twitter.
00:16:04.000 --> 00:16:05.840
Uh you can't verb X.
00:16:06.000 --> 00:16:10.080
Uh, but there's a recent viral uh tweet out on Twitter, uh, right?
00:16:10.159 --> 00:16:14.639
So ChatGPT had a recent integration with Starbucks, right?
00:16:14.720 --> 00:16:20.559
And someone said, oh, you know, why are all of these apps, right, why do they exist?
00:16:20.720 --> 00:16:22.080
It doesn't make sense, right?
00:16:22.240 --> 00:16:30.639
Because it's gonna take me, you know, two minutes at the absolute fastest to make an order on this Chat GPT Starbucks app integration.
00:16:30.799 --> 00:16:32.240
I'm just using this as an example.
00:16:32.480 --> 00:16:44.799
Throw in any uh you know business app that you're using inside or you know, connector that you're using inside Gemini Copilot uh Claude or ChatGPT, but this kind of viral incident with Starbucks, right?
00:16:44.879 --> 00:16:55.360
It's like, okay, well, it takes two minutes to order it via the the the app and if uh the chat gpt app, but if I go into the actual Starbucks app, I can do it 20 seconds, right?
00:16:55.440 --> 00:16:57.120
So this is done, right?
00:16:57.360 --> 00:17:07.119
Um, but I think people are missing the point because it's not just about one app, it's not just about chat GPT interfacing with one app.
00:17:07.359 --> 00:17:18.319
Because in the new agent builder, right, as an example, the brand new uh agent builder, you can go to create an agent all by hand.
00:17:18.480 --> 00:17:20.960
I'm literally clicking around as I do this now.
00:17:21.440 --> 00:17:25.200
You can connect 20 apps, right?
00:17:25.440 --> 00:17:33.680
Your Gmail, your Slack, your Notion, your Teams, your Outlook email, your Google Calendar, Google Drive, whatever, right?
00:17:33.920 --> 00:17:36.000
MCP servers, you can do all those things.
00:17:36.079 --> 00:17:41.680
Uh, you can connect agent skills, you can upload files, you can manage the memory, right?
00:17:41.839 --> 00:17:52.960
So it's not just about, oh my gosh, you know, using a single app to do a task is so much slower than it is to just do it individually in that platform or on that website.
00:17:53.119 --> 00:17:54.000
That's not what it is.
00:17:54.079 --> 00:17:56.559
It's about eliminating that human AI duct tape.
00:17:56.720 --> 00:18:01.279
It's about, you know, the the 30 small human steps in between that are required.
00:18:01.440 --> 00:18:03.119
That is the context carrying.
00:18:03.359 --> 00:18:08.880
Us humans have been the one carrying the context because AI agents didn't have the ability.
00:18:08.960 --> 00:18:12.720
They didn't have the um the tools to do that.
00:18:13.279 --> 00:18:14.319
Now they do.
00:18:14.480 --> 00:18:15.200
That's the thing.
00:18:15.359 --> 00:18:28.240
Yes, I can much more quickly go open my Gmail, read an email, and respond to it uh than uh a connection in Gemini, ChatGPT, Quad, et cetera, right?
00:18:28.400 --> 00:18:33.359
But what about when there's a Google Doc that goes with it?
00:18:33.519 --> 00:18:34.960
I have to look in my calendar.
00:18:35.200 --> 00:18:39.279
Oh, there's actually three or four different emails, um, right?
00:18:39.440 --> 00:18:42.400
Oh, there's that file in my drive.
00:18:42.480 --> 00:18:45.119
There's a Slaughter conversation about that, right?
00:18:45.279 --> 00:18:53.519
Now all of a sudden, yes, it might be quicker to do all of those small tasks individually in those apps or on those websites.
00:18:53.680 --> 00:18:59.039
But when you have to carry the context yourself manually as the human, that's where you can start, right?
00:18:59.279 --> 00:19:10.880
This has essentially been the uh mundane nature of knowledge work in front of a computer over the past 20 years, you know, as SaaS and applications have exploded.
00:19:11.039 --> 00:19:12.400
But that's what we do.
00:19:12.640 --> 00:19:29.759
And that is where the true benefit of now agentic context carry because us humans no longer have to do the duct tape and have to remember and have to bring that context from app A to app B to app C to storage D to messaging platform E and F.
00:19:30.720 --> 00:19:34.000
The agent does it all for us in one swoop.
00:19:34.160 --> 00:19:49.119
So, yes, it might take you five times as long to accomplish a goal inside of an AI agent, but that's not counting the human error, the human lookup, the human retrieval that has to happen every single step of the way in between.
00:19:49.359 --> 00:19:52.079
That's the big unlock here, y'all.
00:19:52.319 --> 00:19:56.319
Uh, but also, I don't know, I start to forget things fairly quickly.
00:19:56.480 --> 00:19:57.440
Maybe it's just me.
00:19:57.599 --> 00:20:03.359
Like I literally use the you know this concept of agentic context carry all the time, right?
00:20:03.519 --> 00:20:08.720
I was actually walking, uh, I was actually walking to uh my office, right?
00:20:08.799 --> 00:20:12.480
Sometimes I record um, you know, from my kind of home office.
00:20:12.559 --> 00:20:15.359
Sometimes I uh, you know, record from my actual office.
00:20:15.440 --> 00:20:18.799
And you know, I'm working on a cool partnership here uh with a group Sage.
00:20:19.200 --> 00:20:28.960
And I had a couple different email threads with with travel, there was some Google Docs, there was all these things, and I'm like, ah, which, you know, and I have multiple emails, right?
00:20:29.119 --> 00:20:36.160
I have multiple email accounts, certain forms go different places, and I'm like, my gosh, like this is gonna take me a long time.
00:20:36.400 --> 00:20:40.319
Instead, right, just use in this instance, use Claude.
00:20:40.480 --> 00:20:42.000
It went and carried that context.
00:20:42.079 --> 00:20:44.640
But what about when you can schedule those things, right?
00:20:44.880 --> 00:20:59.200
Uh, and to say, hey, every day at you know, 2 a.m., I want you to go through my my email, my calendar, uh, notion, Slack, all of these things.
00:20:59.279 --> 00:21:06.720
Yeah, it might take the agent longer to do that if than if you were to, but it's gonna do it on its own schedule.
00:21:07.119 --> 00:21:09.680
And it's gonna carry the context from app to app.
00:21:09.759 --> 00:21:11.759
So that's where the new breakthrough comes.
00:21:11.920 --> 00:21:15.200
It's the capabilities that have made the cross-app technology possible.
00:21:15.359 --> 00:21:19.519
So here's where the unlock and the timing all comes into play, right?
00:21:19.680 --> 00:21:22.319
Um, I love Venn diagrams, right?
00:21:22.559 --> 00:21:30.799
This is where it's kind of the capabilities and the technology and the need have all overlapped with this perfect timing.
00:21:30.960 --> 00:21:40.240
So this is, you know, if you think of like cowork or agentic scheduling, the features and then the context window all coming together and exploding at the same time, right?
00:21:40.319 --> 00:21:43.359
So Claude and Frapic has really led the way in this.
00:21:43.519 --> 00:21:48.559
So now they have that 1 million token context window by default, right?
00:21:48.720 --> 00:21:59.759
Codecs a little bit uh, you know, behind, although there is an experimented uh experimental uh 1 million token context window uh in the command line interface, but on the app, I believe it's 250.
00:22:01.680 --> 00:22:02.400
So what does that mean?
00:22:02.480 --> 00:22:02.720
Right.
00:22:02.799 --> 00:22:11.359
If you're not too technical, that just means, right, the free version of Chat GPT, last time I checked, uh I didn't check the free version in a while.
00:22:11.519 --> 00:22:15.839
But let's just say in 2025, it was about 8,000 token context window, right?
00:22:16.000 --> 00:22:17.839
So now you're looking at 1 million.
00:22:17.920 --> 00:22:19.039
So do the math there.
00:22:19.119 --> 00:22:22.559
Or you know, or you know, going to 258,000.
00:22:22.720 --> 00:22:31.839
Essentially, now AI models and AI agents can remember things over a very longer period of time, right?
00:22:32.000 --> 00:22:34.960
Whereas before, they essentially had very short-term memory.
00:22:35.119 --> 00:22:42.480
You would, you know, especially if you were on a free plan or, you know, early in 2024, 2025, AI models forgot things very quickly.
00:22:42.640 --> 00:22:45.200
So especially when it came to handling your data.
00:22:45.359 --> 00:22:48.000
So if you upload a file, you're working with it, right?
00:22:48.079 --> 00:22:55.599
Maybe updating a job description and doing some research on recent law changes to make sure that your you know job description reflects those or something like that, right?
00:22:55.680 --> 00:22:59.039
And it's going well, and all of a sudden it's, oh, wait, it's dumb, right?
00:22:59.200 --> 00:23:00.720
That's because it ran over the context widow.
00:23:00.880 --> 00:23:11.519
Context widows used to be very small, but now as they become bigger and bigger and bigger, right, it's essentially you're working with an AI model that has a bigger brain that's able to carry the conversation for longer.
00:23:12.480 --> 00:23:17.200
So now you know this kind of trending concept that's happening.
00:23:17.440 --> 00:23:23.200
It's not just going in and working with one app or one connector, right?
00:23:23.359 --> 00:23:41.200
It's bringing in all of the different tech stack that you have to use on a daily basis, that your company has to use on a daily basis, eliminating all of those manual steps in between, because the reality is, just like a large language model, us humans, we have a context window as well.
00:23:41.440 --> 00:23:43.039
How much time do you spend, right?
00:23:43.119 --> 00:23:45.359
Even pre-AI, it was obviously way worse.
00:23:45.599 --> 00:24:05.440
But sometimes you spend just as much time trying to either track, remember, or find certain information where it lives within your kind of SaaS database as it actually takes to create that new business value once you do find it, or reply to a certain email, or to finish a certain deck or a project or fill out a spreadsheet, right?
00:24:05.519 --> 00:24:10.240
Sometimes you spend as much time just trying to retrieve that information.
00:24:10.480 --> 00:24:17.920
So that's where the multiple apps, because the big context window and the new agentic capabilities, those three things coming together come to play.
00:24:18.079 --> 00:24:25.839
So now that you know it's here and you know that this I think is the intermediate step stepping stone until we have those fully autonomous agents.
00:24:26.000 --> 00:24:31.920
You need to take advantage of this scheduled agentic context carry sack, right?
00:24:32.319 --> 00:24:35.359
Here's how step one, three steps, ready?
00:24:35.440 --> 00:24:36.480
I'm gonna go quick.
00:24:36.640 --> 00:24:40.799
Uh, connect your live data sources and your preferences first.
00:24:41.200 --> 00:24:42.319
Make sure to do this.
00:24:42.480 --> 00:24:46.160
And you know, I do have to you know put out the normal disclaimer, right?
00:24:46.400 --> 00:24:48.880
The uh responsible AI person I am.
00:24:48.960 --> 00:24:49.359
Uh, right.
00:24:49.599 --> 00:24:50.640
I'm a business owner.
00:24:50.799 --> 00:24:54.480
I decide if this is safe for my organization, but you need to do the same, right?
00:24:54.640 --> 00:24:57.519
You shouldn't be doing this with shadow AI tools.
00:24:57.599 --> 00:25:00.319
Uh, you need to be doing this with approved tools.
00:25:00.480 --> 00:25:02.240
So make sure you go through the proper channels.
00:25:02.319 --> 00:25:07.599
But let's just say you have, you know, Claude approved or you have ChatGPT approved, whatever it is, right?
00:25:07.680 --> 00:25:10.880
Uh, and you have these connectors or apps approved as well.
00:25:11.039 --> 00:25:11.279
All right.
00:25:11.359 --> 00:25:21.119
So you need to authorize those live connectors to your, as an example, your email, your calendar, your Slack, your drive, all of those important things, your CRM, right?
00:25:21.200 --> 00:25:21.839
It's huge.
00:25:22.000 --> 00:25:30.480
Then you need to understand how each system's computer use and access works, and then ensure your custom instructions and memory are updated accordingly.
00:25:30.640 --> 00:25:36.160
So, first, you have to get your data sources, uh, your preferences, and your memory in line.
00:25:36.240 --> 00:25:40.960
Because when you talk about context theory, well, context is the base, right?
00:25:41.119 --> 00:25:45.519
We did an earlier show in the start here series on the importance of context engineering.
00:25:45.599 --> 00:25:47.920
So make sure you go back and listen to that one as well.
00:25:48.079 --> 00:25:51.359
And then also, all these platforms, they support MCPs.
00:25:51.519 --> 00:26:02.720
So even if your uh, you know, your app of choice, whatever you're using, doesn't um have any of the, you know, oh, it doesn't have an app connection to ChatGPT or it doesn't have an app connection to Claude.
00:26:02.799 --> 00:26:06.960
Well, chances are you can just use a an MCP server and get that cooked up right away.
00:26:07.119 --> 00:26:08.240
So that's step one.
00:26:08.319 --> 00:26:13.039
Step two, you need to context stuff in a dedicated memory thread.
00:26:13.200 --> 00:26:17.759
Here's a little, I wouldn't necessarily call this a cheat code per se.
00:26:17.920 --> 00:26:20.079
And this is much different, right?
00:26:20.160 --> 00:26:29.680
I I I've obviously talked taught uh you know this concept of of prime prompt polish, you know, the basics of prompt engineering 101 with the context window.
00:26:29.759 --> 00:26:33.920
It doesn't throw away those best practices, but it does kind of change what can get done.
00:26:34.079 --> 00:26:41.440
So here's a little uh little cheat sheet one for you for listening to this uh episode now for 26 minutes and running.
00:26:41.599 --> 00:26:49.279
Uh these systems now, the context windows are enormous.
00:26:49.440 --> 00:27:01.599
You can work on it in theory for a very depends on what tools you're calling, but you know, in Claude Code as an example, uh, you know, running your routines all in the same thread, a daily schedule, it's gonna hold, right?
00:27:01.680 --> 00:27:11.359
I have some that run every single day that started uh when it first came out, you know, two-ish weeks ago, and they're not even close to hitting uh the context window.
00:27:11.599 --> 00:27:11.839
Right.
00:27:12.000 --> 00:27:18.799
So at a million tokens, the agent can hold just weeks of regular usage working memory at once.
00:27:18.960 --> 00:27:25.039
So here's what I like to do connect everything to one thread, right?
00:27:25.119 --> 00:27:27.440
They these all work a little differently, right?
00:27:27.599 --> 00:27:38.720
Uh you know, codex works a little bit differently than Claude Code works a little bit differently than these brand new, you know, ancient builders essentially that we got from uh from ChatGPT and from Google Gemini.
00:27:38.880 --> 00:27:46.240
But essentially, uh, if you do connect things on a thread by thread or a folder-by-folder basis, have one where you just context stuff, right?
00:27:46.400 --> 00:27:48.400
Put all your context in there at once.
00:27:48.480 --> 00:27:50.240
And then that can be your daily drive.
00:27:50.400 --> 00:27:57.519
Because the good thing is, is then also if you need to take it into a different direction, you can just fork that threat at that point, right?
00:27:57.599 --> 00:28:15.599
So you at least have this unified base where you can every single day, right, have it be the one that brings you your morning triage, the one for your you know, most common day-to-day tasks, but aren't necessarily specifically project-based that require a lot of different uh directional feedback, uh, et cetera.
00:28:15.680 --> 00:28:15.920
Right.
00:28:16.079 --> 00:28:19.599
So from that, you can iterate on the reasoning until it matches your standard.
00:28:19.680 --> 00:28:20.400
That's the big thing.
00:28:20.559 --> 00:28:21.440
You need to context.
00:28:21.599 --> 00:28:25.039
Step two is technically context stuff and iterate as well.
00:28:25.200 --> 00:28:27.759
Well, iterate will actually get a little bit more into step three.
00:28:27.920 --> 00:28:28.160
Sorry.
00:28:28.319 --> 00:28:29.920
I I jumped ahead of myself.
00:28:30.000 --> 00:28:35.119
So step two, context stuff and uh context stuff in a dedicated memory thread.
00:28:35.200 --> 00:28:38.480
And then step three, iterate with chain of thought.
00:28:38.720 --> 00:28:41.359
If you listen to the show at all, you know how important this is.
00:28:41.519 --> 00:28:46.480
You saw this in my little demo uh that I did yesterday on codecs, right?
00:28:46.720 --> 00:28:55.599
You need to understand um how these models work because they are generative, they are not deterministic, they're gonna work slightly different each time.
00:28:55.680 --> 00:28:58.000
So you can't just run something once, right?
00:28:58.160 --> 00:29:04.000
Especially as these uh the capabilities become uh greater and greater, right?
00:29:04.160 --> 00:29:06.240
A lot, I I see a common mistake a lot.
00:29:06.400 --> 00:29:09.440
People will run something once and they're like, oh yeah, this is great.
00:29:09.519 --> 00:29:10.640
Let's put it out in production.
00:29:10.799 --> 00:29:16.319
Okay, well, that's could be dangerous, especially if you're doing something uh public-facing or client-facing.
00:29:16.400 --> 00:29:21.039
You probably wouldn't want to do that just yet, uh, right, because there's always going to be edge cases.
00:29:21.200 --> 00:29:34.799
You can run the same scheduled run every single day for seven days, and it might two of the days it might call the tool that you didn't want it to, or uh maybe it's uh not calling a tool that you are telling it to.
00:29:34.960 --> 00:29:38.480
So you really do have to review the chain of thought and iterate.
00:29:38.559 --> 00:29:46.559
So what that means, most systems, you can kind of have some level of observability, traceability uh as they go by looking at the system.
00:29:46.720 --> 00:29:52.079
So if they're scheduled, you can go, you know, usually you might click on, you know, might say, oh, thought for one hour.
00:29:52.160 --> 00:30:00.960
And you can click that and then see every single tool, every single step, and kind of get how the uh that schedule would run or that co-work session, how it worked.
00:30:01.039 --> 00:30:03.039
And you can kind of trace it the same way, right?
00:30:03.119 --> 00:30:05.839
When you think of like Bath, right, where you had to show your work.
00:30:05.920 --> 00:30:09.039
I don't understand like the new common core bath stuff, right?
00:30:09.200 --> 00:30:14.640
But back in my day, right, we just, I don't know, wrote down the the numbers in a column, right?
00:30:14.720 --> 00:30:15.759
But you had to show your work.
00:30:15.920 --> 00:30:24.160
So you should always be checking the work of your, you know, co-working run of your scheduled agent task and then iterating it.
00:30:24.240 --> 00:30:36.160
You need to uh refine the prompt, make it better, and then once it is kind of quote unquote ready for production and you've built those guardrails in place, that's when you can save it as a routine or a scheduled automation.
00:30:36.240 --> 00:30:39.039
Then what's refined, that is that hidden workflow.
00:30:39.119 --> 00:30:43.599
So it is those three steps that really allow that agentic context carry.
00:30:43.759 --> 00:30:44.559
Again, faster.
00:30:44.799 --> 00:30:48.000
Step one, connect your live data sources and preferences first.
00:30:48.240 --> 00:30:51.200
Step two, context stuff in a dedicated memory thread.
00:30:51.359 --> 00:30:56.799
And then step three, iterate with chain of thought reasoning before you put it out into production.
00:30:56.960 --> 00:31:04.079
But then schedule that thing and take advantage of this stepping stone that I think is going to be huge.
00:31:04.240 --> 00:31:06.160
And the time is now.
00:31:06.480 --> 00:31:09.440
So, like I said before, this is not the final destination.
00:31:09.599 --> 00:31:15.599
I think that truly autonomous agents with persistent memory are the next big deal, but we that could be far off.
00:31:15.680 --> 00:31:16.000
Who knows?
00:31:16.079 --> 00:31:28.880
I mean, maybe we'll have that, you know, next month, but it could still be another year, two years or more until we actually see autonomous agents that you can give them a goal and they don't really require much else.
00:31:29.200 --> 00:31:32.079
This is the now or the next.
00:31:32.240 --> 00:31:37.920
So understand um and really push this agentic context carry.
00:31:38.000 --> 00:31:49.359
The leaders pulling ahead this quarter are building scheduled context on autopilot, not just you know, better one-off prompts, not just you know, sharing skills within your organization.
00:31:49.519 --> 00:31:53.519
That's no longer enough, uh, right, to really be pushing in your space.
00:31:53.680 --> 00:32:08.240
So pick one recurring task this week, take it through those three steps, and then deploy your first kind of hidden workflow inside of it that's taking advantage of scheduled agentic context carry.
00:32:08.400 --> 00:32:10.079
I hope this was helpful, y'all.
00:32:10.240 --> 00:32:13.759
If it was, please go to starthearseries.com.
00:32:13.920 --> 00:32:22.880
That's gonna take you straight to a signup form to get access to our community for free, the everyday AI inner circle.
00:32:22.960 --> 00:32:36.079
And then in the Start Here Series space, you can go find every single Start Here Series podcast, read every single Start Here Series newsletter, all in one space, connect and network with others who are doing the same.
00:32:36.240 --> 00:32:36.559
All right.
00:32:36.720 --> 00:32:37.759
I hope this was helpful.
00:32:37.920 --> 00:32:38.720
Thanks for tuning in.
00:32:38.880 --> 00:32:42.640
Hope to see you back tomorrow and every day for more everyday AI.
00:32:42.720 --> 00:32:43.759
Thanks, y'all.
