Greg Isenberg gets a live, screen-shared tour of Jack Dorsey's new agent-native chat app from an early user — and presses him on whether it actually beats Slack.
Posted
yesterday
Duration
Format
Interview
educational
Views
9.3K
396 likes
Big Idea
The argument in one line.
Buzz's real advantage over being called a 'Slack killer' is that it's built on an open protocol where agents are first-class teammates, so a team can swap the AI model or coding harness under any agent without losing shared chat context.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You run a solo business or small team (roughly 1-10 people, under $10M revenue) and already coordinate work in Slack or Discord.
You use AI coding harnesses like Claude Code, Codex, or Goose and feel 'model fatigue' from re-explaining context every time you switch tools.
You want a concrete sense of what agent-native team chat looks like in practice, beyond the marketing claims.
You're evaluating whether owning your AI model choice and data matters for your business's long-term flexibility.
SKIP IF…
You need production-grade, complex software engineering tooling today — the guest is explicit that Buzz isn't there yet.
You just want a definitive verdict — this is an early, honest first-look, not a finished review.
TL;DR
The full version, fast.
Buzz, Jack Dorsey's new open-source chat app from Block, gets called a Slack killer, but its real bet is openness: agents are first-class teammates, not integrations, and the AI harness under any agent swaps freely between Claude Code, Codex, and Goose without losing chat context. An early user demos it building a live CRM app and a tweet-analytics workflow, shares shared-compute for pooling local models across a team, and flags that data lock-in with any one vendor (Slack or an AI model) is a real long-term risk. The honest verdict: it's early, sometimes slow, and not for complex engineering yet, but worth trying now if you're a solopreneur or small team, because openness means it can improve faster than a closed competitor.
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.
Greg frames Jack Dorsey's new agent-native app as suddenly viral and brings on early user Vinny to give a live tour and settle whether it actually beats Slack.
02:57 – 03:49
02 · Agents as First-Class Team Members
Vinny's core pitch: Slack bolts agents on as integrations, while Buzz makes them full team members from day one, each with editable instructions like a system prompt.
03:49 – 06:55
03 · Swappable Harnesses Under Any Agent
Every agent's underlying harness — Claude Code, Codex, Goose, or OpenCode — swaps without losing chat history or project context, which Vinny frames as the cure for model fatigue.
06:55 – 08:34
04 · Audio Huddles With Agents
Buzz supports live voice calls where an agent joins carrying full chat context; Vinny explains it without a live demo, and Greg shares his dream of eventually FaceTiming an agent.
08:34 – 11:20
05 · Git, Feature Branches, and Parallel Worktrees
Agents build in isolated parallel worktrees so nothing touches local files directly; Buzz also runs its own Git hosting on a self-hosted relay, extending its ambitions past Slack toward GitHub.
11:20 – 13:43
06 · Building a CRM App with Buzz and Agents
Vinny asks Buzz to spin up a CRM with the Wasp framework, deployed to Railway; the agent returns a live app, screenshots, and links with almost no back-and-forth, then the team keeps iterating on it in the same channel.
13:43 – 18:13
07 · From Chat to Shipped Workflow
The conversation widens to agency use cases like auto-generating client proposals from meeting notes, then Vinny shows his own build: a tweet-performance dashboard piping daily stats into a Buzz channel via a public API for the agent to analyze.
18:13 – 23:53
08 · How to Actually Talk to the Agents
Vinny's advice is to just talk normally — the agents infer intent well. He walks through a real report his agent generated on his tweet performance, then Greg asks about Buzz's rumored Bitcoin backdoor.
23:53 – 25:42
09 · Shared Compute and Local Models
A team can turn on shared compute to run one local model on one member's laptop and let everyone else's agents use it remotely — Vinny frames it as a way for cash-strapped teams to pool one good machine.
25:42 – 27:36
10 · Model Choice, Data Ownership, and Lock-In
Vinny connects shared compute to Buzz's larger bet: a single AI vendor can restrict or reprice a model overnight the same way Slack locks a team's chat history inside one platform, so owning model choice protects the business long-term.
27:36 – 29:39
11 · Context as the Foundation
Both hosts land on the real differentiator: Buzz treats the shared chat as a context foundation that every other feature plugs into, which is what they think a prior tool, OpenClaw, was missing.
29:39 – 31:21
12 · Setting Up and Managing Agents
Vinny's setup advice: pin cheaper tasks to a lighter model and save the frontier model for hard problems, and consider building a 'Chief Agent Officer' agent whose job is routing tasks to the right specialist.
31:21 – 33:03
13 · Skills Carry Over, and the Software Is Rough
Globally installed Claude Code skills work inside Buzz automatically. Vinny is candid this is early alpha-or-beta software — recurring workflows didn't land well and the app feels slower than a direct CLI harness.
33:03 – 34:36
14 · Who Should Try Buzz Today
Vinny's honest fit: solopreneurs and small teams iterating on small ideas, not complex software engineering. He'd use it at his own startup for private team channels plus public channels where the community can chat directly with the team's agents.
34:36 – 38:10
15 · Greg's Take: A Glimpse of the Future of Work
Greg calls Buzz an early but genuine glimpse of a workplace with more agent teammates than human ones, discloses he has no affiliation with Buzz or Block, and argues that even trying it and going back to Slack still teaches you something.
38:10 – 38:44
16 · Closing Thoughts: Open Source as the Edge
Vinny closes on why being open source and built on an open protocol could be Buzz's real advantage — anyone can improve it, so the product evolves through its own users rather than one company's roadmap.
Atomic Insights
Lines worth screenshotting.
Buzz treats AI agents as first-class team members with editable instructions, not bolted-on integrations the way Slack does.
The AI model or coding harness under any Buzz agent can be swapped — Claude Code, Codex, Goose, OpenCode — without losing that agent's chat history or project context.
Buzz agents work in isolated parallel Git worktrees, so they can build multiple versions of something at once without touching your local files.
Buzz runs its own Git hosting on a self-hosted 'relay' server rather than only integrating with GitHub, extending its ambitions past chat and toward code hosting.
A single well-specified request produced a live CRM app, deployed to Railway on the Wasp framework, complete with a returned link and screenshots, before the user even reviewed the code.
Piping a live app's data back into the same chat where you talk to your agents, via a public API, turns a one-off build into a closed feedback loop you can question in plain language.
"Novelty gets you reach. Friction gets you engagement" — the user's own tweet data showed his most viral posts got the least engagement, and vice versa.
Buzz is built on Nostr, an open protocol tightly coupled with Bitcoin Lightning, which the guest expects will eventually let people tip each other or pay agents for compute directly inside the app.
A team can turn on shared compute to run one local AI model on one member's machine and have every teammate's agents use it remotely, which lets cash-strapped teams pool one good machine instead of buying individual premium plans.
Just as Slack locks a team's chat history inside one company's platform, depending on a single AI vendor risks losing access or facing a price hike with no way to move your data elsewhere.
Globally installed AI coding skills carry over into Buzz automatically, so an existing Claude Code setup doesn't need to be rebuilt from scratch.
The guest is candid that Buzz is early alpha-or-beta software: recurring scheduled workflows didn't work well for him, and the app feels slower than working directly in a CLI harness.
The guest's fit assessment: strong for solopreneurs and small teams iterating on small ideas, weak for complex, high-stakes software engineering.
Being open source and built on an open protocol means anyone can improve Buzz, which could let it evolve faster than a closed competitor even from a smaller starting team.
Takeaway
The Real Lesson Isn't the Slack Comparison
AGENT-NATIVE TEAMS
Buzz's real bet isn't the Slack-killer headline — it's a portable, shared context between you, your team, and your agents that outlasts any single model or vendor.
02Agents as First-Class Team Members
Tools that bolt AI agents on as an integration treat them as an afterthought; tools that make agents full team members with editable instructions treat them as an operating model, and the difference shows up in how naturally the team actually uses them.
An agent's 'instructions' field is functionally a system prompt — writing one well is closer to onboarding a new hire than configuring a plugin.
03Swappable Harnesses Under Any Agent
Decoupling the chat layer from the model or harness underneath means switching AI models doesn't cost you your project history, so 'model fatigue' from re-explaining context stops being a tax you pay every time a better model ships.
If you're juggling multiple AI coding harnesses across projects, look for tools that let you swap the engine without losing the conversation — that portability matters more day-to-day than picking 'the best' model.
04Audio Huddles With Agents
Live, spoken interaction with an AI agent surfaces ideas that typed back-and-forth doesn't, particularly if your thinking works better out loud than through drafting and editing text.
An agent joining a live voice call with full context of the existing chat is meaningfully different from a fresh voice session — it remembers what you were already building.
05Git, Feature Branches, and Parallel Worktrees
Letting agents work in isolated parallel worktrees rather than directly on your local files means multiple agent-built variants can run at once without risking your working copy.
A tool that hosts its own Git remotes, not just a GitHub integration, is betting to become infrastructure rather than just a chat client — worth noticing when picking where your code and data live long-term.
06Building a CRM App with Buzz and Agents
A well-specified one-shot request — name the framework, the host, the deliverable — can produce a live, deployed, screenshotted app with almost no back-and-forth, changing the cost of prototyping an internal tool from days to minutes.
Getting links and screenshots back automatically, before you've even reviewed the work, is a meaningful trust signal that an agent workflow is mature enough for non-critical internal tools.
07From Chat to Shipped Workflow
Any recurring 'boring but essential' business process — client proposals from meeting notes, a performance dashboard — is a strong candidate for an agent workflow because the value compounds every time it runs.
Piping a live app's data back into the same chat where you talk to your agents turns a one-off build into a closed feedback loop you can question in plain language instead of re-exporting data by hand.
08How to Actually Talk to the Agents
You don't need special prompt engineering for day-to-day agent requests — talk to it like a normal person and trust it to infer intent, then get specific for tasks that need precision.
'Novelty gets you reach, friction gets you engagement' is a data-backed content rule worth remembering: what makes something go viral once is not what makes people stick around and engage repeatedly.
09Shared Compute and Local Models
A small team that can't afford individual top-tier AI plans can pool money into one capable machine and share its local model across every member's agents, instead of everyone paying separately for hosted access.
Local and open models are closing the gap with frontier models fast enough that 'share one good machine' is now a real cost strategy, not just a hobbyist workaround.
10Model Choice, Data Ownership, and Lock-In
Treat AI model choice like platform lock-in: a single vendor controlling your only usable model can change price or access overnight, often with no clean way to take your data elsewhere.
Government-level restrictions on what a model is allowed to answer are a real, current risk — building a business entirely on one closed model means inheriting whatever restrictions get added later.
11Context as the Foundation
The shared conversation history between you, your team, and your agents is the actual product — every other feature is just a way of acting on that context, so protecting and centralizing it matters more than any single feature.
Tools that produce a 'wow' moment but don't solve shared context between teammates tend to fade once the novelty wears off — durable agent tools need a place where the context accumulates.
12Setting Up and Managing Agents
Pin your cheaper, high-volume tasks to a lighter model and reserve the frontier model for genuinely hard problems — burning premium tokens on routine requests is the most common waste in a multi-agent setup.
As your number of agents grows, a 'router' agent whose only job is delegating each task to the right specialist is a simple fix for forgetting which agent is good at what.
13Skills Carry Over, and the Software Is Rough
If your AI coding setup already has globally installed skills, check whether a new tool can inherit them automatically — rebuilding your skill library from scratch for every new app is unnecessary friction.
Early or beta agent software is often honest about its own gaps rather than hiding them — take a builder's own list of what doesn't work yet as more trustworthy than the marketing.
14Who Should Try Buzz Today
Match the tool to the job: a shared-context team chat with agents is a strong fit for solopreneurs and small teams iterating on small ideas, and a weaker fit once you're doing complex, high-stakes software engineering.
Separating public channels for your product's community from private ones for your internal team, inside the same agent-aware chat, lets outside bug reports become an '@agent' away from being prototyped instead of sitting in a queue.
15Greg's Take: A Glimpse of the Future of Work
Judge genuinely new categories of tool by whether trying them teaches you something, not only by whether you end up adopting them — you either find a new default or learn enough to configure your current tool better.
A market this early rewards hands-on experimentation over waiting for the 'finished' version — the tools worth watching are usually still rough when the people who matter start talking about them.
16Closing Thoughts: Open Source as the Edge
An open-source app built on an open protocol invites outside contributors to fix what's missing, which can let it improve faster than a closed competitor even from a smaller starting team.
When a tool's roadmap is effectively crowdsourced, expect it to bend toward whatever its most vocal power users are missing from the tools they're used to — that's an advantage, but also unpredictable.
Glossary
Terms worth knowing.
Buzz
An open-source, agent-native team chat app from Block (Jack Dorsey's company) where AI agents are full members of the chat, not add-on integrations.
Nostr
The open protocol Buzz is built on. It's decentralized and tightly coupled with Bitcoin Lightning, which is why Buzz is expected to eventually support native Bitcoin payments.
Relay
A server, self-hosted or hosted by Block, that stores a Buzz team's chats and data and hosts their agents' Git repositories — the backend a team's information actually lives on.
Harness
The underlying AI coding engine running behind a Buzz agent, such as Claude Code, Codex, Goose, or OpenCode. It can be swapped without losing the agent's chat history.
Worktree
An isolated Git working copy an agent creates to build or test changes in parallel, without touching the files on your actual local machine.
Shared compute
A Buzz setting that lets one team member's machine run a local AI model and share it with the rest of the team's agents remotely.
Huddle
A live audio call feature in Buzz where an AI agent joins as a participant, carrying the full context of the existing chat into the conversation.
“In Buzz, they're like first class citizens, right? So they're members of your team.”
One-line pitch that reframes the whole 'Slack killer' premise into something more specific.→ TikTok hook↗ Tweet quote
05:55
“The great thing here is that your agent, you don't lose anything in that switching process.”
Names the exact pain (model fatigue) and the fix in one breath.→ IG reel cold open↗ Tweet quote
23:40
“Novelty gets you reach. Friction gets you engagement.”
A punchy, standalone content-strategy rule backed by the guest's own tweet data.→ newsletter pull-quote↗ Tweet quote
25:00
“Some random person in some graphic design community on Buzz gives you an awesome logo and you just boop, tip them.”
Vivid, concrete image of the Bitcoin-tipping future the app is teasing.→ TikTok hook↗ Tweet quote
25:48
“Slack is the controller of your data, and that data is very valuable.”
Blunt one-line articulation of the platform lock-in argument.→ newsletter pull-quote↗ Tweet quote
33:10
“This is beta software. It's like it might even be alpha. I'm not sure.”
Rare moment of unvarnished honesty about the product's maturity, cuts through the hype.→ IG reel cold open↗ Tweet quote
35:10
“I have no affiliation with Buzz or Block or Jack Dorsey, but I do think it's worth trying these tools.”
Disclosure line that earns trust right before the episode's thesis statement.→ newsletter pull-quote↗ Tweet quote
38:38
“Things change so fast, so who the hell knows?”
Punchy, honest closer that undercuts any overclaiming from the rest of the episode.→ TikTok hook↗ Tweet quote
Topic Map
Where the conversation goes.
00:00 – 03:49denseWhat Buzz is and why 'openness' is the real pitch
03:49 – 06:55denseSwappable AI harnesses and portable context
06:55 – 11:20steadyHuddles, Git worktrees, and projects built in
11:20 – 18:13denseLive demo: CRM app and tweet-analytics workflow
18:13 – 23:53steadyHow to prompt agents + Bitcoin/Nostr tangent
23:53 – 27:36denseShared compute, model choice, and data ownership
27:36 – 31:21steadyContext as the core product, agent setup advice
31:21 – 34:36denseHonest state of the software and who should use it
34:36 – 38:44steadyClosing takes: future of work and open source advantage
The Script
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.
17px
metaphorstory
Jack Dorsey just launched Buzz and it's gone completely viral. Yes. It's the same Jack Dorsey who co founded Twitter, who co founded Block.
This guy is building what people are calling a Slack killer. Now, why does this matter for you? Well, it's being called a Slack killer because it's almost like this agentic version of Slack.
It's a way to have, uh, conversations with your teammates except their agents.
It's a way for you to huddle agents. It's a way for you to have software being built on the fly via agents.
So in this episode, we're going to break down it so clearly for you so that you understand what it is, why you should use it, and what are the best tips and tricks to use Buzz. My take on this whole thing is it's a glimpse into the future of work.
I can't wait to see what you think about it. And the by the end of the episode, you'll understand if it's for you, why you should use it, how you should use it. Enjoy the episode.
What's up, Vinny? I saw your tweet.
Jack Dorsey retweeted it. By the end of this episode, what are people gonna learn? They're gonna learn how to use
Buzz. Mostly, I think it's like the sweet spot for Buzz right now is I think solopreneurs, small teams.
They're going to learn how to use it to kind of brainstorm, keep everything in context between their agents, and just build products, build solutions, ideate, and like, yeah, use it to to get stuff out there quicker and be more productive.
Cool. Okay. So what I what I hope to get from this is basically understand what Buzz is, understand why it matters, understand some best practices on how I can use it to actually use agents to do things for me.
And by the end of this episode, I want to understand if I should actually use this thing. Or should I just basically continue using Slack? And that's what I hope to get out of this.
Do you think you think you could commit to that? And also, while you're going through it, just share the best practices that haven't been shared anywhere so that, you know, we we've we've got a little edge on people.
You think you can do that? Yeah. I think so.
I just wanna start off by saying, like, I think the Slack killer yeah. It's a good line that it's a Slack killer. But I think the real, like, the real selling point for me for with Buzz, like, from the beginning is just the the openness of it.
So do we wanna start there? Just, like, what I mean by open and and what kind of, like, what that allows?
Like, the power how powerful the openness feature of Buzz is? Yeah. I mean, sell me on Buzz.
Like, why Okay. Okay. Yeah.
So first off, is Slack with agents inside and a lot of people like, it's kind of like Slack.
You got agents inside. So Slack allows you to do that by adding an integration.
But in Buzz, they're like first class citizens, right? So they're members of your team.
You don't think of them as like add ons or just kind of a side feature.
They're an integral part of the app. They're part of your team.
Think about it that way. So I'm kind of I like to keep things simple.
And I just started using this not too long ago. I mean, it's only been out for a little while. I just discovered it.
And these are basically what you see here are the default agents you get when you start. But under the surface, there's a really powerful thing, which is you can edit and add agents. And when you give agent instructions, that's just basically like a system prompt.
The really cool thing underneath is the harness. So we're all probably using codecs, Cloud Code, maybe something like OpenCode to do our coding work, build products, whatever.
Be more productive. The great thing here is that your agent harness is swappable under any agent.
So I can go under here and say, I want to use Cloud Code. I want to use Codecs. I want to use Goose.
And, like, for me already, that point there was like, okay, this is very cool. And the reason why this is so cool is because, I mean, if any of you, all your listeners probably follow this stuff, AI moves at a breakneck pace.
Right? So today, it's this model. The next day, it's the other model.
And sometimes, I don't know about you, but personally, I get model fatigue where I'm just like, gosh. You know, I I just switched over to this.
I just got used to this model. Like, I don't even want to hear the good things about model x because I'm still over here using model y. But the cool thing about this is that this is a layer on top of that and you can just switch.
And the best part is is that your agent, you don't lose anything in that switching process. You can change the harness under your agent and all the context that's in your chats. So everything you've been chatting about with this agent, like all the stuff you're trying out, the apps you're building, that all of that all that context comes with to that new harness, to that new model.
So first off, that's something I think is really cool about Buzz.
That's a that's a huge deal because I feel like a lot of us are ping ponging between different models and harnesses and stuff like that, and it's tiresome.
You know? When you're when you're starting from scratch, it does feel like and then I need to understand, okay, what is Sol good at too, you know?
Like, and what are the context for this particular model to get the most out of it? So I think the fact that this is built natively into the product that you can go and switch, just like, I like that because it's thinking about how people are working in today's day and age, and they're acknowledging it, and they're like, we understand that, yes, there's gonna be new models all the time, and you might wanna be switching, but you probably don't want to spend a lot of time adding context, figuring out things, because you're busy trying to run a startup, trying to make money, trying to be productive.
Yep. Yep. Exactly.
And you can take all your stuff with you. Know, like if you're if you've built all these skills and skill files and folders and stuff, you know, as long as they're in like the global directory, if they're accessible to globally and they're not project specific, like Buzz can access them.
So you can install agent skills like normal. You can change then you can change harnesses, change models, and and your your agents in the chat have access to all of them. So I think that's really cool.
I'm really happy about that. Okay. What else do we what else do we need to know?
We've got audio huddles with agent. So that allows you to have like an audio chat with your agents. And they'll be in the chat and they'll respond to you.
And so you can, you know, have a voice call with someone else, bring an agent in and communicate. And they also get the context of your whole chat in the audio huddle.
So that's pretty cool. I haven't tried that out yet though. So I have seen some good videos on YouTube that run through it and show you how to do it.
So What's cool about huddling
an, you know, agents or an agent I I tweeted I tweeted about this. I feel like what I'm missing from building in the AI age is I feel like I'm asking my agents to do things.
Maybe I'm using WhisperFlow or even just typing it, and it feels very much like back and forth, back and forth. What's cool about a Huddl is it's live.
And I I the way my brain works when I'm when I'm being creative and stuff like that is I need, like, live interaction in order to really, like, extract what's going on in my brain.
I think there's probably a lot of people like me like that, so it's cool that that that's a feature. My dream, and I tweeted this, my dream is eventually to have, like, the ability to FaceTime agents.
And that to me is like the holy grail of the future of work. But, you know, cool that you can see that Buzz is scratching the surface here.
Is it good or if it's not good? We'll have to test it. People in the comment section, please let us know.
But it's interest it's interesting that that feature exists. What what's another feature that you're excited about? I mean, well, right under the surface, you've got these agents, and you probably want them to do work.
So they're like integrating they have Git pretty tightly integrated into it.
And your agents will actually you know, they're doing coding. They can create projects.
They'll create feature branches, and they work in parallel work trees. So they're not actually messing with stuff that's locally on your computer.
They'll make a work tree, a copy, and they can work there. So they can do things in parallel. So you could, for example, you're talking to your teammate about a UI, the design of your product or your web page or something.
And you can just tell the agent spin up three different versions of landing pages based on the ideas that you guys had in the chat, right? And they've just added this projects view.
So if you go into settings here, there are experiments. You can turn on workflows and projects, for example.
And then you'll see that here. And this is basically like your apps that you're coding on with your agents and working on with your agents.
And another interesting thing is that you can integrate with GitHub, but under the surface they have their own hosted Git hosting.
So you can push to remote repositories that are actually on your relay.
So that's probably a more technical thing, but Buzz works on relays. So these are like servers that you host and or that get hosted for you by Buzz, by Block.
And then they'll actually push that code to the remote repositories on your relay. So the relay is where all your information, all your chats, everything, that's where all that stuff is happening and where it's stored and where it's getting sent and pushed to the other people chatting with you and the agents chatting with you.
So I can see in the future that they're trying to take on more than just Slack here. They're trying to take on
GitHub as well. So that's pretty crazy. Cool.
And I think, you know, for the nontechnical audience, I feel like a lot of my nontechnical friends are actually starting to use GitHub now in the AI age. Don't know if you've noticed this.
No. I I I'm just kind of in the I'm a I'm Bubble. Mostly mostly in the, yeah, the nerd bubble.
Yeah. It's it's interesting. It's just I feel like, you know, five years ago, if you were nontechnical, ten years ago, if you're nontechnical, you barely knew what GitHub was.
The word repo, you're like, are, you know, are are you repossessing my car? You know?
But I think now in an AI age where we're kinda all technical in some way, something like this makes makes a lot of sense to be built natively.
Yeah. And the the interesting thing is you don't really have to know that much about it. You don't really have to even consider it because your agents will just kinda take care of stuff for you.
Like I was doing stuff here where I was asking you, you know, pull up a simple CRM app and, you know, I I told it here's the chat right here.
Spin up a simple CRM app using the Wasp full stack framework and deploy it to Railway for me. So there's some technical knowledge there.
I'm asking for a specific framework, a specific hosting provider. But it'll do that and even gets stuff, pushes it to these remote repositories, deploys it live on the web, and even sends you screenshots.
So it's like, oh, here's your CRM. Here's your that I just made.
And that's all before I even checked out what it was even doing. So it just it did it for me, gave me the links, gave me a little preview of it. And yeah, I thought that was pretty cool.
That's ins that's insane. Yeah. There's a lot of I feel like they've baked in a lot of workflows and a lot of just prompts into these agents so that these agents can work end to end and really get tasks done.
I think that's the idea. Yeah. That's insane.
Like, you think about it. Like, we're kind of, like, glossing over it. But, like, look what you've got.
Yeah. And it's, like, it's live on the internet, right, already. So the nice thing is, you know, I already had a Railway account.
I, you know, I know a little bit about working with full stack apps, but I didn't really tell it much. I just said, you can see the chat, put this online.
So now I have a fully functioning CRM dashboard that I can share with team members. And then we can the cool thing is then what you would do is you'd add them to this chat here.
And you start talking about what features are missing, what things you don't like, like and then you guys, you you know, you have a conversation with your teammates and you come to a a conclusion, and you just tell your agents to go ahead and start working on it, and they'll do it.
So why this is a big deal is if you think about it,
you know, you're able to create software on the fly, essentially automated, that's high quality, that could be deployed, that you can manage in a view that, you know, is clear.
And I think, you know, if you think about your business, let's say you work you have an agency business, proposals is, you know, the lifeblood of your of your business in a lot of ways. Right?
And, you know, imagine that you can create proposals based on people's data.
Maybe you're pitching the NFL, and you're you know, you could put in the context of the pitch meeting notes, let's say, from granola or something else.
It ingests it, and then on the fly, you have these agents based on that create proposals that are unique to those clients. And it's vibe coded or or it's using agents to actually, you know, create them.
You know, there's so many ways that you can use this interface to think about, okay, I have, you know, context over here.
This is how it comes in, And then here's how it's gonna get spin up on the Internet or through, you know, an asset that is built within code. And then here's how you're gonna view it.
Is that right? Yep.
Yeah. Exactly. In one of the of the videos I made that I posted on Twitter, that was what I was going through.
So I was taking more of like a a marketing angle and I built a tweet leaderboard for for Twitter.
So like my team, like how much we're tweeting, how good, what are our top tweets, all that stuff, all that information was being collected in this app, this Fullstack app.
And then I was like, oh, well, um, instead of just building it with Buzz, why don't I get Buzz to also access the app, get that information, then we can get like, you know, daily reviews here, daily digests. And, um, we can also like kind of brainstorm on what's working in our Twitter strategy for marketing, what's not working, get it to find links between the popular tweets and brainstorm, come up with new stuff.
So that's exactly what I did. I had it I told it let's build a publicly facing API, and then let's build a workflow.
So I just tell it daily, check that API and give me the numbers back.
And then the awesome thing is that's what you're seeing right here in this tweet stats for app from app. It's getting the stuff from that dashboard through an externally facing API and putting it here into this channel.
And then I can reply to it and be like, Okay, Fizz, which is one of the agents, like Fizz, let me know what's the common thread between my successful tweets this last seven days?
Because it also sends those over the API. So if I understood you correctly, I think you're talking about kind of these workflows where you're like this circular stuff where you're giving information in, getting it out through apps, and it's all going into the context of your your agents and stuff like that. Exactly.
Yeah. Yeah. I mean, I'm interested in this because this is like the quote unquote boring stuff that every business has that if you can figure out how to have some unfair advantage here.
It's it's what separates a good business from a bad business.
Mhmm. Yeah. And and that's the thing I really like is that this openness of buzz, like, it allows this kind of these kind of integrations to happen very easily.
So with Slack, would have to to get all this kind of stuff up and running, you'd have to get API tokens, create an app on their platform.
All their stuff is proprietary. You have to learn the structure of things. I mean, with agents, it's easier.
But here, because it's built on these open protocols, you can get stuff like this running really easily. And then you have this just crazy context engine that just can help you do work.
If I wanted to, I could ask Fizz right now, connect up the app or connect yourself to the X API and make it available so that we can start tweeting from Buzz.
Like, that wouldn't be that hard. We could probably do that in fifteen minutes, you know? Crazies.
One thing that I really like about this kind of we were talking about this closed circle of context that I'm getting back from my app.
So that tweet dashboard app is that here it sends me the tweets every day and the stats, like the top tweets and my stats and impressions. And I can just ask the agent, let's make sure I'm replying there.
And I'll just say, what can like, is the common thread between my top tweets?
And sometimes, like, I used to do this in a session with Claude Code or ChatGPT where I'd be like literally exporting tweet data or copying, pasting tweet data and putting it into ChatGPT on the desk or in the browser, getting information chat about it and pulling it back.
But this is so cool because it's getting the information from the XAPI. It's in the context window of the agents here.
I can chat about it here. And then we can chat with colleagues about what are some more tweets we want to push out?
What's a different marketing angle we could take? And that's what I really like about this here.
And you can see there is the agent working. So what's actually going on in the background, that's nice if used to looking at this kind of stuff in a terminal.
You actually see the tool calls that it's tools that it's calling and think or here it is down here, like what is happening. So it's getting information from the channels, checking for newer stats and things like that, gets the results.
And so that's all like kind of under the curtain stuff, but we should get a response back soon.
And is there is there any like tips around how you is best to talk to them? Like, I see here is you just wrote a short and concise question, but how how do you should how should people think about the best way to to talk to some of these agents?
I mean, they're pretty good at pulling out the the intent behind your questions these days. So just have a conversation with them like like you would like a normal person. So I don't think I at least don't have any special special approach there.
I kind of just say this is probably nicer than what I would normally type. Usually I got typos in there and stuff.
Like, give me the sauce. They're usually pretty good at that because they have all this context. So I'm usually pretty straightforward and concise with the questions and the stuff, the prompts I give.
Reported back. So it's showing what I did.
I tried Jack's Buzz. I deployed OpenSaaS to OpenShift, my first time successfully self hosting.
So it's like it's showing that I'm doing these things as a beginner, and I'm reporting back on what I learned.
And those are the things that are doing well. So it worked. My top two tweets got over a million impressions.
But this sucked theme got they're the bottom half, and they got a collective 88,000.
So novelty gets you reach. Friction gets you engagement.
So there's some actually really good advice there. And that's kind of the power of using it here and having all of it in Buzz is that makes stuff like this easy and illuminating.
So that's really cool. I like that.
By the way, I saw one of the one of the pieces of feedback is nobody is mentioning that Buzz has a backdoor for Bitcoin native payments.
What what does that what does that mean?
So Buzz, we're talking about open protocol. So I kinda glossed over this.
It's built on Noster. And Noster is the name of the open protocol that Buzz is built on.
And I personally believe, because you know if you know Jack Dorsey, he's a big fan of Bitcoin. Noster has tightly coupled with Bitcoin Lightning, which is a very fast and, like, almost fee less way to pay and transfer Bitcoin.
And my guess here is that because it's built on Nostr, Nostr is open, you'll be able to integrate Bitcoin Lightning payments into Buzz.
And you'll either be able to pay agents will be able to pay for compute, this shared compute idea that we talked about, or that's in there in the settings.
Agents will be able to pay for work and compute or people will be able to pay other people tip other people for tasks they get done. So I can honestly imagine a big open community of, let's say I don't know why I keep gravitating towards graphic designers, but like I say designers. And you need a really cool logo and AI image gen isn't really doing it for you.
And you ask some people spit out some ideas and some random person in some graphic design community on Buzz gives you an awesome logo and you just, boop, tip them.
Thank you for that idea. And then you go ahead and keep working.
And yeah, we'll see. It's not integrated at the moment, but I I think it will come.
Very cool. Awesome.
What else do people need to know about Buzz?
So let's see. You mentioned that shared compute idea.
And I think that goes back into the openness of Buzz, and that's really cool. And that was something that really surprised me too. So why don't we just talk about that for a minute?
So we've got this compute setting here, and you can turn it on.
And it will automatically suggest some local LLMs, some local model, AI models that you can download on your laptop or in computer, and you can literally share them with the other members of your team.
So let's say that you are just starting out.
You're a college student, and you want to start building a business.
You've this awesome idea. You don't have a lot of time. You don't have a lot of help.
And you don't have, you know, like the money to buy one of these Max Pro plans, whatever they're called, with like unlimited token budget.
So you could get together with some friends, buy a decent sized laptop or a Mac Studio or something or maybe even like a beefy Mac Mini. You put this on, you share compute, and all of you can use this one model running from one machine even though you're chatting from different from different laptops, from different computers, whatever.
So that's just like a very simple example. But you can see how they're paving the road here for real openness in model choice and being able to really harness the power of these open source and local models that are always getting stronger and starting to compete with the frontier models from Anthropic and OpenAI.
So I think that's really cool.
And why do you think that even this is worth playing with?
You know? Or, you know, why is this really important? Why is local important in the grand scheme of of the future of work?
I mean, model choice is important depending on what you want to do. And, you know, it's the same idea where Slack is the controller of your data, and that data is very valuable.
You're sitting here spending I mean, this stuff makes work a lot easier.
But you still have these really cool ideas. You're iterating on these ideas. You're putting stuff out there that you work pretty hard on and that you think a lot about.
And then to have just one company control all that data, basically. Like Slack, for example, have the whole all your data in that sessions. And if you're tired of Slack and want to move to something else, it's really hard or it might not even be possible to take all that data with you to some other platform.
So you just stay there and you're stuck with them. It's the same thing for models. You know, you see people talking about why, you know, there there are who knows?
Today it costs $200 a month for an unlimited plan. Tomorrow it might be $22,000.
We don't know. We don't know how how what's gonna happen. And the future, you know, having choice in the future is very important for the integrity, for the sustainability of any business because that's your data.
That's what you want to be able to do with it. I don't know. Some of these models have restrictions on what you're even able to ask them.
So governments have top down mandates on what they're allowed to do with the models or not. So we saw models get pulled back after people started using them for a couple of days.
So you're building your businesses on top of these tools.
You should be you should have flexibility and freedom and control of these tools, basically.
Yeah. And I think the big insight of all of this is a lot of us haven't realized how much of a content hub, Slack, and products like that have become to us.
And that and we're we're learning that if you wanna get the best out of any of these models, you need to have the most amount of context possible. So what's really cool about Buzz is it's basically like, okay, we recognize that this is your con your context foundation, and we're gonna go help you do a bunch like, we're gonna help you pivot into a bunch of directions.
You wanna, you know, have these projects that, you know, are integrated at GitHub, we can go do that. If you wanna do local stuff and do shared compute, we can help you do that, you know, etcetera, etcetera.
You wanna huddle an agent? We can help you do that. But it you know, what's so cool about it is whatever you decide to do, the context is at the core of it.
Yep.
Yep. Exactly. And that's what I like about it.
That's what I think is really cool is that I think they really hit on something here. So, you know, I think we'll see.
It could go like, I saw people commenting OpenClaw, you know. Nobody's talking about OpenClaw anymore.
OpenClaw was a real wow moment and like, okay, we can get these agents to do a lot of productive work for us. But I have the feeling that it was missing something and this shared context because we were working in teams. We're working with other people.
Teams might even just mean your other agents, right? Like you just You're going to have conversations whether it's with one agent or another, but now you have the ability to expand that, your team set, expand your, you know, your global context window, so to speak, and then make it globally available to everyone in your team, including the agents.
So, yeah, I think it's super cool. Around setting up and managing your agents, is there any
sauce here? Is there anything people need to know besides what you've already shared?
I I wouldn't say so. I've seen it depends on how how deep in the weeds you like to go. My honest opinion would be you probably don't need to do that much.
All I did the first thing I did was just I wanted one that's Fable. I'm using specifically or exclusively, I'm using Claude code as the harness under here.
But they have adapters so you can adapt. You know, you can even adapt OpenCLAW or Hermes or OpenCLAW or I'm sorry, OpenCode or Goose.
So like any of these harnesses can be added under the surface here, under the hood. But the first thing I did was I just pinned them to models.
So Fizz is a Fable model and Honey is Sonnet because there are just certain tasks that I don't need to use the power of Fable and burn through tokens so fast. So that would be my main piece of advice.
Besides that, just start playing around. I made this chief agent officer. So if I do start to create more agents with more specific kind of instructions and system prompts, then I might delegate the delegation to Mr.
Chief Agent Officer here and be like, Okay, I have this task. Who's the best for it? Because you might have a copywriter.
You might have a brainstormer, you might have a code reviewer, things like that. And you might forget who they are or, I don't know, you just get an agent to kind of delegate. So that would be pretty much my only advice on that side there.
Cool. Anything else you want to show? People were asking in Twitter about skills.
And I think I mentioned it already, but like all the stuff under the surface that you're used to using if you use Cloud Code or something, that's all available as long as it's to Buzz, to your Buzz agents, as long as it's globally installed. So skills and things like that, yeah, they can use and they can take advantage of.
So you don't lose that stuff. So that's also very nice. And I would say that this is beta software.
It's like it might even be alpha. I'm not sure. It's like early preview software.
So some things don't work that great. I was trying to create workflows and set up recurring tasks and things like that.
And weren't really landing great. That was one thing I noticed.
Another thing I noticed was things can be kind of slow. If you're used to it being like working directly in Cloud Code or in a codex.
It seems to be a bit faster there. And I think it's because it's communicating with your server, your relay, and communicating back.
And because of that, I think it's at the point now where it's really good for, like I said, solopreneurs, small teams iterating on small software or small ideas.
And if you're really doing complex software engineering, this might not be the tool that you want to use.
But if you're just looking for something that can bring you value and, like I said, bring all your knowledge into one central tool, then this is definitely a great a great thing to use.
So if you're a founder, small team, maybe, you know, one to 10, under $10,000,000 in revenue, 0 to $10,000,000 in revenue.
Like, should should you use this thing? Maybe you're using Slack today.
Should you download Buzz and actually try it out?
So I'm pretty convinced that like, I'm trying to convince my team at Wasp that we use this internally and exactly for that reason.
So we are a small company, a small startup, and we're developer tool, we're a software company.
And you have all this stuff that you talk about in Discord. We talk about in Discord. And we have our community there, too.
So that's one thing I should mention is that you can create channels that are private and public, right?
So you could invite people from your community using your product to come chat with you. And then you have private channels just for your team.
And then that context is shared, which is also a huge thing because if someone comes into your chat and says, hey, we want to get this fixed, then fixing it with an agent is just an at away.
You would just tell Fizz, let's tackle this bug. Whoever just did a bug report or said, you know, wanted a feature, you could just have them prototype it right there.
So yeah, I I think for small teams, this is definitely
worth checking out. So my take on my take on this is this is so cool.
It's clearly, you know, scratching the surface around what the future of work would look like in an a in a in a world where you have more agent employees than teammates.
Mhmm. It does feel early. It does feel like they're they're just, you know, like I said, scratching the surface.
But, you know, my passion is I I don't know if you know this about me, Vinny, but I'm a cofounder of a in a company called LCA, Late Checkout Agency. And that business designs, like, the world's biggest AI native software, not Buzz, but, like, you know, companies like that.
And when I see something like this, I'm I'm just like it's close. It's not there yet, but it's close.
And I do believe that it's probably worth and I have no affiliation with Buzz or Block or or Jack Dorsey, but I do think it's, like, worth trying these tools just so you can you might learn a thing or two, and then even if you don't end up using Buzz, like, you might just end up, you know, configuring Slack in a way that, you know, is best for you, and you learn something here.
So I think or you might realize, like, you love Buzz and forget about Slack, and this is what the world you wanna live in. Regardless, you you're gonna learn.
There is no losing in this situation. So I think the most important thing we can one of the most important things that we can do in the AI age and the agentic era is just getting our hands dirty, learning how these tools work, playing with them, and seeing how we could live in the future.
Because the people that live in the future, Vinny, those, as you know, those are the people that could, you know, look around the corner and who can create products that people really want.
And Yeah. You know, I think that's really important.
Yeah. And and yeah.
You bring up a really good point, and and I think that's why they're they're kind of looking around the corner with this already. And what I think could allow them to really be a strong player in this arena is the fact that it's open source.
So anybody that has an idea of how to so it's an open source app built on an open protocol. And so anyone who has an idea of how to improve this can.
And that might take it in the direction because the hive mind, so to speak, is going to be using this thing and working on it at the same time.
So they'll be like, Okay, this isn't working for me. I like this about, like you said, trying out from some other tool. I like this about the tool I usually use.
I'm going to add that in here or whatnot. So that might work to their advantage very well.
And it might become the tool that people just kind of default to. But things change so fast, so who the hell knows?
Amen, brother. Vinny, thank you so much for coming on, giving us a tour, sharing some I'll include links in the show notes, in the description where you can follow Vinny for more of this sort of stuff.
I would love to have you back on the podcast, Vinny. The comment section on YouTube, please let let me know if you enjoyed Vinny. I certainly did.
If you enjoyed this topic, what you want me to cover next.
Thank you so much, Vinny. Any last words for the people? Well, if if they want me to come back, I can show them some sweet stuff about how you can, like, use the open protocol that it's Buzz is built on.
So you can go pretty deep. And I think there's some pretty powerful stuff there that we both alluded to in our tweets. So, yeah, it's worth worth checking out.
Cool. Yeah. Let us know.
And we we live to serve. So if if that's what people want, we can go deeper.
Thank you, Vinny. Have a creative day, everyone, and I'll see you next time.
The Hook
The bait, then the rug-pull.
Jack Dorsey's new chat app has AI-Twitter calling it a Slack killer within days of launch. Greg Isenberg brings on an early user for a live, unscripted tour to find out whether that label actually holds up.
A two-hour SIP Live AMA and timeline breakdown covering getting your first 100 customers, a Cloudflare protocol for charging AI agents to browse the web, and the viral chart showing AI spend rising while headcount rises with it.
Cody Schneider maps the exact infrastructure — pipeline, warehouse, agent — behind a Facebook ads system that researches, creates, publishes, and kills its own losing ads.
Nick Vasilescu spins up an AI co-founder running on Grok 4.5 and, in one live session, takes it from idea to landing page, thumbnail, and cold-email sequence.
00:34
01:08
01:43
02:19
02:30
03:23
04:04
04:35
05:06
05:37
06:08
06:39
07:33
08:10
08:34
09:15
09:43
10:10
10:38
12:02
12:18
13:29
13:58
14:28
14:58
15:28
15:58
16:28
16:58
17:28
18:31
19:09
19:47
20:25
21:03
21:40
22:18
22:56
23:34
24:33
25:01
25:56
26:24
26:53
27:51
28:22
28:52
29:51
30:17
30:42
31:33
31:59
32:24
33:18
33:49
34:20
35:21
35:52
36:23
36:53
37:24
37:41
38:30
38:39
1:43:41
37:47
44:47
51:34
39:43
Video of the Day56:00