文字起こし 🔗
WEBVTT
00:00:01.199 --> 00:00:11.039
As smart as today's AI systems are, do you ever wish that they would just watch you do all your work and then magically do it all for you?
00:00:11.439 --> 00:00:19.359
Well, that's what we have now with a new feature from OpenAI and Codex called Record and Replay.
00:00:19.920 --> 00:00:59.679
Because here's the reality: as we are continuing to shift from the responsive chatbot era to the proactive agent era, there's still a lot of manual work that we have to do, whether that's scheduling and automating your workflows, creating new skill files by hand or whatever it may be, whether you're using Chat GPT, Codex, Claude, Copilot, or Gemini, there's a lot of manual work that needs to go in to create these repeatable and autonomous processes.
00:01:00.000 --> 00:01:02.399
But I think that may start to change.
00:01:02.479 --> 00:01:19.439
And I think that with what this new feature is in Codex Record and Replay, we have maybe a preview of what's coming next, as we have these extremely powerful super apps that live on our desktop and can do work for us.
00:01:19.599 --> 00:01:22.480
So that's what we're going to be going over on today's show.
00:01:22.560 --> 00:01:27.840
But here is the big picture of Codecs Record and Replay.
00:01:28.239 --> 00:01:32.640
So I think prompting is kind of losing to demonstration.
00:01:32.719 --> 00:01:43.280
And maybe that's not a bad thing, even if you've spent a lot of time refining your prompt engineering skills or how you type to a chat bot that could be leaving by the wayside.
00:01:43.439 --> 00:01:47.040
So OpenAI shipped this new feature less than a week ago.
00:01:47.200 --> 00:02:00.000
And essentially, you demonstrate something on your on your actual computer on your Mac right now, and it becomes an editable skill.md file, then the agent can reuse.
00:02:00.239 --> 00:02:12.080
So the AI interface, I think, is moving away from just typing instructions to just showing your work and showing a series of tasks that go into a project that you may want to start automating.
00:02:12.240 --> 00:02:16.000
And I think that this right now it seems like a small feature, right?
00:02:16.159 --> 00:02:19.439
And I'm actually surprised this one didn't blow up more than it did.
00:02:19.680 --> 00:02:25.360
But I think this is a big step in transferring agency and workflow to autonomous systems.
00:02:25.520 --> 00:02:57.680
Because the other thing why I think this is actually a big deal, because this kind of open AI is showing their cards in terms of model capabilities, because now, through whatever magic they did, they are now uh can understand and ingest video, which is huge because this has been a big advantage that Google Gemini has had for the last two years as the only um kind of big player that you could upload a video to and it could actually understand uh what was going on inside that video.
00:02:57.759 --> 00:02:58.960
So that's a bigger picture.
00:02:59.039 --> 00:03:00.479
Maybe we'll talk about that a little bit more.
00:03:00.639 --> 00:03:03.039
But here, stick with me for the next 25 minutes.
00:03:03.199 --> 00:03:05.919
This is gonna be a short one uh on today's show.
00:03:06.080 --> 00:03:06.960
Here's what you're gonna learn.
00:03:07.039 --> 00:03:12.800
You're gonna learn what OpenAI actually changed inside of Codex uh beyond just the launch demo.
00:03:12.960 --> 00:03:21.759
You're gonna know why the skill md or the skill markdown file is the real business artifact here that leaders should understand.
00:03:21.919 --> 00:03:27.599
You're gonna know how the shared format changes uh the codecs versus claude comparison.
00:03:27.680 --> 00:03:28.879
Yeah, stick around for that.
00:03:29.039 --> 00:03:35.280
And you're gonna know which workflows you should try to record first without creating unnecessary security risk.
00:03:35.599 --> 00:03:37.360
All right, I hope you're excited for this.
00:03:37.520 --> 00:03:42.719
This is everyday AI in our weekly segment putting AI to work on Wednesdays.
00:03:42.879 --> 00:03:51.120
So once a week, I get a little brave, I demo some live technology, it's unedited, unscripted.
00:03:51.439 --> 00:03:52.479
We'll see how it goes.
00:03:52.639 --> 00:04:01.039
Sometimes these hands-on demos are amazing, sometimes we get bugs, but that's the reality of generative AI and working on the cutting edge.
00:04:01.199 --> 00:04:02.719
Sometimes you get cut.
00:04:02.879 --> 00:04:03.280
All right.
00:04:03.520 --> 00:04:07.759
So if you haven't already, please make sure to subscribe to the podcast if this is helpful.
00:04:07.919 --> 00:04:20.800
And then make sure you go to your everydayai.com, sign up for the free daily newsletter, and we're gonna be recapping not just the highlights from today's show, but all the other AI news and developments you need to know to be the smartest person in AI at your company.
00:04:21.040 --> 00:04:26.319
All right, so without further ado, yeah, let's get dirty, shall we?
00:04:26.480 --> 00:04:37.360
All right, so this one's gonna be a little tricky because I'm gonna be jumping back and forth uh between some different screens, some different things I have going on, but we're gonna try to make this work.
00:04:37.439 --> 00:04:49.920
And this is one of those that uh if you are uh listening on the podcast, this might be one of those ones that you should tune into uh for the video version.
00:04:50.000 --> 00:04:55.839
So you can always see that video version of this on our website at your everydayai.com.
00:04:56.000 --> 00:04:56.319
All right.
00:04:56.399 --> 00:05:04.000
I'm just making sure, doing one more check here before I start sharing my screen, that I don't have anything uh too private.
00:05:04.160 --> 00:05:04.560
All right.
00:05:04.800 --> 00:05:11.279
So yeah, not gonna be uh doxing myself or anyone else, or I'll only be doxing myself slightly.
00:05:11.519 --> 00:05:11.920
All right.
00:05:12.000 --> 00:05:16.639
Load full transcript
Uh so let's go ahead and share this screen.
00:05:18.399 --> 00:05:21.519
Uh and and one thing to know, this is a plugin.
00:05:22.000 --> 00:05:22.319
Okay.
00:05:22.639 --> 00:05:29.279
So if you don't know too much about codecs, I've done a couple of shows recently uh on codecs.
00:05:29.439 --> 00:05:34.720
But if you have a paid ChatGPT subscription, then you have a codec subscription.
00:05:34.800 --> 00:05:38.319
It is the same, you know, it uses the same subscription, which is really nice.
00:05:38.480 --> 00:05:43.680
So you can go download uh codecs is a desktop program, whether you use Mac or Windows.
00:05:43.759 --> 00:05:48.399
Although these, a lot of the features come to Macs for Mac first, like record and replay.
00:05:48.560 --> 00:05:48.800
All right.
00:05:48.879 --> 00:05:53.040
And then after you are logged into Codecs, if you haven't downloaded it already, go ahead and do that.
00:05:53.120 --> 00:05:54.480
You can follow along with me.
00:05:54.560 --> 00:05:57.680
Uh, and then you're gonna make sure to go into your upper left-hand corner.
00:05:57.839 --> 00:05:59.600
There's a plugins section.
00:05:59.759 --> 00:06:04.079
All you're gonna do is search for record and replay.
00:06:04.240 --> 00:06:05.920
Uh, then you are going to add the plugin.
00:06:06.319 --> 00:06:16.959
After you do that, if you are brand new to Codecs, you're probably gonna have to go through your Mac security settings and grant it some access uh to some uh different um privacy features that it needs.
00:06:17.040 --> 00:06:19.680
Because as an example, this uses computer use.
00:06:19.759 --> 00:06:24.000
So you do give it access to essentially read and write to your entire computer.
00:06:24.160 --> 00:06:29.439
So, you know, if you are using this in an enterprise setting, make sure to get the permissions you need before doing this.
00:06:29.680 --> 00:06:37.199
Understands the pros and the cons and the risks, and that just goes for not just the record and replay in codecs, but for anything desktop, right?
00:06:37.279 --> 00:06:46.639
So whether you're talking about codecs, uh clawed, um, you know, desktop using clawed code, claud cowork, uh anti-gravity, whatever you're using on your desktop, always keep that in mind.
00:06:46.800 --> 00:06:53.680
So after you have codecs, you're logged in with your ChatGPT plan, you've installed the record and replay plugin.
00:06:54.160 --> 00:06:55.040
Here's what we're gonna do.
00:06:55.120 --> 00:06:56.480
So I'm gonna do this.
00:06:56.720 --> 00:07:00.959
It might seem a little backwards, all right, but stick with me here.
00:07:01.040 --> 00:07:05.920
I'm doing it like this for a reason, mainly so hopefully the demo can work.
00:07:06.079 --> 00:07:10.560
All right, so live stream audience, let me know if you can see my screen here.
00:07:10.720 --> 00:07:13.920
All right, I think we're pretty good, not giving away too much.
00:07:14.079 --> 00:07:20.240
Uh, aside from the fact, yes, I do have uh, you
know, codecs and claude code running here at the bottom, like I always do.
00:07:20.319 --> 00:07:24.160
Uh, and a couple other secret, uh, secret projects I'm working on down there.
00:07:24.399 --> 00:07:29.920
Okay, so I'm going to put in a prompt first describing what I'm going to do.
00:07:30.000 --> 00:07:33.279
And you might be saying, Jordan, that defeats the entire purpose.
00:07:33.519 --> 00:07:34.959
Uh, yes and no.
00:07:35.199 --> 00:07:38.720
Okay, because my example here requires a little nuance.
00:07:38.879 --> 00:07:47.360
So also think of what this is and what it isn't, because right now, this feature all it does is it sees your screen.
00:07:47.519 --> 00:07:54.240
So uh I do think uh that they're going to uh add in the feature where you can talk to it at the same time.
00:07:54.399 --> 00:07:58.000
So you can talk it through and walk it through visually at the same time.
00:07:58.079 --> 00:08:00.720
But essentially, this is gonna record my screen, everything that I do.
00:08:00.879 --> 00:08:10.480
So, because I'm doing a live demo and I want this to hopefully work and you know, not have to accidentally log into all these different places, I'm first gonna give it a precursor on what I'm doing.
00:08:10.639 --> 00:08:12.720
So I'm just putting in a simple prompt.
00:08:12.800 --> 00:08:19.839
Uh, and for this, I'm gonna go ahead um, I'm gonna go ahead and put this prompt in.
00:08:20.000 --> 00:08:20.319
Okay.
00:08:20.480 --> 00:08:21.519
Nothing crazy here.
00:08:21.600 --> 00:08:22.879
I'm just telling it what I'm gonna do.
00:08:23.040 --> 00:08:30.319
So I want you to think of that manual thing that you do that usually requires what I call kind of this contacts carry.
00:08:30.480 --> 00:08:33.360
So, you know, carrying contacts from different programs.
00:08:33.519 --> 00:08:37.440
I'm doing something extremely simple, very straightforward.
00:08:37.679 --> 00:08:38.080
All right.
00:08:38.159 --> 00:08:41.519
Um, I'm just having my thing that I'm trying to automate.
00:08:41.600 --> 00:08:45.279
This is something I don't have time for, even though it's not super time consuming.
00:08:45.519 --> 00:08:47.360
It's just all of these small little tasks.
00:08:47.519 --> 00:09:00.240
When I think about them in my head, I'm like, oh, this requires three, four, five different websites, all these different steps, even if it only takes 10 minutes, a lot of times I just don't do that because my day-to-day schedule is absolutely bunker bananas, right?
00:09:00.720 --> 00:09:13.919
In in my mind, even if a task is 10 minutes, if it requires me going into five, six, seven, eight different systems, doing a lot of copying and pasting, carrying contacts over, even a little bit of brain power, sometimes I just don't do them, which is why I really like this new feature.
00:09:14.000 --> 00:09:21.039
So, all I'm doing in this prompt, I'm saying for this chat, I'm going to be using the record and replay eventually, but do not use it yet.
00:09:21.200 --> 00:09:24.799
And I'm just saying, here's a primer on what I'm going to do before I hit record.
00:09:24.879 --> 00:09:30.559
Yes, I know this is kind of counterintuitive, but because it requires a little bit of nuance, um, this is why I'm typing it out.
00:09:30.639 --> 00:09:41.279
And the reason being, what I'm trying to do is I'm gonna go into my Beehive newsletter, that is our email newsletter provider, and I'm gonna find the five most clicked on stories from our last newsletter.
00:09:41.440 --> 00:09:45.919
But here's the thing most days that's our actual website, right?
00:09:46.080 --> 00:09:49.039
It's everyone clicks on the episode page, they go to our website.
00:09:49.200 --> 00:09:50.639
I don't care about that, right?
00:09:50.720 --> 00:09:58.720
So, what I'm ultimately trying to do is I want to see the five most clicked stories that are not ours, right?
00:09:58.799 --> 00:10:01.279
That aren't also not something, you know.
00:10:01.360 --> 00:10:05.200
Sometimes we link out to Twitter or, you know, a lot of times it's Spotify.
00:10:05.279 --> 00:10:06.879
People click on our Spotify link.
00:10:07.039 --> 00:10:10.960
So, you know, I can't just go click on the five uh, you know, top stories.
00:10:11.039 --> 00:10:12.720
I do have to explain that nuance.
00:10:12.799 --> 00:10:25.919
And Codecs would get this right because this is an editable skill that ultimately this is gonna do, that then I can run uh, you know, whenever I want by calling on the skill, or I can make this a daily um automation, which is what I'm going to do.
00:10:26.000 --> 00:10:37.519
But all I'm gonna do, the five most popular stories, take all the text in there, copy and paste it, put it into notebook LM, and then I'm going to create an audio overview, the debate style.
00:10:37.679 --> 00:10:44.320
So essentially every day I can wake up, I can know the five most important stories that our audience thinks about.
00:10:44.480 --> 00:10:50.159
Not only that, the content of everything, have it be extremely relevant, but then have two hosts, right?
00:10:50.240 --> 00:10:55.360
I love notebook LM to have two hosts kind of debate different sides of those stories.
00:10:55.519 --> 00:11:00.559
So I can hopefully better understand how our audience is thinking about these things.
00:11:00.639 --> 00:11:00.879
All right.
00:11:00.960 --> 00:11:02.399
So that's all I'm doing.
00:11:02.720 --> 00:11:05.840
So uh I went through, I told it, this is what I'm doing.
00:11:06.000 --> 00:11:10.559
So now I'm gonna click the at button, all right, and then I'm gonna click record and replay.
00:11:10.720 --> 00:11:11.039
All right.
00:11:11.200 --> 00:11:14.720
And so I'm all I'm doing is saying, uh, you know, let's start.
00:11:14.960 --> 00:11:15.360
All right.
00:11:15.600 --> 00:11:17.600
So I just clicked the at button.
00:11:17.679 --> 00:11:19.679
Uh, I typed in record and replay.
00:11:19.840 --> 00:11:20.799
I click let's start.
00:11:20.879 --> 00:11:28.399
So now what should happen is it's gonna say, okay, we're recording now, and then it's gonna uh create a little recording thing at the bottom.
00:11:28.480 --> 00:11:30.559
And I want to make sure I have everything ready.
00:11:30.720 --> 00:11:31.679
Yes, I do.
00:11:31.759 --> 00:11:34.320
I'm gonna drag this over here, make sure I'm not showing anything.
00:11:34.480 --> 00:11:35.519
All right, we're good.
00:11:35.679 --> 00:11:41.919
Yeah, I can't even show all my notebook LMs uh because yeah, there's some some private uh things in there.
00:11:42.080 --> 00:11:42.960
All right, so here we go.
00:11:43.039 --> 00:11:44.879
I'm just doing inside codecs.
00:11:44.960 --> 00:11:47.279
I at tagged record and replay.
00:11:47.440 --> 00:11:48.720
I said, let's start.
00:11:49.519 --> 00:11:52.399
It should let's see if it goes here.
00:11:55.279 --> 00:11:59.120
All right, it says it's reading the skill, it's thinking.
00:12:01.519 --> 00:12:02.240
There we go.
00:12:02.399 --> 00:12:04.960
And so now I have a little recording at the bottom of my screen.
00:12:05.039 --> 00:12:08.799
So I'm gonna try to go as fast as I can, but I'll talk you through it for a podcast audience.
00:12:08.960 --> 00:12:12.399
Nothing crazy is happening here, but Codex is now recording my screen.
00:12:12.879 --> 00:12:15.840
I do have dual monitor, so hopefully this will work okay.
00:12:16.000 --> 00:12:18.159
All right, so now I'm going into my Beehive account.
00:12:18.240 --> 00:12:27.360
I'm logged in, I'm going to the newsletter, I'm going to yesterday's um yesterday's uh newsletter.
00:12:27.600 --> 00:12:29.759
I'm going to uh where am I going?
00:12:29.919 --> 00:12:30.720
Performance, yes.
00:12:30.879 --> 00:12:31.120
Okay.
00:12:31.279 --> 00:12:34.159
So now here's all the different ones, the different clicks.
00:12:34.399 --> 00:12:35.759
These are all the different stories.
00:12:35.919 --> 00:12:41.759
So uh I'm gonna click what I think are five of the most read stories.
00:12:41.840 --> 00:12:43.759
I'm just going in order here.
00:12:45.759 --> 00:12:46.399
All right.
00:12:49.679 --> 00:12:53.039
I want to make sure all of these are actual news stories.
00:12:53.200 --> 00:12:54.000
There we go.
00:12:54.240 --> 00:12:56.000
It looks like we have them all.
00:12:56.960 --> 00:13:03.120
Okay, one is just for uh GLM, uh, which just takes you to their website.
00:13:03.200 --> 00:13:05.039
So that that's not gonna tell me anything.
00:13:05.120 --> 00:13:09.759
So I'm just actually gonna Google it because it looks like we just linked to their actual website.
00:13:09.919 --> 00:13:13.759
Um, but let me just go ahead and grab the actual announcement post there.
00:13:13.840 --> 00:13:14.559
There we go.
00:13:14.799 --> 00:13:16.960
Okay, so I think I'm good.
00:13:17.840 --> 00:13:20.879
All right, this one couldn't be reached for whatever reason.
00:13:21.120 --> 00:13:23.200
All right, we're good, we're good.
00:13:23.840 -->
00:13:24.559
What is this?
00:13:24.720 --> 00:13:26.080
This was a random tool.
00:13:26.240 --> 00:13:27.919
All right, so I'm not gonna do that one.
00:13:28.080 --> 00:13:30.480
We got to verify myself via Cloudflare.
00:13:30.720 --> 00:13:31.360
Yeah, this is fun.
00:13:31.440 --> 00:13:32.480
All right, let's do one more.
00:13:32.639 --> 00:13:37.840
I do want to make sure that we have five that we can actually copy and paste and grab the text.
00:13:38.720 --> 00:13:42.559
All right, it looks like this fortune article is not working.
00:13:42.960 --> 00:13:43.519
Cool.
00:13:43.840 --> 00:13:50.399
So maybe Codecs will see that I can't copy and paste this because it's, you know, apparently behind a paywall.
00:13:50.720 --> 00:13:53.519
All right, we'll see if Codex picks up on that or not.
00:13:54.159 --> 00:13:55.679
But again, this is editable.
00:13:55.759 --> 00:14:02.879
So I can always go and say if you can't access um the text, then you know, go down to the next story.
00:14:03.039 --> 00:14:05.600
Okay, so it looks like we're good here.
00:14:06.799 --> 00:14:10.559
So I'm gonna go ahead and grab all of the text from these.
00:14:10.639 --> 00:14:12.559
So I'm gonna copy and paste.
00:14:12.720 --> 00:14:15.039
Um, I'm gonna go into copied text.
00:14:15.200 --> 00:14:17.600
I'm gonna paste, insert, next.
00:14:17.759 --> 00:14:20.480
I'm clicking, uh, let me just X out of that one.
00:14:20.639 --> 00:14:25.360
That was something about NVIDIA, about the uh water sustainability.
00:14:25.519 --> 00:14:27.360
I'm gonna click on add sources.
00:14:27.519 --> 00:14:34.399
Here's OpenAI announcing um how they're using AI to diagnose rare genetic diseases.
00:14:34.559 --> 00:14:36.799
Uh, I'm pasting that into Notebook LM.
00:14:36.879 --> 00:14:40.399
Again, I'm just copying and pasting the context of all of these pages.
00:14:41.120 --> 00:14:44.960
All right, next, there was a micron in anthropic uh announcement.
00:14:45.120 --> 00:14:46.639
I'm copying and pasting that.
00:14:46.799 --> 00:14:49.519
I'm adding this inside of Notebook LM.
00:14:49.679 --> 00:14:51.279
We have three stories so far.
00:14:51.519 --> 00:14:52.559
Let's get two more.
00:14:52.639 --> 00:14:55.919
I'm copying and pasting the GLM 5.2.
00:14:56.240 --> 00:14:58.000
All right, putting that in there.
00:14:58.240 --> 00:14:59.120
There we go.
00:14:59.279 --> 00:15:05.440
And then our last one here something about um someone leaving DeepMind for Anthropic.
00:15:05.600 --> 00:15:06.399
Let's go ahead.
00:15:06.639 --> 00:15:09.279
Actually, there's a ton of extra stuff on that page.
00:15:09.440 --> 00:15:11.679
So we'll see if Codex uh picks this up.
00:15:11.759 --> 00:15:13.360
That there's a lot of extra stuff here.
00:15:13.440 --> 00:15:14.799
It's a shorter article.
00:15:14.960 --> 00:15:19.120
So uh I'm gonna copy and paste that in and done.
00:15:19.279 --> 00:15:22.240
All right, the next thing, and this is all laid out in the instructions.
00:15:22.399 --> 00:15:27.600
I want Codecs to go on here, click audio, and then, but I want to customize this.
00:15:27.679 --> 00:15:28.799
So there's a little carrot.
00:15:28.879 --> 00:15:29.759
I'm gonna click that.
00:15:29.919 --> 00:15:31.600
I want it to go to the debate.
00:15:31.759 --> 00:15:36.559
All right, and then I actually have a prompt that I'm going to want it to copy and paste.
00:15:36.639 --> 00:15:37.600
And I told it this.
00:15:37.840 --> 00:15:48.399
So the prompt, the custom prompt that I'm using for this audio overview is this the included sources are web page context of the most clicked articles from the everyday AI newsletter.
00:15:48.559 --> 00:15:59.360
Please simulate how business, I'm gonna say how busy business leaders who are leading AI implementation at their companies.
00:15:59.519 --> 00:16:00.559
That's you guys, right?
00:16:00.639 --> 00:16:12.559
Uh, maybe interpreting these stories and debate how each side uh thoughtfully thinks about these with pinpoint detail, avoid generalities or soft takes.
00:16:12.720 --> 00:16:13.039
All right.
00:16:13.120 --> 00:16:17.120
And then I told Codecs, once I hit generate, it is done.
00:16:17.200 --> 00:16:18.320
So let me go back.
00:16:18.480 --> 00:16:23.600
Uh, all I have to do is I just have to click the record button one more time.
00:16:23.840 --> 00:16:30.320
And then what that does is it sends a message to Codecs actually that says, I'm done recording.
00:16:30.480 --> 00:16:35.279
Okay, so now what it says, what it's saying is it says I'm stopping the recorder now.
00:16:35.440 --> 00:16:38.799
Then I'll inspect the captured event stream directly.
00:16:38.960 --> 00:16:43.840
After that, I'll build the reusable skill rather than just summarizing the demo.
00:16:44.080 --> 00:16:46.480
So um, pretty good.
00:16:48.159 --> 00:17:00.399
So I'm gonna give that a minute to cook as we go over some of the other details, and then hopefully we'll see it at the end and I'll see if I can actually um you know go and run this skill.
00:17:00.480 --> 00:17:01.600
We'll see how long it takes.
00:17:01.759 --> 00:17:02.080
All right.
00:17:02.240 --> 00:17:09.759
Also, I just realized I forgot to do a couple of things that I told Codex I was gonna do, like renaming the notebook, uh, which I didn't do.
00:17:09.839 --> 00:17:13.039
So we'll see what it picks up on, some of the nuance there.
00:17:13.279 --> 00:17:15.440
But let's just quickly go over what's new.
00:17:15.519 --> 00:17:20.720
I talked about a lot of this as I was going through, but right now this is Mac only.
00:17:20.880 --> 00:17:24.000
And during the recording, Codex observes everything you do.
00:17:24.160 --> 00:17:26.240
So actions, what you click on.
00:17:26.319 --> 00:17:27.599
You can do different programs.
00:17:27.680 --> 00:17:31.200
So I just told it to go to a specific Chrome profile.
00:17:31.359 --> 00:17:37.279
So when it builds this skill, uh, it is going to tell itself to use that
Chrome profile.
00:17:37.440 --> 00:17:40.319
So uh Codex has all of these different skills, right?
00:17:40.400 --> 00:17:47.359
So the big one of the big differences between Codex and ChatGPT is well, Codex is an autonomous coding agent.
00:17:47.519 --> 00:17:48.960
Don't worry if you're not technical.
00:17:49.119 --> 00:17:53.759
All that means is it can use your computer, it can use any programs, any other desktop programs.
00:17:53.920 --> 00:18:01.279
One of the most used things I have Codecx do is to use Claude Code on the desktop, uh, to use Google Gemini on the desktop.
00:18:01.359 --> 00:18:03.599
So I just have it use all my other desktop programs.
00:18:03.680 --> 00:18:05.519
It can read and write files, all that good stuff.
00:18:05.599 --> 00:18:08.319
But Codex observes everything you do.
00:18:08.480 --> 00:18:14.960
So different actions you take, what you click on, what you highlight, what you type in, etc., and also the window content.
00:18:15.119 --> 00:18:23.200
And then those observations, in order sequentially, essentially become this skill file that changes what you're doing, right?
00:18:23.920 --> 00:18:30.319
So when the recording stops, and this is what's happening now live on my other screen, uh, Codec starts to create this skill.
00:18:30.480 --> 00:18:35.839
And this is what I think is really important because a skill is an editable file, which is really nice.
00:18:35.920 --> 00:18:38.559
At any time, I can go in and I can update this.
00:18:38.640 --> 00:18:42.319
But here's the big unlock that I don't think most people are talking about.
00:18:42.400 --> 00:18:53.200
Aside from the fact, like, oh my gosh, apparently uh GPT-5.5 via codecs can ingest and understand video files, which is a first for open AI at scale.
00:18:53.359 --> 00:18:58.160
So before it could use Python, it could go in, you could upload a video, it could look at things frame by frame.
00:18:58.240 --> 00:19:04.079
But as far as I know, this is the first time that open AI slash codec slash chat GPT can actually ingest and understand video.
00:19:04.240 --> 00:19:08.319
And like I said, this is a huge advantage that Google has had for like a year and a half.
00:19:08.480 --> 00:19:13.920
It's something this process here, I've actually been doing this well before Codecs recorded and replayed.
00:19:14.079 --> 00:19:19.279
I was using Google Gemini because I would dictate something, I would upload these little things, and I was creating skills.
00:19:19.359 --> 00:19:21.839
I was already doing this, but again, I'm a little dorky.
00:19:22.000 --> 00:19:24.960
I always try to break these things, but um, this is pretty big.
00:19:25.039 --> 00:19:31.119
But keep in mind this skills MD file, this is an asset that doesn't just have to live inside of Codex.
00:19:31.440 --> 00:19:42.160
Obviously, it's gonna work best in Codex because everything that it writes, you know, if it's using different skills, if it's using different apps, it's going to be
tagging those in a codecs manner that codecs can read.
00:19:42.319 --> 00:19:45.039
However, you can use this in Claude Desktop, right?
00:19:45.119 --> 00:19:48.960
You'll you'll have to go in, but you can import this skill into Claude Desktop.
00:19:49.119 --> 00:19:53.920
You can change, you know, if it does call a codec specific skill, you can change that over.
00:19:54.000 --> 00:20:01.839
If it's, you know, using Claud's or sorry, if it's using Codecs computer use or Codex Chrome, you know, you can go and updo update those things accordingly.
00:20:02.000 --> 00:20:08.400
But essentially, you are using this record uh feature to literally do anything, right?
00:20:08.640 --> 00:20:13.039
Uh and this is why I think codex and claude can actually become more connected now.
00:20:13.200 --> 00:20:23.279
This is, you know, don't think of this as just locking yourself into codecs, although I do think the hardness and codex is much better than Claude Desktop and all the benchmarks show that.
00:20:23.359 --> 00:20:35.039
But if you are a heavy uh clawed desktop user and that's where you get the majority of your work done, but you still also have a ChatGPT account, you should probably go in and give this a try because that shared structure matters.
00:20:35.200 --> 00:20:40.160
Because now these replays, the recording replays, can be used across many different surfaces.
00:20:40.400 --> 00:20:44.720
And remember, it works through your screens, your browsers, and your plugins.
00:20:45.119 --> 00:20:51.839
So you can use other saved skills, you can use other plugins because this is an editable feature.
00:20:51.920 --> 00:20:54.640
So anything that I would have typed in codex, right?
00:20:54.799 --> 00:20:58.079
All I did when I started, I just said, let's start, right?
00:20:58.160 --> 00:21:02.559
I could have typed something out and said, let's use this skill, let's use these apps.
00:21:02.720 --> 00:21:04.960
You, you know, you're gonna watch me work on my computer.
00:21:05.119 --> 00:21:07.359
I just did something simple that would hopefully work.
00:21:07.519 --> 00:21:07.839
All right.
00:21:08.000 --> 00:21:14.400
Also, scripts and connectors can increase the reliability, but screen control, you know, it does increase those trust requirements.
00:21:14.480 --> 00:21:20.400
So if you are rolling this out, you know, in your enterprise, make sure that you get all those proper permissions that you need.
00:21:20.960 --> 00:21:24.720
Uh, and the limits right now are privacy and scope.
00:21:24.880 --> 00:21:26.799
So, you know, not everyone has this.
00:21:26.880 --> 00:21:27.839
You do have to be on a Mac.
00:21:28.079 --> 00:21:32.640
Uh, if you are in the EU, UK, Switzerland, you don't have uh access to this.
00:21:32.799 --> 00:21:35.680
And I would always start off with low stakes workflow.
00:21:35.759 --> 00:21:39.759
So what I just showed, right, this isn't something that's being published live.
00:21:39.920 --> 00:21:41.519
This is something that I'm using internally.
00:21:41.599 --> 00:21:43.359
This is still a very new feature, even for me.
00:21:43.440 --> 00:21:44.319
I've been using it a lot.
00:21:44.400 --> 00:21:44.960
I like it.
00:21:45.039 --> 00:21:46.400
Uh, it's still a new feature.
00:21:46.559 --> 00:21:49.440
So I would start by using this with low stakes things.
00:21:49.599 --> 00:21:56.720
Uh so you know, I would go through list your weekly computer tasks uh with all the steps and clear success criteria.
00:21:56.799 --> 00:22:01.680
So for me, the success criteria for this is okay, is it gonna actually Accurately find those five things.
00:22:01.920 --> 00:22:04.400
And I open my notebook LM app every morning.
00:22:04.480 --> 00:22:10.720
If I go take a walk, I can listen to this and be better up to date on probably hopefully what all of you are thinking.
00:22:10.880 --> 00:22:19.599
So I would start just recording one short workflow using realistic inputs, no secrets, no you're right, nothing, no passwords, nothing like that.
00:22:19.759 --> 00:22:24.799
Read each skill when it's done before rerunning it or setting it up as a schedule of automation.
00:22:24.880 --> 00:22:30.400
And to do that in codex, literally, uh, so it looks like this is done recording.
00:22:30.480 --> 00:22:34.559
It called this skill just the Beehive Notebook LM debate.
00:22:34.720 --> 00:22:34.960
All right.
00:22:35.119 --> 00:22:38.160
So now I can say run this every day at 6 a.m.
00:22:38.400 --> 00:22:38.640
Right.
00:22:38.720 --> 00:22:39.599
That's all you have to do.
00:22:39.759 --> 00:22:47.039
And then every day at 6 a.m., I can open my notebook LM app and it's that uh going to the studio section and it's waiting for me.
00:22:47.200 --> 00:22:52.720
All right, so we can read each skill file, refine pre uh preferences, and build your automation library.
00:22:52.880 --> 00:22:53.920
All right, let's check in.
00:22:54.079 --> 00:22:57.599
I don't think I'm gonna have time to rerun this skill, but I do want to check it.
00:22:57.680 --> 00:22:58.960
All right, so it says done.
00:22:59.039 --> 00:23:00.559
So it took three minutes.
00:23:00.799 --> 00:23:02.640
Oh wow, and impressive here.
00:23:02.799 --> 00:23:06.160
That was in 167 event capture.
00:23:06.240 --> 00:23:10.079
And I would go through, always encourage people, read the chain of thought.
00:23:10.240 --> 00:23:13.359
So I'm going through here, I can see exactly what it did.
00:23:13.519 --> 00:23:15.599
Uh, make sure everything is correct.
00:23:15.759 --> 00:23:19.839
I'm looking through this, it looks like it's doing a pretty good job.
00:23:20.160 --> 00:23:21.759
Um, I don't see anything that's wrong.
00:23:21.920 --> 00:23:26.160
And then it says uh validation passed with official skill validator.
00:23:26.319 --> 00:23:33.039
The only hiccup was uh something with uh Py YAML, uh so something with the programming language there.
00:23:33.279 --> 00:23:36.400
And uh instead it kind of wrote
its way around that.
00:23:36.480 --> 00:23:37.119
So pretty good.
00:23:37.200 --> 00:23:43.200
And then it says that anytime I can invoke this later using the Beehive uh notebook LM debate skill.
00:23:43.359 --> 00:23:45.359
So, as an example, let me just open this.
00:23:45.440 --> 00:23:53.359
Uh, I'm gonna say run the Beehive notebook uh skill uh right now.
00:23:53.680 --> 00:23:54.079
All right.
00:23:54.319 --> 00:23:56.880
So we'll just see, make sure it's gonna kick off.
00:23:57.119 --> 00:23:59.920
It should start uh taking over my browser here.
00:24:00.000 --> 00:24:07.200
So we'll probably see that a little bit in the background, just to show everyone that things are well, hopefully, uh running and working.
00:24:07.279 --> 00:24:07.920
And there we go.
00:24:08.000 --> 00:24:09.119
So it's running it now.
00:24:09.200 --> 00:24:15.599
I'm I'm not gonna, you know, sit around for another, you know, three to 10 minutes and blab on because I wanted to keep this one tight.
00:24:15.759 --> 00:24:17.359
Uh, you know, hey, under 30 minutes.
00:24:17.440 --> 00:24:17.920
There we go.
00:24:18.000 --> 00:24:19.599
Uh, but it's going, it's working.
00:24:19.680 --> 00:24:21.920
And now from here, I would validate it.
00:24:22.000 --> 00:24:24.799
Uh, I would run it once, I would make sure everything is correct.
00:24:24.960 --> 00:24:28.240
I would go listen to it on uh my notebook LM app.
00:24:28.319 --> 00:24:29.599
I would see what I wanted to tweak.
00:24:29.680 --> 00:24:32.480
Once I get it exactly how I want it, then I schedule it.
00:24:32.559 --> 00:24:33.920
You schedule it and forget it.
00:24:34.079 --> 00:24:39.839
But as we wrap up today's show, I want you to really understand what this unlocks.
00:24:40.160 --> 00:24:42.160
Because a couple of things.
00:24:42.400 --> 00:24:47.839
Number one, it is that I think now the easiest way to create skills.
00:24:48.000 --> 00:24:55.440
Skills, still to this day, skills and uh, you know, plugins, which is, you know, Codex's version of using skills and apps.
00:24:55.599 --> 00:25:01.279
So skills and plugins are, I think, probably the new measure, right?
00:25:01.440 --> 00:25:04.079
It's not about how many tokens you and your teams use.
00:25:04.240 --> 00:25:10.880
It's about how often is your agent running a skill or a um or a scheduled skill.
00:25:11.039 --> 00:25:20.000
I think that's huge because that's where you are starting to actually win back time by running repeatable workflows without you, the human, having to go in, type something, find something.
00:25:20.079 --> 00:25:21.279
Oh, where's this Google Doc?
00:25:21.359 --> 00:25:22.799
Where's this Obsidian Vault?
00:25:23.519 --> 00:25:24.400
Where's all this at?
00:25:24.480 --> 00:25:24.720
Right.
00:25:24.799 --> 00:25:30.559
So, you know, really recording your mundane workflows across different browser tabs.
00:25:30.720 --> 00:25:34.000
Obviously, you can do things where you're logged in like this, right?
00:25:34.160 --> 00:25:37.599
Obviously, I was logged into my Gmail to reach uh Notebook LM.
00:25:37.759 --> 00:25:41.599
I was logged into my um my Beehive account.
00:25:41.680 --> 00:25:44.880
So I told it to use my Chrome profile because I keep those logged in.
00:25:45.039 --> 00:25:51.519
So think about all of those things that you do over and over and over, uh, right, and not having to write something out by hand.
00:25:51.680 --> 00:25:56.400
Although I kind of did a little bit of that just to ensure that this would hopefully work zero shot.
00:25:56.559 --> 00:25:58.799
But think of what this unlocks and what this changes.
00:25:58.880 --> 00:26:02.720
And it doesn't require literally anything, it doesn't require any technical skill.
00:26:02.880 --> 00:26:04.319
I'm not an overly technical person.
00:26:04.480 --> 00:26:13.440
You go in, if you have a ChatGPT account, download codecs, go through again, assuming you have all permissions, record these workflows, and it's done.
00:26:13.599 --> 00:26:13.920
Right.
00:26:14.000 --> 00:26:19.519
And now you have a worker working for you that you just showed it what to do and it did it.
00:26:19.599 --> 00:26:26.000
And I think this is a breath of fresh air, a very small feature that is going to make a big difference.
00:26:26.079 --> 00:26:27.200
So thank you for tuning in.
00:26:27.359 --> 00:26:30.880
I hope this was helpful doing our AI Wednesday at work series.
00:26:31.039 --> 00:26:34.640
Like I said, I've done a couple other uh codecs walkthroughs.
00:26:34.720 --> 00:26:40.160
Uh, so make sure you go check out 781, episode 781, where we went over codex sites.
00:26:40.319 --> 00:26:42.799
Another great one with I think a lot of value there.
00:26:42.880 --> 00:26:43.920
So make sure you go check that out.
00:26:44.000 --> 00:26:47.920
And then go to your everydayai.com, sign up for the free daily newsletter.
00:26:48.000 --> 00:26:50.880
We're going to be recapping the highlights from today's show and a whole lot more.
00:26:51.039 --> 00:26:51.839
Thanks for tuning in.
00:26:52.000 --> 00:26:54.880
Hope to see you back tomorrow and every day for more everyday AI.
00:26:54.960 --> 00:26:55.759
Thanks, y'all.