13:38Alex Finn · Tutorial
The new Hermes Agent update has me speechless
Eight features in Hermes Agent 0.17 that move it from chatbot to ambient operating system.
June 24thThe argument in one line.
OpenAI's new voice mode isn't dictation — it lets you talk to a single agent that can see every device and chat you have open and spin up dedicated agent threads to act on what you say, turning your voice into a chief of staff instead of a text-input method.
Read if. Skip if.
READ IF YOU ARE…
- You already run multiple AI agent projects (coding, business ops) and want a faster way to check status and delegate next steps without typing.
- You spend real hands-off time walking, commuting, or otherwise away from a keyboard and want AI-driven projects to keep moving during it.
- You've dismissed AI voice features before as 'just dictation' and want to understand what specifically changed.
- You manage several devices and want one of them to act as a command hub that the others control remotely.
SKIP IF…
- You don't already use agent-based AI tools like Claude Code, Cursor, or Codex — this workflow assumes existing agent projects to delegate to.
- You want a basic ChatGPT voice tutorial rather than a multi-device agent-orchestration setup.
The full version, fast.
ChatGPT's new voice mode differs from prior voice AI because the agent can see every open chat, connected repo, and device on the account, then spin up an isolated chat thread per task instead of just converting speech to text. The creator demonstrates asking for a cross-project status update, delegating fixes to new threads, and running the same agent from a phone in 'remote mode' while a separate desktop does the actual work. He structures his day around three voice rituals — morning status check, mid-walk brain dump, end-of-day debrief — each closed by explicitly telling the agent to spin up a thread, and recommends writing a short 'compass doc' per project so the agent has goals to reverse-engineer next steps from.
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 – 00:50
01 · Cold open
Alex Finn claims OpenAI's ChatGPT Voice mode is the biggest AI release from a major lab yet, previewing that it will change not just AI use but how people interact with computers.
00:50 – 02:23
02 · AGI moments
He frames the release against three prior personal 'AGI moment' tools — Cursor with Sonnet 3.5, Claude Code with Opus 4.5, and OpenClaw — positioning ChatGPT Voice as opening his eyes to 'Ambient AI.'
02:23 – 10:06
03 · Live demo
A screen-recorded walkthrough: asking voice mode for a status update across three projects, spinning up a dedicated agent thread per project, and using 'remote mode' from a phone during a six-hour hike to keep directing work on a separate desktop.
10:06 – 12:31
04 · Why it's so good
Five stated reasons voice beats typing for this use case: it blocks doom-scrolling, lets you juggle multiple projects, tolerates imperfect phrasing, and works from anywhere.
12:31 – 16:45
05 · Three workflows
Morning status check, mid-walk stream-of-consciousness brain dump, and end-of-day debrief — each closed by explicitly telling the agent to spin up a thread for the recommended action. Includes the 'compass doc' technique for giving agents durable project goals.
16:45 – 20:35
06 · Best tips + headquarters setup
Closing rules — ask more questions than commands, don't worry about misspeaking, and designate one always-on desktop as a 'headquarters' machine that every other device remotely controls — plus a walkthrough of the settings screen used to connect devices.
Lines worth screenshotting.
- The difference between AI dictation and AI delegation is whether the model just converts your speech to text or actually spins up agents to act on it.
- A voice agent that can see every open chat, connected repo, and tracked issue can compress a multi-project status check into one spoken answer instead of five separate app checks.
- Voice conversation structurally blocks doom-scrolling mid-task because a live back-and-forth demands an immediate reply, unlike a typed prompt you can walk away from.
- Spoken instructions tolerate filler words, false starts, and topic switches in a way typed prompts don't, because the model parses intent from the whole stream instead of one precise sentence.
- The fixed phrase 'spin up a thread for that' turns a casual voice conversation into real work delegation instead of just more chat.
- A one-paragraph 'compass doc' stating a project's short- and long-term goals lets an AI agent propose useful next steps instead of guessing from an open-ended question.
- Asking an AI 'what should we do next?' instead of giving it commands offloads the cognitive load of planning, at the cost of giving up some control over prioritization.
- Running the same voice agent from a phone in remote mode still lets it direct code changes on a separate desktop machine that's actually signed in.
- Designating one always-on desktop as a 'headquarters' machine, with every other device set to remotely control it, keeps a person's code and project state from fragmenting across devices.
- Three daily rituals — a morning status check, a mid-walk stream-of-consciousness dump, and an end-of-day debrief — are enough structure to keep multiple AI-driven projects moving without sitting at a desk.
Voice AI Works Best As Delegation
VOICE AS DELEGATION
Voice mode only pays off once you stop dictating text and start delegating tasks to isolated agent threads across every device, guided by written goals instead of moment-to-moment decisions.
02AGI moments
- The presenter frames ChatGPT Voice as his fourth personal 'AGI moment,' following Cursor with Sonnet 3.5, Claude Code with Opus 4.5, and OpenClaw — a personal timeline for judging which AI releases actually change how you work.
- Rating tools by whether they change your daily workflow, not by benchmark scores, is a simpler way to decide what's worth learning next.
03Live demo
- ChatGPT's new voice mode can see and control every device signed into the account and can open a new chat thread on each one, not just answer inside a single conversation.
- Asking for a 'status update' works because the voice agent reads across existing chats, connected repos, and issue trackers, then compresses that into one spoken next step per project.
- The agent doesn't do the requested work itself — it spins up a separate, isolated chat thread per task, and 'remote mode' lets the same agent run from a phone while still directing the machine that's actually signed in.
04Why it's so good
- Voice conversation structurally blocks doom-scrolling mid-task, because a live back-and-forth demands an immediate reply the way a sent text prompt doesn't.
- Spoken requests tolerate filler words and topic-switching in a way typed prompts don't, because the model parses intent from a messy stream rather than one precise instruction.
- A spoken request can be followed immediately by a second one about a different project, which is what makes working several projects in parallel practical.
05Three workflows
- Three recurring voice rituals structure a day: a morning status-and-next-step check, a mid-walk 'stream of consciousness' brain dump, and an end-of-day debrief asking what to push forward overnight.
- The instruction that turns any of these rituals into real action is a fixed phrase — 'spin up a thread for that' — which the agent uses as a trigger to delegate instead of just chatting back.
- A short 'compass doc' — a markdown file per project stating its short- and long-term goals — gives the agent a target to reverse-engineer next steps from, instead of guessing from an open-ended question.
06Best tips + headquarters setup
- Favor questions over commands when talking to a voice agent; it has more context and can think through options faster than manually specifying every step.
- Treat one always-on machine, not a laptop that sleeps, as the 'headquarters' that every other device remotely controls, so all code and project state stay in one place.
- Setting this up is a one-time step in the app's connections settings: enable 'control this Mac' on the headquarters machine, then approve each phone or tablet as a device allowed to control it.
Terms worth knowing.
- Ambient AI
- AI that's always listening in the background and can act on an idea the moment it's spoken, without opening an app or typing a prompt.
- Chief of staff pattern
- Using a voice AI to triage and delegate work to separate specialized agents rather than having it perform every task itself.
- Compass doc
- A short markdown file written for each project stating its short- and long-term goals, so an AI agent can reverse-engineer useful next steps instead of guessing.
- Headquarters device
- The single always-on desktop machine a person designates to hold all their code and project files, which every other device remotely controls.
- Remote mode
- A setting that lets a phone or tablet run the same voice agent and direct work on a separate headquarters computer while away from it.
- Reverse prompting
- Asking an AI open-ended questions, like 'what should we do next?', instead of issuing direct commands, and letting the model propose the plan.
Things they pointed at.
Lines you could clip.
00:00
“OpenAI has quietly released what I believe to be the biggest AI tool a major lab has ever put out.”
02:53
“This isn't dictation. This is being able to command an entire army of AI agents with just your voice.”
07:51
“I actually got more work done walking around outside in the redwoods of Northern California than I do sitting at my desk with a keyboard and mouse.”
16:12
“Agents are a lot more productive when they have a goal to work for, not when it's just some sort of arbitrary question, what should we do next.”
20:19
“I promise if you do the things I listed out in this video, the way you use AI will never ever ever be the same.”
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.
metaphoranalogystory
OpenAI has quietly released what I believe to be the biggest AI tool a major lab has ever put out. I think it's not only the future of AI, but the future of how we interact with computers as a whole.
Now this isn't hype. They're not paying me. None of these major labs pay me or even really talk to me at all.
This is based on my experience of using ChatGBT voice over the last few days. It has completely changed how I use AI.
In this video, I will show you how to do the exact same thing. I will show you why this has been an AGI moment for me, how it's completely changed the way I use my computer, and also give you a few workflows you can use immediately to 100 x your productivity.
I'm not even kidding. If you stick with me to the end of the video, your entire life is gonna change. Now let's lock in and get into it.
Let's first talk about how big this moment is with Chad GBT voice. Let's first talk about these AGI moments. Moments where I used a technology that completely changed how I view the world.
The first was Cursor with Sonnet three five. This is when I really discovered vibe coding. When I saw how much AI was going to disrupt programming, this got me into AI.
This inspired me to start my YouTube channel, all my content, everything like that. Next was Claude Code with Opus four five. This was the moment where I was convinced software is basically solved.
AI can handle everything software related. Then there was OpenClaw, which really opened my eyes to AI agent and made me believe we would all have our own army of thousands of agents working for us in the near future.
And now is ChatGBT Voice. This has opened my eyes to Ambient AI.
What is Ambient AI? Ambient AI is the idea that there's gonna be an AI waiting for you always, always listening, and can build things for you no matter where you are in the entire world.
It completely removes the barriers and the friction between you coming up with an idea and it being created and appearing in front of you. Over the last several days of me using ChadGBT voice, I have used it everywhere.
Walking around outside, going on a hike at a cafe, around my apartment, in my bed, no matter where I am, I can just say ideas that come to me, and they get built out and created on my computer. Let me show you how this works.
So up until now, the voice mode for every AI tool ever made has been dictation. What is dictation?
You click a button, you say things out loud, and then they appear in your text chat, and you hit send on it. See, there's the thing I just said.
You say things out loud, and they appear, then you hit send. Right? It's the same thing as using the AI you've always used.
You're just saying it, and then the text appears on your screen. That's it. That's all dictation is.
This isn't dictation. This is something completely different.
This is being able to . Let me show you. So if you hit this new button in ChatGBT right here, start new voice chat, it opens up chat GBT voice.
So I'll mute this real quick. We're now in chat GBT voice. You're probably pretty familiar with this interface before because they had a sort of voice like interface before where you can chat with an agent, but now this is super powered.
What makes this super powered? Now your ChatGBT voice agent can control any of your devices.
Any device you have that has the ChatGBT app on it, your iPad, your iPhone, your computer, your voice agent can now go on any of them and control them. On top of that, that agent can go and spin up new agents and new threads on each of those.
So, basically, the power of your voice can control all your technology, get things built in an instant, and explode your productivity anywhere you are.
And I'm gonna demo this for you in just one second and show you why this is so different than anything else you've ever used, then go into the workflow. But when you were using voice before with AI, it was basically dictation where you'd say things, it'd get converted to text. The AI would read it, then it would come up with text and convert it to voice for you.
Right? It's just a simple AI chat. Now it's an entire command center.
So watch how I typically start my days. Hey, Chad GBT. Can you give me a status update on school OS, on personal OS, and on Henry?
Then give me a recommended next step with each. I'll check each project's current status and then boil it down to one practical next step for each. So before when using voice AIs, it would have access to just the chat.
It can't see other chats you're doing, other AI agents you're using, other things going on on your entire computer and all your devices. Now because voice has access to all your devices, all the chats, everything all your different apps are doing, it go found.
School OS is deployed, but the course sync is paused after an upstream error, so no member course data is flowing. Next is restoring a valid course response and running a one member canary before turning sync back on.
Personal OS is production stable, including the video ideas release, but Google Calendar is waiting on your approval at Google's security warning.
That's the next move. But the founder attended three cycle production proof is still undone, so no external beta yet. The next step is to run that proof in the controlled window and record the verdict.
So I just got a status update on the three biggest major projects I'm working on right now. It checked out all the chats. It looked at the GitHub to see all the code updates.
It looked at my linear to see all the issues we have queued at the moment. It can see across everything. But here's where it can get really interesting.
Not only does it have a view into everything you're doing, all your chats, all your devices, but it can spin up agents to work across all of those. So watch this right here.
Okay. Can you spin up a new thread in school OS to fix that error and get the connector working? Inside the personal OS, can you spin up a new thread that makes the Google that makes the Google Calendar connector work and get that approved for me?
And then for Henry, can you actually build a ChadGBT site that has a plan for how we can do the entire production test and run? And then send that all to me when it's all ready and done.
Got it. I'm lining that up. All three can be set in motion.
I'll prepare and verify the Google connection,
and I'll prompt you in a visible browser for the one approval only you can complete. I'll keep each effort isolated so nothing spills across repos. And now, boom, look at this.
Three new chats spun up to do the work I just approved. So you can see the Henry test plan is building a chat GBT site for me right here. There's a new chat.
Personal OS, it's fixing a calendar issue, and that's running for me. And school OS started as well. So it spun up these three new agents to begin work for me.
So . I tell it what's on my mind, and it goes and it just delegates the work to all the agents that need to get it done. Now you take it a step further and you think about what this is like when you start using it cross device.
So for instance, I was out yesterday. I was going for a hike for like six hours. I was on my iPhone.
And what you can actually do when you're on your iPhone or your iPad or any other devices, you can go into remote mode. And when you're in remote mode, you can actually start up your voice chat there as well.
And again, this voice chat has control over all your devices. So I'm on the go. I have my AirPods in, and I'm able to talk through all my work without looking at a screen, without touching a keyboard, without touching a mouse, and get agents spun up across all my devices to keep all my projects moving forward.
I not only got the same amount of work done that I typically do sitting at my computer, I actually got more work done. I'll tell you why it's so much more efficient in a second, but I got more work done walking around outside in the redwoods of Northern California than I do sitting at my desk with a keyboard and mouse.
And what's amazing is it's ambient. And so you can give commands, go on mute like I just did here, and then walk around outside, sit at your computer, read a book if you want.
And then when the work is done, Chad GBT voice will just wake up again and say, hey. By the way, we just finished this. We're working on this now.
This was complete. You can check this out. Personal OS calendar is built and verified locally, including the daylight saving fix.
Google's approval page is open waiting on your consent, so nothing's deployed yet. Do you want me to push it live once you approve Google's read only calendar access? Yes, please.
You can do that once I approve. Got it. I'll handle the deployment right after your approval.
And as you can see here, it just goes. It's just ambient. I can be doing other things.
So once you approve global's read only access, it'll go live and get checked. And this is what makes it so great. It's ambient.
So while it's working, I can be doing other things. I can be reading. I can be working on other things.
I could be I could be going for a walk, and then it just wakes up and says, hey. We just completed this work. Here's the next thing you need to do.
And you can do this on the go. I was sitting by the pool yesterday as well. I had one headphone in.
I'm just sitting there relaxing. Every once in a while, my AI wakes up, says something to me. I go, okay.
Yes. Continue. Okay.
Work on this. And then it just goes. I'm staying productive even when I'm walking around, sitting by the pool.
I mean, you can look at me. I'm sunburnt all over because I was spending so much time out yesterday being productive. I'm gonna go into two workflows you can use immediately in a second, and I promise will change how you work with AI forever.
Also, feel free to jump around down below if you want the different chapters. If you learned anything at all so far, make sure to leave a like down below. Subscribe and turn on notifications because all I do is make amazing videos about AI.
Also, join the Vibe Coding Academy. I'm doing a live boot camp on voice chat this week. I will live show you way more workflows, so make sure to join that down below.
Promise you'll be the best decision you ever make. But before we go into these workflows, let's talk about why this has been such a massive productivity gain and all the benefits you get from using this workflow. Number one, you can't doom scroll while conversating.
If you're anything like me, whenever you send a prompt to an AI, it's so easy to kinda just pop open Twitter or look at a YouTube video and do things while the AI works. And then all of a sudden, it's half an hour later and you haven't responded to the prompt since you've sent it. It is impossible to doom scroll when you're deep in a conversation.
You're talking to an AI. I can't say something, and then it says something back, and I sit there and scroll for an hour. You're forced to be engaged.
So you just get way more done, and you're way more focused. You can chat about multiple projects at once. So this, again, acts as your chief of staff.
This is how you want to think about ChatGPT voice, which is it's your chief of staff. It's not doing the work for you. It's delegating your work to other agents.
So much like a CEO would have a chief of staff that interacts with officer, the chief product officer, the CTO. It's the same way where you're talking to it, and it's giving work to all your other agents.
So you're able to move the ball forward with multiple projects at the same time. When you're typing and you're texting your agent, you can only text one agent at a time and give one command at a time for each project. So you're able to multitask way better.
You don't need to be perfect. Right? When you're typing a text prompt out, you kinda need to be perfect because it's gonna read it very literally.
But when you're talking, you can mess up. You can say, um, ahs.
You can talk about multiple things, and the AI will truly sit there and think about what it is you're saying, parse it, and figure out the right things to do. And this is maybe the biggest game changer for me, and it can be done anywhere. All of this can be done anywhere.
My work doesn't need to be here at this desk. In fact, when I'm done filming this video, I'm gonna go outside and go for, like, a four mile walk and just have one AirPod in talking to my agent the entire time. When ideas come to me, I'll say them out loud.
The agent will take it, assign the task, and keep it moving. Found the calls. Opening the second connection over the first task, secure callback.
Still giving me updates on everything we talked about. I'm getting work done while building a video. While signed in as Alex at one percent better dot I o.
I'm literally this is the first time I've ever gotten work done while filming a video. This is pretty amazing. So I'm gonna show you three amazing workflows with this right now, and then I'm gonna, at the end, right after this, give you a bunch of huge tips you need to know so you can get the absolute most out of this feature.
Let's talk about morning kickoff. I actually showed this workflow to you at the beginning of the video, which is this. You ask for a status update on all your active projects plus one step you can take to move the project forward.
Then once the agent gives you the recommendation, go, okay. Spin up a new thread for every action you recommended.
That's the key here. That's one of the biggest prompts you wanna give when you're talking is, okay. Sounds good.
Spin up a thread to do that work. You never want your voice agent to actually do any of the work itself. You want it to delegate.
So everything you do, say, yep. Okay. Spin that off to a new thread, and it will start a new chat in whatever the appropriate project is and make sure that work gets done.
Next workflow is stream of conscious. This is just when you go for a walk, put in your AirPod, and just start talking about whatever's on your mind.
Whatever you think about the current projects you're working on, your career, your goals, your ambitions, what you're thinking about, you just brain dump it all out to your voice agent. It's just listening the entire time as you're walking around. And the very and you go, okay.
Based on everything I just told you, what should we build out? What can we build to get me closer to my ambitions and goals? Then it will give you a bunch of ideas.
Know, okay. Spin up a bunch of threads to get all that done. And then it goes you come back to your computer at the end of your walk, and all your work is done for you, and you moved everything closer to your goals.
And then the end of the day debrief, I like to do this right before I go to bed. So right before I put my phone on the charger, I go, I'm about to go to sleep.
What are three things we can do tonight to push our project forward? Gives you a bunch of ideas. You say, hey.
Spin up some threads to do that, and it's good to go. Here's the key thing. Here's the one connector between all of these use cases and workflows.
You should be, when using ChatGBT voice, asking way more questions than commands you give.
This is the beauty of having super intelligence at your fingertips. I guess not really at your fingertips, at your mouth. That's weird.
It has way more context, can think way faster, knows way more things than you. No offense. So you should be reverse prompting as much as you can.
What would you do here? How would you push things forward? The advantage here is, one, it comes up with ideas you never would have thought of.
But two, it lightens your cognitive load. So if you're walking around outside, you're enjoying your day, you're talking to Chad GBT voice, don't wanna, as you're walking, have all these really deep thoughts about what is every step your agent should be doing next. Instead, you go, hey.
What do you think we should do next? It gives you a bunch of ideas, then you just approve. That's so much easier on your cognitive load than having to manually think about every single thing you wanted to do.
So you should be asking way more questions. As long as you're using it and saying, hey. What should we do next?
You're going to get good results from it. How do you improve that, though? How do you improve the recommendations it gives you?
Well, that's through what I call a compass doc. So one of the first things you wanna do in ChadGBT voice when you boot it up is say, hey. Can you build a compass doc for each one of our projects that explains to our agents what the short and long term goals of this project are?
What ChadGBT voice will do is create a markdown file in each one of your major projects that explains what you wanna do with that project in the long and short term.
Now when you ask, hey, what should we do next on these projects? Your agent can actually look at your compass doc and figure out, okay. What are actions we can take to reach that goal?
Agents are a lot more productive when they have a goal to work for, not when it's just some sort of arbitrary question, what should we do next? If it has a goal to work for, it can reverse engineer that and figure out all the steps it should take to get there. So if you have it, first thing you do, create a compass doc for every project.
You can now go anytime you want to your voice agent and say, hey. What should we do next? It'll make sure it creates a step that gets you closer to those short and long term goals.
So here are my best tips I'm gonna give you on this. As I said, number one, ask more questions than prompts. You shouldn't be giving as nearly as many commands as you give questions.
What should we work on next? How would you handle this? What's the best way to accomplish this goal?
You'll get way more better results, and it'll be easier on your brain to just ask questions. Don't be afraid to rant or mess up. I find that Chad GBT voice is incredible at finding the unknown unknowns.
Right? The things you don't know that you don't know. So if you go and you just rant and you say everything on your mind, Chad GBT voice can find the angles that you haven't thought of yet and find the solutions to those angles.
You also don't have to worry about messing up. If you misspeak, it can figure it out. You can say, um, uh, uh.
It'll be able to parse through that and figure it out. They really train this voice model really well to understand the true intentions of everything you're saying. And then most importantly, choose one device as your headquarters.
I have a bunch of devices. I have three Mac studios, an iPhone, an iPad, a DGX spark, a home PC. I have all these devices.
What you wanna do is this. You wanna choose one of those devices preferably an always on desktop computer like a Mac Studio, a Mac Mini, or a PC, and have that be your headquarters, then have every other device be a node for those headquarters so they control that main computer.
So I'm on my Mac Studio right now. Because this is my headquarters, when I go on my iPhone and I give it commands, it'll control and write the code on my Mac Studio.
Then if I go on my iPad and give voice commands, it'll control everything on Mac Studio. So it keeps all your code in all your projects in one place. This will allow you to go mobile and do things a lot easier on the go.
To set all that up, you just go into your settings, then you go to connections right here. And then on your main headquarters computer, you click control this Mac, and then you add all the devices that you want to allow to control this computer. Once you do that, you can go on your iPhone, your iPad, wherever you are, and you can have that device from ChadGBT voice control your main computer, which is amazing.
And, again, it controls your entire computer. So you can say, hey. Build a PowerPoint.
Hey. Edit this doc. Open this up on my computer.
Load this local model. Whatever it is, you go into voice. It controls it on your computer.
This has completely changed the way I work. The moment I'm done editing this video and post it online, I'm gonna be out in the sun in the 80 degree weather of Northern California walking around with my AirPod and getting work done.
I think this completely changes the road map of every AI lab. I mean, this is so powerful. I think Anthropix tries to copy this as quickly as possible.
I bet SpaceX comes out with a new kind of voice product like this really, really soon. I actually personally guarantee you that the Johnny Ive device that's gonna come out later this year is gonna be some sort of microphone product that ties directly to this functionality. Maybe a little pin or something that sits around in your house that can hear you talk anytime that you can just say prompts to anywhere you are, and it takes it and does the work on your computer.
I would like bet anything. That's what the product is that's coming out. It's just basically ChadGBT voice, but in hardware form somewhere around your house.
I have, over the last few days, turned this product on to, like, 10 of my friends, and all 10 of them are like, wow. This is the most amazing thing ever. I thought this was just voice dictation.
This is so much bigger than that. I promise if you do the things I listed out in this video, you take it, do it right now after this video ends.
I promise the way you use AI will never ever ever be the same. This was helpful. You learn anything all?
Leave a like down below. Let me know down the replies. Have you used ChadGBT voice yet?
What do you think? Subscribe. Turn on notice.
I'll see you in the next video.
The Hook
The bait, then the rug-pull.
OpenAI's new voice mode isn't a better way to talk to a chatbot — it's a voice-controlled command center for an entire fleet of AI agents. The creator walks through how a single spoken sentence, delivered from a hiking trail or a pool chair, can check the status of several separate software projects and then spin up a dedicated agent thread to fix each one.
Named ideas worth stealing.
02:11list
The 4 AGI Moments
- Cursor w/ Sonnet 3.5
- Claude Code w/ Opus 4.5
- OpenClaw
- ChatGPT Voice
The presenter's personal timeline of tools that changed how he works, used to argue this release deserves the same weight.
Steal forA simple frame for evaluating whether a new AI release is actually worth learning, based on whether it changes daily workflow rather than benchmark scores.
09:38list
Productivity Gain (5 reasons voice beats typing)
- Can't doom scroll while conversating
- Can chat about multiple projects at once
- Can take multiple actions at once
- Can stream of conscious
- Can be done anywhere
Five stated reasons conversational voice mode produces more output than typing into a chat window.
Steal forA checklist for deciding when a voice-first workflow suits a given task better than a typed one.
12:31list
Three Daily Workflows
- Morning kickoff
- Stream of consciousness
- End of day debrief
Three fixed voice rituals tied to specific times of day, each ending in delegating action items to new agent threads.
Steal forA ready-made daily structure for keeping multiple AI-agent projects moving without sitting at a keyboard.
16:02concept
Compass Doc
A short markdown file per project describing its short- and long-term goals, so a voice agent can reverse-engineer useful next steps.
Steal forA lightweight way to give any AI agent, voice or text, durable context about what a project is actually trying to achieve.
16:45list
Best Tips
- Ask more questions than prompts
- Don't be afraid to rant or mess up
- Choose one device as your headquarters
Three closing rules for using voice mode effectively.
Steal forA short checklist to sanity-check before adopting any voice-agent workflow.
How they asked for the click.
VERBAL ASK
09:52product
“Also, join the Vibe Coding Academy. I'm doing a live boot camp on voice chat this week.”
soft plug woven in mid-video, right after the productivity-gain framework and before the workflows section, pointing to his paid Skool community rather than a single product purchase
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
