I quit OpenClaw for this
Publicado no IA Dots ·
Agentes de IAAI SignalAssistente pessoaldot-ai_agentsdot-ai_assistantsdot-artificial_intelligenceidioma-enInglêsMuseyt-ik8cyqLfLDE
Vídeo em Inglês, publicado pelo canal Tech Unicorn.
About this content: The summary, key points and topics below are generated automatically from the video transcript. They reflect the content, opinions, claims and predictions presented in the video and do not represent a position, validation or fact-check by IA Dots.
Video summary
The video discusses the speaker's transition to a new AI agent tool called Instinct, detailing its capabilities in handling daily tasks, travel, and purchases. The speaker compares it with OpenClaw, outlines different levels of trust and security considerations when using AI agents, and shares personal rules and experiences.
Key points
- The speaker transitioned from OpenClaw to Instinct, an AI agent that recently received a large valuation and funding.
- Instinct is described as easy to use and plug-and-play via WhatsApp, handling tasks like booking flights, ordering items, and managing calendars.
- The video categorizes AI usage into different levels of trust, ranging from basic daily assistant features to granting access to financial accounts.
- The speaker evaluates security aspects, noting potential risks such as an incident where an AI changed a password via email reset links.
- Personal rules were established for payment commands to ensure the AI only purchases items when explicitly told to do so.
- OpenClaw was abandoned due to maintenance issues, speed limitations, and high operational friction compared to Instinct.
Topics
AI agentsInstinct AIOpenClawAI trust levelsTravel automationAI securityProductivity tools
What you will learn
- The speaker uses a strategy of riding the AI wave by migrating skills and memory from one free tool to the next.
- Setting explicit language rules with AI agents helps prevent unintended purchases or actions.
- Evaluating the worst-case scenario and expected value helps determine the appropriate level of trust to place in AI tools.
- AI agents can streamline personal logistics, such as booking travel, ordering items while sick, and managing daily schedules.
Questions and answers
Why did the speaker switch from OpenClaw to Instinct?
The speaker switched because OpenClaw was not fast enough, hit usage limits, and required constant maintenance compared to Instinct's plug-and-play setup.
What is Instinct's current pricing model according to the video?
Instinct is currently free to use, with the founder's plan to monetize in the future by taking a cut of purchases made through the tool.
What specific rule did the speaker set for making payments with the AI?
The speaker explicitly stated that casual questions like 'can you do this' do not authorize a purchase, requiring specific phrases like 'buy this' or prior confirmation.
Video transcript
Transcript from available captions; it may contain errors.
This AI does everything that my automations
[music] team that I paid $50,000 to does like right out of the box. I just landed in
Bali and this AI [music] booked my flight, did my visa arrival card, handled the entire immigration
process for me. I just texted it like a friend. Tomorrow is my birthday and it's planning me
a birthday surprise. Like it got me a birthday present and I guess in the next video I'm going
to share with you what it planned for my birthday. I'm obsessed with this new AI and I have quit
OpenClaw and I travel constantly and it's just done all of my travel for me. I don't even know if
I need my human team anymore because the pace of AI agents is just so fast. And today we're going
to talk about the fastest growing one because it just got a 10 billion valuation and it raised $1
billion. A 23-year-old can build a $10 billion company. Why can't you at least use the product
to make your life better? Okay, just saying. And today I'm going to show you how to do just that.
I'm going to walk you through the different levels of trust that you're not all going to trust AI the
way I trust AI. And I get that. So we're going to talk about the levels of security and what you
are agreeing to at each level. I'm going to give you the full reasons on why I switched, what this
costs, and my rules for it that I set it up for so that I can trust it. And the scariest thing
that I gave it honestly wasn't my passwords, but we'll talk about that at the end. So first
of all, I was sick last week in Germany. Not only did it automatically book my train tickets when I
would get on the German train and I told it like, "Hey, I'm going here. Can you get my ticket?"
Done. I don't speak German. I was sick. It ordered me drugs to my German hotel and I just like had
them deliver downstairs. I don't know what kind of app it used. I don't know anything. It just made
a new account for me. So being sick is no longer a problem. It's insane. I'm doing this thing that
I'm pioneering. It's called riding the AI wave. But this is how I ride it. Okay. You export your
skills, memory, and everything from OpenClaw or whatever AI you're using, and then you import
it to the next free tool. This isn't all going to be free forever. So, you keep riding the wave
of whichever AI tool is free. And right now, this just got a billion dollar influx of money, and
that means the security is going to be good. Well, doesn't mean that, but I do think that it's going
to be better than a lot of other AI agents. And it means that this is going to be a really good tool
to use. So, I just imported all of my open claws, brain, skills, memory, everything into which thing
are we using? Instinct. I hate that I'm talking about it like this because it's not sponsored.
I did use Instinct to reach out to Instinct for partnerships. So, that's kind of funny. So, maybe
in the future it'll be sponsored, but right now, these are all just my honest raw thoughts. I'm
going to I'm going to talk [ __ ] about it a little bit as well. So, the way I'm thinking
about it right now is OpenClaw is like Linux. If you want a lot of customization, you want to
build your own [ __ ] from scratch and you really care a lot about the the nitty-gritty and the ins
and outs and you're okay with that, use OpenClaw, okay? And if not, this is like the Apple of
AI agents because it's so easy to use. It's plugandplay. It's no headache. It's don't even
lift a finger. You don't even have to lift the finger because you can whisper flow directly
to instinct. And people think I'm schizophrenic because I do it all the time. But that's okay. I'm
the most efficient person I know. So, let me take you through all the things I use it for. Starting
with the lowest to the highest level of trust because the lowest level of trust is like things
you might be comfortable with as a normal person, but the highest level of trust like damn. I
don't think anyone should be doing it, but I'm doing it. So, we're going to we're going to
see what happens. Guys, I do not think you should give all your passwords to AI or your credit
card or whatever. But the reason why I'm able to accomplish so much with AI agents is because I
have trust in it. Do I trust the company building it? No, absolutely not. Like I just trust that
these systems are built in such a way that if I prompt it correctly and I set boundaries and and
rules and guard rails like in me speaking to it that it's not going to do anything too crazy that
I don't want it to do. So obviously do with this as you will. I trust myself to talk to the agent
in a way that I know it won't do anything I don't want it to do. So the the first level of using
it without any trust is still quite useful but it's like you might as well just use chat at that
point. The second level is giving it access to your notion, your calendar, emails and obviously
I already do that in Claude and Chad GBT as well, but it is much more convenient talking through
WhatsApp about it. So I have a daily standup with my instinct where it tells me all of my
to-dos for my emails, my calendar for the day. um like you know it's basically my assistant. So
quick thing about email or Gmail specifically. I'm just fact-checking this because any public app
that reads your Gmail has to pass Google's review or you get a big scary red flag and I didn't get
that. And if your mail goes through its servers, it needs a security assessment every year. So that
part I can't look. I'm I'm just saying Google does review this stuff. It's not that sketchy. Okay.
So Instincts founder says there's protection and anything it reads goes through firewalls.
A separate system checks every action. He's also said security teams battle tested the
system that catches its hallucinations. He didn't name them and I couldn't find any
published results, but apparently it's secure. So that's what level two trust is. Now level
three, level three level trust is kind of scary and I was really not going to do this one. Okay, I
only did it because one of my friends that invited me to Instinct already gave Instinct their their
cards and had everything booked and I was like, "Okay, I'm kind of tempted to do that." So, my
assistant, my physical human assistant that is right now somewhere in the States traveling. Uh,
he has one of my cards that's like the single-use card uh because I trust him enough, you know, that
he's not going to go buy a Hermes bag with it or something. You actually couldn't do that. It has
a limit, I think, of $1,000. I want to tell you my thought process because I trust AI way more than
the regular person. But I also trust the universe and everyone in it a lot more than most people.
So that's always worked out for me. I have a lot of trust and that brings a lot of opportunities
to me. And the worst case scenario I'm always comfortable with. So this is how I think about
it. You think about the best case scenario and the worst case scenario. You have to be comfortable
with the worst case scenario. The worst case scenario is it books something crazy and I spend
$1,000. But the best case scenario is way better than $1,000 because every day, you know, I could
be making an extra thousand dollar just from the things that it's helping me do. So, the best
case scenario is already like the expected value of using the AI in this way is much higher than
not using it. So, I will use it. That's the logic. And you just have to make sure you're comfortable
with the worst case scenario. So, obviously, if you're not comfortable with losing $1,000,
give it a card that has a max limit of $100 or don't give it a card at all. Um, by the way, you
can also give it a Stripe link where you approve each purchase. So, you don't have to give it
your card. It can still buy things for you. Now, what it costs, this is the part of the terms
I should have read first. I read it last. I do things in the wrong order, so you don't have
to. It says, I appoint it as my agent to make agreements, commitments, or transactions for
me. And any such agreements are quote binding on you as if entered into directly by you.
Translation, if it clicked buy, I clicked buy. It's like sending a friend to the store with your
card basically, right? And remember who's paying cuz open clock cost money to run on API credits
and heavy use was like something $25 to $50 a day. I've definitely spent that before. This cost
me nothing. Like this is totally free. Yeah, they haven't disclosed any revenue. The plan
to monetize this, according to the founder, is to take a cut of what it buys for you. So, this
is actually the end goal is that this is going to be a a shopping agent. Yeah, I looked for an audit
or like a security page. Couldn't find one. So, this is just information that I did from like
deep research. Yeah, some scary incidents that happened. A VC, Ooh, let's read this one. A
VC posted that it changed her password without asking, which is kind of scary. So in her own
screenshot, the assistant tells that she said yes to getting access to her health portal. There was
no saved password, so it clicked forgot password and the recent link went to her Gmail and it can
read her Gmail. So it sent a new password and 8 minutes later it told her that it changed her
password. But the reset link thing is is fixable. So Meta says its assistant email connection
filters out login codes and reset links. So the AI isn't supposed to see them. Instinct
hasn't published anything like that. But yeah, so let's talk about my rules and the way
I think about this and things I'd want to know before giving it more information. I explicitly
told it my language for making payments. So for making payments, I specifically said if I say can
you do this, that doesn't count as you should buy the thing. Only if I say buy this or like confirm
with me first. So, you really want to explicitly talk to your AI agents in in general about
language and like when I explicitly say this, it means buy. When I ask about can you do this,
it doesn't mean buy. So, number one reason I quit OpenClot is it was just not fast enough to
build the things I wanted to build. Secondly, I would hit my limits. So, I couldn't even use
my Claude plan anymore. As you all know, that got banned. But then I used my codeex plan and I would
hit the limits on that which is annoying. I I am a I'm a heavy user super user over here as you
may know but that's no good and thirdly the maintenance there everything like keeps breaking
and you have to repair it and you have to log into your VM and repair it every time when things
break on this I just text my WhatsApp okay so this is why I quit openclaw now would I go back
to openclaw absolutely not with openclaw I was my maintenance guy I was the security team I mean
now that I don't have my automations guy I have to do everything and so no if this works for me
I'm going to keep using this when This is paid. I'm going to pay for it because it's just so worth
it to have a more carefree life. I personally love it. Some of you that don't use AI agents, you're
going to be like, "Oh, that's too much. You can't even make your own decisions." No, I can't make my
own decisions. Okay, Amanda, I like my AI making my decisions because I know like my higher self
knows the things I want to accomplish. Okay. So, I always was thinking, why is there such a lag,
such a barrier between what I want in my head and what I can materialize in life? Not only when
I'm like booking appointments, creating apps, or anything that I want to materialize in the
world, but there's just so much unnecessary in the middle like bureaucracy, interfaces,
communications, different apps, different everything. AI agents are the future because they
remove all of that. I can literally go almost nurling brain ideiation to creation instantly
and that's what I want. Like I am on the planet to create and if AI agents are allowing me to
create more and just like have my ideas become into reality faster, that's great. That's what I
want to use it for. Let me know next if you want to see more unhinged use cases cuz I have some
crazy unhinged I've done. Let me know below what other videos you want to see. I appreciate you for
watching. I'm going to get way more into YouTube now that I have AI helping me. I'm going to get to
100K in 6 months. Okay, that's my goal. I think I will reach that goal because AI is on my side. So,
uh, come back in six months and see if I did it or stay tuned for the next video. Yeah, I love
you guys. I will see you in the next one. Bye.
Assistir ao vídeo original no YouTube
Assuntos: Agentes de IA, Assistente pessoal, Assistente pessoal, Muse. Publicado no YouTube em 2026-10-15.
Seleção automática por assunto e idioma. O vídeo pertence ao canal de origem.
About OpenClaw and Instinct
OpenClaw is an open-source AI assistant written in TypeScript and released under the MIT license. Wikidata lists Peter Steinberger as its developer and records the earlier names Clawdbot and Moltbot. According to the official README on GitHub, OpenClaw runs on the user's own computer and talks to them in channels such as Discord, iMessage, Slack, Teams, Telegram and WhatsApp, with native apps for macOS, iOS, Android, Windows and Linux. State, memory and credentials stay on the user's hardware, models and agent harnesses (Claude, Codex or local models) are swappable plugins, and the project is stewarded by the OpenClaw Foundation, an independent 501(c)(3) with no paid tier, hosted service or token.
Instinct, the assistant the video switches to, is described on its official site as an AI personal assistant without an app interface of its own: the user texts or calls it, and it can use a phone and a computer the same way a person does. The site gives examples such as ordering ingredients for a recipe, booking a train trip, ordering groceries, sending flowers and finding a sports court.
Fact sheet
- OpenClaw – type: Open-source autonomous AI assistant
- OpenClaw – developer: Peter Steinberger
- OpenClaw – earlier names: Clawdbot, Moltbot
- OpenClaw – language and license: TypeScript, MIT
- OpenClaw – steward: OpenClaw Foundation (501(c)(3))
- Instinct – type: AI personal assistant used by text message and phone call
- Instinct – official site: instinct.com
Read also
- The $10B AI Assistant Challenging Meta's Muse, Grok's Bots, and OpenAI's Dots
- Instinct AI is For Real. What You Need to Know.
- The risks of trusting personal AI agents with your money, data
Sources and references: GitHub – openclaw/openclaw (README) · Wikidata – OpenClaw (Q137916205) · Instinct – official site · video by the Tech Unicorn channel on YouTube.