3:02:05Ray Fernando · Livestream
Live build with GPT-5.6 Sol
A three-hour first-look stream where one /goal command builds an entire multiplayer game — and then an iOS app — while the host mostly watches.
July 10thThe argument in one line.
A persistent project page that tracks status, pull requests, and agent activity separately from the chat log removes the coordination overhead that breaks a single-chat AI-agent workflow once enough requests pile up.
Read if. Skip if.
READ IF YOU ARE…
- You're running multiple AI coding agents in parallel and keep losing track of which pull request is stuck or which agent forgot context.
- You're a solo developer or small team shipping client work through cloud coding agents and want an audit trail beyond one long chat log.
- You're curious what a dedicated coordination layer for agent-driven development looks like before deciding whether to try it.
SKIP IF…
- You're not yet using cloud-based coding agents — this is a workflow upgrade for people already deep in agent-driven development, not an onboarding tutorial.
- You want a settled verdict — this is a first-day hands-on look at a beta feature, not a long-term review.
The full version, fast.
A new beta feature called Projects gives a codebase one persistent page instead of one growing chat thread: it tracks the agents currently working, the pull requests they've opened, an exec-style status board of what's decided versus outstanding, and a docs area the agent writes to for context that used to get lost. The problem it addresses is real — a single chat-based agent setup starts forgetting in-flight PRs and burns through usage once requests stack up. Projects runs agents in the cloud, surfaces PR reviews and bot comments automatically, and syncs the same project chat to a phone app. The lesson: coordination overhead, not model quality, is what breaks agent-driven shipping at scale.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:44
01 · This Thing Called Projects
Cold open: he frames Projects as a possible answer to running an AI agent as his CTO, and introduces the project-page concept.
00:44 – 01:53
02 · Grok Bot Went Crazy Viral
Backstory on his prior single-chat multi-agent setup (GrokBot), the viral video that came from it, and the usage-burn problem that followed.
01:53 – 02:55
03 · Grok Bot Forgets Sometimes
The specific failure mode: the single-chat agent forgetting an in-flight PR or a request once things piled up, prompting him to try Projects.
02:55 – 04:10
04 · The Project Page
First tour of the Projects UI: the project page, mapping out an API's shape before writing code, avoiding breaking changes for existing users.
04:10 – 06:02
05 · Agents, PRs, and Listening
How separate agents spawn per request, how PRs and their reviews surface inside the project, and how the project keeps listening to PR comments.
06:02 – 08:41
06 · The Exec Status Board
He voice-dictates a status update and walks through the generated exec status board: decisions made, work in progress, and outstanding concerns.
08:41 – 10:35
07 · Files That Live on the Cursor Side
The project's own docs/notes area, separate from the repo, that agents actually read from and that self-archives notes.
10:35 – 11:26
08 · The Same Chat on My Phone
The same project chat and agent list mirrored in the Cursor mobile app, letting him check and manage agents away from his desk.
11:26 – 12:12
09 · Still in Beta, Cloud First
Caveats: still in beta, requires cloud agents to work well, and a closing invitation for questions and future updates.
Lines worth screenshotting.
- A single chat-based AI agent workflow starts forgetting in-flight pull requests once enough requests pile up in one thread.
- Letting one agent handle everything can burn through usage quota fast, because it keeps spinning up its own background routines.
- A persistent project page that tracks agents, PRs, and decisions separately from the chat log reduces the chance a stuck request gets lost.
- An exec-style status board — decided, in progress, at risk — gives an AI agent's output the same accountability a human status report would.
- Automated code-review bots catching issues in a pull request only help if something is actually watching for their comments and routing them back to an agent.
- Storing project documentation somewhere the tooling actually monitors, instead of buried in the repo, means agents pick it up instead of walking past it.
- A workflow built once for shipping a feature can double as a reusable template that keeps generating the same style of status update automatically.
- Running coding agents in the cloud instead of locally is what makes an always-on, persistent project page possible in the first place.
- A mobile app mirroring the exact same project chat means a stuck agent can be checked and unblocked without opening a laptop.
One Project Page Beats One Long Agent Chat
WHAT TO LEARN
The failure mode in AI-agent development usually isn't the model, it's losing track of PRs, decisions, and usage once everything routes through a single chat thread, and a persistent project-level view fixes that by design.
02Grok Bot Went Crazy Viral
- Routing every AI agent request through one chat thread works, but the agents it spins up spawn their own routines that quietly burn through your usage.
- A single-agent setup can produce huge, shareable results and still hide a scaling problem underneath.
03Grok Bot Forgets Sometimes
- A general-purpose chat agent can lose track of an in-flight pull request or a spun-up background task once enough requests stack up.
- If an agent doesn't fully register a multi-minute explanation of a problem, the fix can silently fall through the cracks instead of erroring loudly.
04The Project Page
- Centralizing a project's context in one page, instead of one growing chat log, gives every new request a shared frame of reference from the start.
- Mapping out an API's shape at the project level, before writing code, catches breaking changes to existing production users early.
05Agents, PRs, and Listening
- Spawning a separate visible agent per request, instead of burying everything in one chat, makes it possible to audit what each task actually did.
- An agent that keeps listening to a pull request after opening it can react to review comments and bot-flagged bugs without a human relaying them, so automated reviewers actually get acted on.
06The Exec Status Board
- Rolling a project up into decided, in-progress, and at-risk sections mirrors the status report a human engineering lead already owes their team.
- A status page built once for shipping a feature can keep generating the same kind of update automatically for every future feature in that project.
07Files That Live on the Cursor Side
- Storing project documentation somewhere the agent's tooling actually monitors solves the problem of notes and skill files getting written but never read.
- A dedicated, self-archiving notes area can replace the workaround of piping project context into an external notebook tool.
08The Same Chat on My Phone
- Mirroring the exact same project chat and agent list on a phone app means a stuck agent can be checked and nudged without opening a laptop.
Terms worth knowing.
- Cloud agent
- An AI coding agent that runs on the tool's own servers instead of the developer's machine, so it keeps working even when the laptop is closed.
- Pull request (PR)
- A proposed set of code changes submitted for review before being merged into the main codebase.
- Exec status board
- A running summary of a project's decisions, in-progress work, and open risks, written the way an engineering lead would report status to executives.
- Feature branch
- A separate line of code changes for one specific feature, kept apart from the main branch until it's ready to merge.
- BugBot
- An automated review bot that scans pull requests for bugs and leaves comments before a human looks at the code.
Things they pointed at.
00:44linkHow To Make Your GrokBot A CTO
04:21toolConvex
06:26toolDeepgram
06:26toolGemini
08:51toolVercel
Lines you could clip.
00:00
“I finally have my answer to maybe see if I can replace my chief technical officer.”
01:57
“with Grok Bot, it will actually forget sometimes that I spun up another request or a PR is going through”
05:34
“I feel like this is the first time I feel like I've been able to delegate a little bit more here in the last 24 hours”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphor
I finally have my answer to maybe see if I can . And that's what this thing called projects. So this is announced newly from cursor and guess what?
I got access to it and I've been using it for the last day or so. And it's been really, really handy to ship software in this way. And what is really happening here is that you can create your own project page.
And this is basically one chat or one specific area where all of the knowledge for the project is saved on here. And what this will allow me to do is track everything from. like pull requests as you can see uh things that are happening that are getting merged in i can then create documentation my problem right now and i'll show you this right here is that i'm currently using grokbot and i love grokbot in fact i made an entire how to make your grokbot a cto and this went crazy viral this got millions of views from my live stream And even Elon Musk retweeted it.
And then the article already has a bunch of views here. And I actually teach you how to set up Grokbot to have multiple agents to do things. And what I've started to realize is after a little while, I noticed that the agents set up a lot of routines and a lot of stuff that's going on.
And it starts to pretty quickly. So a lot of people who were following my workflow ended up having to like eat up their usage super fast. But...
The advantage of it is that now you have one central place that you're basically doing all the communication with. As you can see, I've been shipping software like crazy, literally just using a chat window just within Grokbot. And so my single Grokbot is actually managing this whole thing, and this is my direct responsible agent.
And I wanted to see with the new projects, can I ship a feature and have it keep track of everything on here? The problem I started to really experience with... using Grokbot.
And this is where I think projects is going to be really interesting for us here is the fact that with Grokbot, it will actually forget sometimes that I spun up another request or a PR is going through, or maybe it just didn't fully understand everything that I was saying for three minutes as I was trying to describe a problem.
And that's kind of what I was running into. So I just said, let's just go ahead and add this new feature. And so this is going to add an API to my whole transcription app.
I have a transcription app and it's really great because it has a great user interface to it. And what we're doing here is a lot of different workflows because the code touches a lot of the back end and I have to deal with a lot of complexity. And I'm onboarding my first like enterprise customer, but he's actually just a client who's using this for another enterprise client.
And what we're trying to do is try to map out what the shape of this API will. actually end up looking like and so that it doesn't also break my existing production users and as we can see here this is the way that the project layout is kind of shaped what we have here is our basically our project page if we go view project page this is kind of what we're brought to and this little thing up here is called project and this kind of describes the agent is kind of writing to this specific area to let you know and let itself know kind of what's going on and i've given it a lot of different requests and it spins up its own cloud environment so this specifically works in the cloud and you can select the model that you want to work with in the cloud for this type of thing I also want to give you a heads up to here as far as agents.
When you're having a communication in this chat for this project page, this is where all the agents will be spawned out from. So as I've been able to send information or send different requests. as you can see here this is all from my grokbot all in like one day and i haven't even read through them because there's so many of these requests and there's so many things this comes from brockpot so i've shipped so many features just in one day just from talking that as you can see that's just a pain in the butt to keep track of and i'm not really trying to do that i think projects are a really good place if you're used to a software development workflow and you want kind of everything to kind of land in one place and so In this area here, as kind of what I was pointing out, is that you have an agents area.
So this shows you all the agents that have been spawned up to take care of your various requests. And if you want to go through their different chat workflows and understand what's going on, you have the ability to deep dive there. You can see all the PRs that have been open on your requests, and you can actually review them if you click on them here.
And they can tell you that the feature has been merged in or whatnot. What's nice is that if there's any comments as well, you'll see the comments that are done there. And if there's any reviews, it's all kind of done there as well.
So I have these GitHub actions that kick off as part of my workflow. And what is great is that cursor's continuing to monitor them. And so you see the things here is called listening.
So this is what is actually kicked off inside of a project. And so this will allow the actual agents to listen to the comments that show up in these PRs. So as comments are coming in, and all my different checks are going.
If there's a new bug that's called out by BugBot or there's another agent that calls out something in your own review flow, I want to make sure that that gets addressed because sometimes that was not getting fully picked up because my single GrokBot that was basically doing my CICD would have so much stuff coming in, it would sort of get overwhelmed.
And while it can handle a lot of tasks, I just didn't want to leave everything to chance because stuff is actually touching a production workflow. And I feel like this is the first time I feel like I've been able to delegate a little bit more here in the last 24 hours to a project page and keep track of everything. So the next thing you're probably thinking is like, okay, now that I've been able to hand off more of the work to this nice workflow, how do I kind of wrangle everything so I can keep track of things?
Because if you're like me, trying to scroll through this big entire chat can be a little bit cumbersome and exhausting. I'm not really all about that. So one of my favorite workflows as an executive is to try to get what I call like an executive summary or high -level overview, or like some people say is a status.
And so if you've ever had to do these before as an engineer, you basically have to roll up everything that's been happening in your project for the past week and kind of roll up what's the risks, where are you as far as red, yellow, green, if there's anything that you need to call out that's still a risk to the project that you need to ship or something.
This is kind of your opportunity to do that. And then you also need to make sure that everyone's kind of on top of the work that they're saying that they're supposed to be done. And so I kind of do a similar treatment here.
And I just speak to my agents just with this little microphone to basically give up a little bit of a status report of some sort. So we can see the documents that are here. This is called the exec status board.
So this gives me the open decisions. And this tells me what's going on for the feature that I'm currently shipping. I can hit this little.
expand to full screen mode here and this will give us that there and if i kind of push this to the side we can really just start focusing on our project here and in our project you can kind of we can kind of get a sense of what this status board is kind of looking like so this is my current goal to try to get this integrated for my my customer here and then this is my current shipping branch that i have everything on and this is my preview branch that i have set up with vercel and all of this stuff is done in the cloud so this is what we've decided on this is currently what's work has been done.
I also have a layout of stuff that's been in progress and then there's any outstanding concerns. This is going to be what's showing up for me. Also, if there's also any decisions that are in progress, sometimes this gets lost, especially if you have a lot of agents that are firing off.
I can kind of miss some gaps or some stuff can become missing because it maybe flew away in my chat and I never followed up with it. So there's an opportunity for this so that I can come back into the loop and make sure that either something stays on track or it needs to be delegated to another agent and kind of follow through end to end.
So this is kind of. what I set up for myself, but this setup is literally can be adopted to whatever your workflow is. An advantage of maybe doing something through a project is now I can ship new features through here.
And when I ship new features through here, it's gonna keep. giving me executive updates for those features going forward. So that's a kind of really nice surprise that I wasn't really expecting is the fact that you can kind of reuse a lot of these templates that you're using for shipping software from your project, especially if you're just continuing to touch this repo and so forth.
That I didn't really get consistently. And this is something if I'm going to be spinning up new projects for new repos, I should be able to somehow transfer the knowledge in those workflows into those workflows just by describing in plain English. and just getting like one specific document to hand over to the next thing.
So speaking of documents, if you click this little book icon here, this will show you the current files in this project. So these files are not necessarily in my repo. This is something that lives on the cursor side as far as documentation or things that are important for the agent to keep in context.
So the promise here is that everything's supposed to be kind of kept in memory for a long time or forever, and this will continue to compound. So if you have been shipping... software a lot of times i write stuff like skill files or documentation or i have specific areas in the repo itself that will stick in there but not a lot of times the agents will pick them up they won't pick them up so like agents will then fly by them and not really do anything with them And in here, you'll have all this information saved.
There's like even a notes section, and you can just type in the notes and do whatever you want here, and it sort of self -archives. So previously, I used to hook up Notion or some other types of notebooks to try to keep track of this, but now it's kind of nicely contained into here, and this is all tied to your project. So if we go back to the tabs up here and scroll all the way to the left, we'll go back to our original project, and this is the little notebook that's currently open, and you can see some of these documents that are down here.
So if I click this again, I go back to this nice little view here that it's generated for me, for my agent. And then I can click any one of these and this will take me right directly to the PR or anything like that. So this project view is kind of telling me what's going on.
And if I go back to this little minimize section up here, what you might notice is that there's actually like, you're actually connected to a cloud. So the cloud that you're connected to. is literally just, you know, cursor zone cloud.
If you see down here, there's a little cloud icon. So this shows you that you're connected to the cloud icon. And this, you know, is giving me a currently update of what's going on.
So there's a lot that's going on here. And this workflow for me seems to be like a really nice way to kind of have everything kind of contained into one area and just basically be able to use one chat. So that's the other part of doing this workflow is the fact that I can.
have everything on my phone so if i launch the cursor app on my phone and i basically just say hey go into here to go to cursor i can actually see a project section so there's literally just a project section that's actually identified there and if i click that project that's the same chat that you're seeing literally on this screen up here so that same chat is basically what i can use and then i can see keep track of the agents that are down there below there's this little glass if i tap the agents i'll see all of the agents that are currently running and those are the agents that i can tap into there and start to manage itself so you can kind of you know herd them together here and it just makes it easier to kind of go back and forth with them and have this conversation this is a really cool workflow this is still in beta and i'm still like kind of getting my hands around what this is looking like but this is kind of my first look and if you've been curious about trying to try out this new projects thing i believe they have a sign up list but i'm not entirely sure i think they're going to slowly start rolling this out
It does require the cursor cloud agents. So if you're using a project and it's not on the cloud, it tends to work really well in the cloud first. Or if you can put your repo onto the cloud and have it manipulate and do those things, I think this is going to be the best way to do that.
But if you have any questions, definitely drop them in the comments below and I'll be kind of giving you some updates as far as what's going on with this app and how this continues to progress because I'm starting to learn more workflows. And I wanted to see if I can hook this up to Slack. It should be able to interact that way as well.
And I think there's a lot of other features that I'm not really sure yet. And I'm going to keep learning and sharing back with you. So I definitely appreciate it.
And I'll catch you in the next video.
The Hook
The bait, then the rug-pull.
Ray Fernando got early access to Cursor's new Projects beta and spent his first day shipping a real API feature for a paying client through it, testing whether one persistent project page can do what a single sprawling agent chat couldn't.
Named ideas worth stealing.
06:26concept
Exec Status Board
- Decided
- Timeline and capacity
- Open review comments and bugs
- Done (merged)
- Shipping branch
A living status document the coding agent maintains inside the project, mirroring the roll-up a human engineer would give leadership: what's decided, what's in progress, and what's still a risk.
Steal forany solo developer or small team using AI agents who needs an audit trail without babysitting one long chat thread
How they asked for the click.
VERBAL ASK
11:35link
“if you have any questions, definitely drop them in the comments below”
The only spoken CTA is a soft ask for comments; the actual funnel link (a free community at rfer.me/startmyai) appears only in the video description, not spoken on camera.
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Visual structure at a glance.
