How to Master Grok Bot in 16 Minutes

Tenex· 16 min· 3,370 words· 15 min read· English ·Watch on YouTube

This is the full transcript of How to Master Grok Bot in 16 Minutes, published on YouTube by Tenex. Every paragraph carries the moment it was spoken, so you can click any line to jump straight to that point in the video, search the whole thing for a word, or copy it out.

0:00Everyone is talking about Grokbot. It is the most viral AI tool I've seen in a long, long time, but people are only harnessing 10% of its potential. So, I spoke with Matt Palmer on the Cursor team at SpaceX AI to get his real workflows all the way from easy [music] mode to god mode. By the end of this episode, you will be able to harness the

0:21full power of Grokbot, so let's hop in. [music] Okay, Matt, the four levels of Grokbot. Let's start with easy mode. What do you got? So, I think the first thing that we can talk about is just information retrieval. So, I have some very salient examples. I can't show those exactly because a lot of them happened in the SpaceX or the Cursor Slack, but as knowledge workers, almost all of us have to work across different tools. And so, one thing I like to use Grokbot for is summarizing information, bringing all these tools together, and eliminating context switching. So, here if I was going to talk through this example, I'll

0:59make this a little bit bigger. I had Grokbot actually summarize what it's been doing for me. So, we shipped a bunch of small features. I had to manage basically the launch of this this change. And the most difficult thing thing was keeping all the content in sync. So, I was looking at our change log, our documentation, Twitter and LinkedIn copy, and video captions. Maybe this is a common workflow for like a product marketing manager or somebody that has to manage multiple work streams. Instead of going through all of those different things, I could have one chat with Grokbot where I just had it answer all my questions, pull in all the

1:32feedback, and then update all of those sources. So, here Grokbot's explaining what it did in sort of our demo. And what it was doing is accessing information across multiple Slacks, going into multi pull different places, and summarizing that information, and then it was actually able to go out and make those changes for me. One of the reasons it could do that is because if we look at our plugins here, Grokbot can actually access multiple accounts per plugin and we can connect all the different accounts that we might need to. So our Google Calendar, our mail, if we go into Notion here, which is another popular one. I had it access the SpaceX and the

2:10Cursor Slack recently. Cursor was acquired by SpaceX and so this has saved me so much time in just going across a bunch of different apps, navigating across a bunch of different windows. And so I think that the ability to just access information and go across all your different accounts and then bring that into one place has saved me so much

2:29time already. One other question I have here is like basically the left pane almost like looks like my text like my iMessage and and kind of the list of my my pinned contacts and then or my pinned chats and then all of my other chats. How did you decide where to have like how do you decide which bot to talk to as well as when to spawn up a new bot with Grok

2:53bot? Yeah, I mean the interesting thing is that it's really dependent on how you like to use things and what we found at SpaceX is that everybody has different usage patterns. And so my preferred way of working with Grok bots is that I have sort of durable bots that I use all the time and then I have more like ephemeral bots that I might spin up for like one-off tasks. The cool thing about bots is that they can call other bots. So for example, I have a writing bot here that I'll use I've given it a bunch of like my writing rules and things that I like

3:25care about when I'm writing. And I really like this bot, but if I'm working with some other bots, I can have them call the writing bot. I can just say, "Oh, you know, like I'm working on this project, this like change log project, but I want the change log to sound like me. I'm going to like at my writing bot and call it." So from that sense it's like okay, well, if I have one bot that knows my writing style, I don't need to have a new bot every single time. So that's one of the ways I like to consolidate bots, and I can think about

3:54creating a new bot that's for a certain chat. For example, for this demo, I just spun up sort of this demo bot, and now I can use it for what I need to use it, and then I can hand that back off. Um more to your point about the pinned chats, these are kind of like I don't know, maybe my core six. Like I have this bot Franco, which I use for like

4:13workout tracking and things like that. And then, you know, like you've mentioned, uh I also have sort of an executive assistant and maybe a chief of staff. And those two have more context over certain things. For example, my chief of staff is really just looking at my emails, my calendar. Um the executive assistant is more for managing Slack. Uh but I do think in terms of like, can I have bots that are like durable and I use them all the time? And then can I have ephemeral bots or bots that I call

4:39for certain use cases? [music] Love it. Okay, that was easy mode. Now, let's talk about intermediate. Um it you you kind of alluded to this with um being able to call different bots, but to show me what it looks like in practice to have this multiplayer experience with a number of bots, and why like why involve multiple bots in the conversation versus just have one

5:05bot? What value does that add? There are several ways of thinking about this, but at the highest level, right? Bots are composable, which means that a bot can call another bot. A bot can create another bot, and that goes both ways. And so, there are a couple ways to think about this. If you're chatting with one bot, and you just want access to another bot temporarily, like I just alluded to, you could say, "Can you rewrite this with @writingbot, right?" And so, now what you'll see is that this bot is is thinking, but it's going to call writing

5:35bot and have writing bot perform a task. So, that's one way to think about this sort of multiplayer experience. The second way is if you have like a team, maybe you're, you know, you've worked in a team in the past, you have some managerial experience, you might bring them together in a group chat. So, we

5:49can see it just message our writing bot. And now what you'll see in the writing bot actually is that it got a message from this other bot. But, you can also think about bringing multiple bots together. So, for example, I'm writing some Twitter articles on Grok bot. And for that, I need three bots. I need my writing bot that has my voice. I need my Cursor product expert if I have questions about how the product actually

6:11works. And I need my X algorithm bot to help consult me on, you know, how to optimize an article or what time I should post it or things like that. So, I brought these three bots together in a group chat here, which you can do really just by clicking this new channel icon. We call them channels. And now I have access to all three bots at once and they can collaborate. So, really the distinguishing trait there is are three bots like talking to each other in parallel and just collaborating a bunch or do you need one bot to sort of message another bot for one time and

6:41then bring that back. For any workflow in business where multiple perspectives are valuable, just like multiple perspectives would be valuable from team members on your team, you can imagine including multiple bots in a single chat. Exactly. And and I like even what you said there because it really drives home this point of like create specialists rather than trying to go broad. Rather than giving one bot the job of like managing your SEO AEO, managing your brand voice, managing all these things, create individual bots cuz they can just call each other and that way you're separating the context and you're going to get better results. The same way you

7:15would from a human that's just thinking about one thing instead of trying to think about four things at once, right? [music] Love it. Okay, let's enter hard mode with the X to Notion integration for your CRM um of your network. Okay, sweet. This is the one you reached out to me for to like make this whole video. So, I think some backgrounds is is interesting on this one because the important thing is that I had this idea just like one morning while I was writing a video uh a video outline and it basically came together in like 10 minutes and that's really why you need to try out Grokbot and you just need to

7:50play out with your workflows because some of these things just click and you can get them done. What I was wondering is oh I'm I'm big on Twitter. I have a lot of followers. I follow a lot of people. Could you help me find my where my mutuals live because a lot of people put that in their bio, right? And then just make me a little CRM so that anytime I want to go visit some some place, I can see which of my connections are there maybe I could get a coffee with them. And I fired this off there

8:13initially this was like an audio prompt. Uh maybe like an hour later I had a working CRM. So right, this is at 7:19 and we kicked this thing off at like 6:46. So it was like 30 minutes. It just created a a list of all the people I followed on Twitter, went through, scraped their profiles, uh their handles and their locations, checked off if they were mutual follower or not and added their state. Really like did all the data extraction and formatting work here. And so you know, this is [laughter] even gave me some views. Like I can I can filter by US state and see who lives

8:49where. And so really this is just kind of like if I had like a really smart intern and I wanted to throw some project at them and like not worry about it, I could just like have that. Uh and so that experience really opened my eyes that I need to think bigger and like anything that I could do if I just had like an afternoon or a day, I should be

9:05trying to give to this Grokbot. Love it. And we were talking about this before, but in some ways I feel like personal CRM is like this low hanging fruit idea that everyone's thought of. On the other hand, everyone's thought of it because it's really important to them and no one's like ever had a great solution for it. Same thing for me. I have this database now and now I run Grokbot on a daily job where it pulls 10 people that I haven't spoken with in the last 90 days. It gives me that list of people, it reminds me on what my last conversation was with them and then it provides some

9:37conversation starters so I can add value to them. And so it allows me to keep my personal CRM warm, where before there was no way I was talking to 2,500 people on an ongoing basis. Well, look at this. We didn't even talk about how easy it was to create routines, which is just like basically tell your bot, "Hey, run every 4 hours." and it just runs every 4 hours. Or tell it to like look at Slack and it just looks at Slack. And so, part of the reason these things are so powerful, to your point, is because you could have this bot run every 2 minutes

10:06and like keep these things updated. And I just love that example that you shared because the other unmaintainable thing about personal CRMs is like information capture, right? It's like, "Oh, I talked to this person and I think their name was like Jessie, but I'm not quite sure. I remember they sit on this floor and like maybe I had a meeting with them

10:24last week, right?" All of these things. You just tell your GrokBot that, it goes into your calendar, it looks at Slack, it pulls all this stuff together, it just remembers the person that you met at work or whatever, right? And now all of that is solved, and so it really unlocks this thing of like number one, we can capture the information, and

10:39two, we can actually keep it up to date. And those were the two things that always broke for me when I tried to do these things in the past. [music] Okay, let's round this out with God mode. It's time to show how you rip some code with GrokBot in Cursor. Okay, cool. So, yeah, for God mode I really wanted to show the integration between Cursor and um GrokBot because I think this is super important. And so, I'm going to talk about this in two ways. First, how I build tech demos, and second, how I actually do like some more serious engineering and how some other members of our team do that as well. So,

11:11you know, my job is like I'm a dev devrel, I get people excited about new technology. Largely what I'll do is I'll see something on X that I think is interesting, I'll hear about some technology, and then I'll have to set aside time to go explore that thing so I understand it. And maybe from that idea becomes a piece of it becomes a piece of content or it becomes something I can

11:30share. Well, actually you can automate almost that entire process, right? Because I have bookmarks. If I could go access my bookmarks, then well, then I could get the name of the library or the package. And then what I do is draft a prompt in cursor. And then I take that and look at a preview of it. So, what this Grokbot

11:45does here is it has access to my XDM's. And it'll uh might sorry, my bookmarks rather. It has access to my DMs as well for other reasons, but it'll look at my bookmarks, find an interesting technology, and then say, "Hey Matt, do you want me to like make a demo for you?" And then I just say, "Yes." And it'll kick off a cursor cloud agent that I could open in cursor, build the demo, verify it, and take screenshots. Uh and then I hooked up Cloudflare, so it'll actually just run preview deployments of these, and I can go to a URL and play with them. Um and so I've been using

12:16this to start to understand some of these different libraries. Um and these are all just kind of like playful demo examples. It's like all of this stuff that I just wouldn't have time to do or explore, now I can. And if I see one of these and it really sparks my curiosity, I just open it in cursor. Because we have cloud agents, like the cursor app literally just pops up and I can

12:34continue prompting there. Uh and it's been a really like magical experience, and it's just I don't know, man. This I think this is the future of how how we're going to be writing code. I love it. And one thing you had also shared with me is that when you think about how a lot of uh Cursor's engineers build, um they're also using this this handoff and this dance between Grokbot uh as well as Cursor. And you you kind of talked to me about these like these two loops within the software development life cycle. Talk about that and how you're seeing kind of the most cracked engineers use these two uh in

13:09unison together. I think the the tech demos example is like my kind of nascent version of this, and then this Loops bot is my more sort of um formalized engineering version. And if you think about you know, there this term has existed in software engineering for a while, these terms, the outer loop and the inner loop. The outer loop being like testing, integration, deployment, and the inner loop being more of like how I write code, how I test that code, how I, you know, debug and do all of the like nitty-gritty. And then the outer loop's kind of more of like the implementation, right? You can think about this GrokBot

13:41being like an agentic outer loop, and then the agentic inner loop is Cursor. And so, typically, when we're building prompts and uh, you know, coming up with new ideas or bug fixes, what we do as engineers is we think, "Okay, well, like what repo am I working in? How does it work? What are similar technologies um or technologies

13:59that I could use to solve this problem? Maybe what are our competitors doing? Um how do I come up with that context, turn it into a prompt?" Typically, my prompts follow the same shape. I might use similar skills. I might use um goal or loops or like some of these other new engineering things, and then I'll

14:15dispatch that to a Claude agent. What GrokBot allows us to do is just think one level higher, right? How do we think about the meta? These are the things I always do. I'm turning that into a GrokBot. So, this bot basically helps me come up with prompts. So, as I'm thinking through ideas, this was just an update to my personal site, but I could also go like a little bit further up and probably find some like more harder like engineering stuff going on. Um what I'm doing is I'm talking with this bot to synthesize my prompts, and then it's just building the prompt

14:44for me and handing that off to Cursor. So, I'm seeing Cursor engineers do this as well, where they're thinking about the skills that they use, the context that they often use, and then just telling the bot, "Hey, do this thing for me." And if you don't follow Lauren Tan, super excellent follow, really, really, really cracked engineer, her whole thing is just like every single thing that you do, turn that into something an agent can do. And with GrokBot, we're like kind of getting there, where if you can describe exactly your process, you can start to create these bots that have access to your tools, and then run these like meta loop

15:16workflows, if that makes sense. Love it. Matt, this was amazing. From easy mode to god mode, It's It's really amazing just what's possible with Rockbot and kind of the output I get from this technology is a not a technology problem. It's a creativity problem. And so I think just spending time and kind of beating up ideas is the

15:36best way to really maximize your work. So thank you so much for joining Talk Nerdy to me and can't wait to have you on in the future. Amazing. Thanks, Alex. Thanks, man.

Where these words come from. This is the caption track YouTube holds for this video, written automatically by YouTube rather than by the creator. We read it, tidied the line breaks and laid it out so it can be read. The plain text version is at https://viewrankai.com/tools/youtube-transcript/C6eItTdAuNw.txt.

All rights in this video belong to Tenex. Watch it on YouTube. If this is your video and you would rather this page did not exist, tell us and we will remove it.