Latest / In The Loop / Why OpenClaw is the most important software invention since ChatGPT

Transcript

Why OpenClaw is the most important software invention since ChatGPT · In The Loop

• 0:00
  This episode is about Open Claw, the open source AI agent tool
• 0:04
  that went from a side project by just a single Austrian developer
• 0:09
  to what the CEO of NVIDIA, Jensen Huang, a couple of weeks
• 0:12
  ago described as probably the single most important release of
• 0:16
  software probably ever. He compared it to Windows,
• 0:19
  Linux, HTML, and announced that NVIDIA were building their
• 0:22
  entire enterprise agent strategy on top of it.
• 0:25
  In today's episode, I'm going to explain how this project became
• 0:30
  such a global sensation. I'm going to explain exactly why
• 0:34
  it's one of the most important software updates ever, why it's
• 0:38
  potentially more important than the web browser, than Linux,
• 0:41
  than even the iPhone operating system.
• 0:43
  And I'm going to also explain what it actually is, what's
• 0:47
  innovative about it, what's new, and what it means for all of us.
• 0:51
  This is In the Loop with Jack Harton.
• 0:54
  Hope you enjoy the show. Let's start with what open core
• 1:07
  actually is, because the first thing to understand about open
• 1:09
  Core is it's it's not just a normal agent.
• 1:11
  You know, yes, it is a chat bot still because, you know, all
• 1:15
  agents are mostly chat bots and it feels mostly the same, but
• 1:17
  underneath it's something very, very different.
• 1:20
  Essentially what open core is, is a long running no dot JS
• 1:22
  background service, often called a demon.
• 1:25
  So in software terms, that is essentially a computer program
• 1:30
  that runs as a background process.
• 1:32
  So it doesn't rely on a user to initiate the process.
• 1:37
  So it might handle typically on a, you know, a device, maybe
• 1:40
  something like automatically connecting to your Wi-Fi.
• 1:42
  Now open claw connects to your file system, your browser, any
• 1:47
  APIs, your calendar, your e-mail.
• 1:49
  It's on your laptop. It can run on a schedule.
• 1:52
  It monitors all those applications for things that are
• 1:55
  happening called events. It can change a series of
• 1:58
  actions together and skills together, and it can message
• 2:01
  you, you know, on text or Messenger or WhatsApp or
• 2:05
  whatever it might be when it needs your input or attention.
• 2:08
  Now, if you compare that to ChatGPT, which is a conversation
• 2:12
  that you must start, whereas as long as your PC is open or you
• 2:15
  got something to host it and keep it running, it's literally
• 2:20
  always on. Then there's the interface, or
• 2:23
  in this particular case, the lack of a web interface.
• 2:25
  So there's no open claw, you know, web app that you make an
• 2:28
  account with and log into. Instead, it works inside over 15
• 2:32
  different messaging platforms that, you know, WhatsApp,
• 2:35
  Telegram, Slack, Discord, Signal, iMessage, Microsoft
• 2:39
  Teams, Google Chart. And you can just very simply
• 2:41
  text it the same way that you'd text someone that is in your
• 2:44
  team, you know, reschedule my Thursday meeting or, you know,
• 2:48
  go to my emails, summarize them all, triage and prioritize them
• 2:52
  all, write e-mail responses, and then I'll approve them.
• 2:55
  Summarize the PDFs that have been sent to me across all these
• 2:58
  different, you know, Microsoft Teams are Slack chats.
• 3:01
  And I have to say, these types of things, when you try them,
• 3:03
  they are quite magical. And that zero friction approach
• 3:06
  to an agent doing something for you is why the adoption curve
• 3:10
  has been so fast. I mean, staggeringly fast.
• 3:14
  It's faster than ChatGPT, roughly 20 times faster by a
• 3:19
  many visitor growth numbers. You know, every AI company spent
• 3:22
  billions trying to get people to come to their own interface,
• 3:25
  whereas Open Cloud simply went to them.
• 3:28
  And you know, Anthropic and all the big platforms have of course
• 3:30
  copied almost all of it. They copied the messaging first
• 3:34
  approach with Cloud Code or Cloud Code channel, shall I call
• 3:37
  it. They've got Bishop, Telegram and
• 3:39
  Discord integrations within a few more weeks that brought out
• 3:43
  dispatch, that brought out schedules.
• 3:45
  So many of the different features that Open Claw built,
• 3:48
  they've provided to users as part of the $20 or whatever
• 3:53
  membership T you're on, but in a much more secure and Safeway.
• 3:57
  And I think the speed of their response tells you a lot.
• 3:59
  An open eye even went and bought Open Claw.
• 4:02
  Now buying Open Claw is more buying the talent and the
• 4:05
  people, which was basically just the main Austrian developer who
• 4:08
  built it all because it was all open source and everybody could
• 4:12
  access the technology and see how it was built anyway.
• 4:29
  Now the skills and MCP ecosystem is where it gets really
• 4:33
  interesting because a skill in Open Core is just a markdown
• 4:37
  file, so a document type, so just a document type containing
• 4:41
  human readable, so normal text instructions for a specific
• 4:44
  task. Now if you've listened to last
• 4:46
  week's episode, you'll understand exactly what a skill
• 4:49
  is. I really recommend going listen
• 4:51
  to that episode if you haven't. And it's one of the most
• 4:53
  underutilized new capabilities that all AI systems globally are
• 4:58
  now adopting. And any of those skills are able
• 5:00
  to call into and connect to MCP servers.
• 5:03
  So model context protocol, which is basically a wrap around of
• 5:08
  integrations APIs. So it makes it really easy for
• 5:11
  an agent to take actions in different systems.
• 5:13
  So suddenly this agent has access to a huge ecosystem of
• 5:17
  skills and a huge ecosystem of integrations and control of your
• 5:21
  entire PC, and it can run around the clock for you.
• 5:25
  You can see why it's an incredibly innovative approach
• 5:29
  to how agents can do work for people.
• 5:31
  And that community really blew up.
• 5:32
  So Claw Hub hosts more than 13,000 community contributed
• 5:37
  skills and basically each skill is like a mini app for an agent.
• 5:41
  And then there's also memory open.
• 5:42
  Claw stores everything in just markdown on your local machine,
• 5:46
  such human readable. You can look at it, you can edit
• 5:49
  it, and it essentially builds a three layer memory system.
• 5:53
  And then what's really interesting about open claw is
• 5:55
  the token economics. So obviously tokens are when
• 5:59
  word start getting generated by an LLMA tickle.
• 6:02
  A typical ChatGPT prompt uses maybe a few 1000 tokens and a
• 6:07
  genetic open claw task where the agent plans, executes, checks
• 6:11
  your PC does work, re pawns, re executes, all whilst you know
• 6:16
  you're having a shower. It can consume 1000 times more
• 6:21
  tokens and chat chi BT. So you can just imagine a
• 6:23
  continuously running background agent, the kind that Jensen
• 6:26
  Huang is envisioning everyone having, is going to burn
• 6:30
  millions times more tokens per day than a single chat chi BT
• 6:34
  conversation. Now obviously the CEO of the
• 6:37
  company that sells everybody computer chips wants this to be
• 6:41
  how everyone consumes credits, because obviously that means
• 6:44
  everyone needs more computer chips from them.
• 7:02
  So yeah, I think this dovetails really nicely into our Jensen
• 7:04
  firing is so excited because he called Open Claw the operating
• 7:07
  system of agentic computers. It's huge big conference just a
• 7:11
  couple of weeks ago. So he said, and I quote Mac and
• 7:14
  Windows are the operating systems for the personal
• 7:16
  computer and Open Core is the operating system for personal
• 7:20
  AI. You know, Windows and Apple made
• 7:22
  personal computers accessible just to everybody, and Open
• 7:25
  Claw, in his framing, does the exact same for personal agents.
• 7:30
  And the deeper analogy is Linux. You know, Linux has become
• 7:33
  basically the invisible infrastructure of the Internet.
• 7:36
  You know, every cloud server, every Android phone, every smart
• 7:39
  devices runs on a Linux, and they built all operating systems
• 7:43
  on top of Linux. And Jensen is betting that Open
• 7:45
  Claw becomes that invisible infrastructure of the agent era.
• 7:49
  Linux took 30 years to reach ubiquitous global adoption.
• 7:53
  Open Claw, on the other hand, surpassed that in three weeks.
• 7:56
  Yes, you heard that right. Linux is one of the most
• 8:00
  important open source technologies ever invented, is
• 8:03
  literally the basis of most technology, and Open Claw has
• 8:07
  become more widely adopted than Linux in just a few weeks.
• 8:12
  So yeah, NVIDIA have obviously announced massive plans to make
• 8:15
  Open Claw the fundamental pillar of its strategy so it can then
• 8:20
  increase token generation globally from say 2 million
• 8:24
  tokens per second, which is where we're at today, to about
• 8:27
  700 million tokens a second, which is about 350% increase.
• 8:32
  The scale of what Jensen Huang is describing is obviously just
• 8:35
  ridiculous. You know, NVIDIA currently has
• 8:38
  43,000 employees and he projects 75,000 human employees alongside
• 8:44
  7 and a half million open call like AI agents over the next 10
• 8:48
  years. For NVIDIA alone, that's about
• 8:50
  100 digital workers per person and the things that you can get
• 8:54
  open call to do, and very soon all the main platforms are going
• 8:58
  to enable you to do the same on their systems.
• 9:01
  The scale of production is quite outrageous and the number of
• 9:05
  tokens that are going to be consumed is ridiculous, which
• 9:09
  means that obviously Nvidia's chips are going to be very, very
• 9:11
  important. So for those that you know, want
• 9:13
  to jump into testing this, what I'd say is that you can a wait
• 9:18
  for this to be fully released on all main systems that you feel
• 9:22
  really comfortable with and trusted.
• 9:24
  When that starts to happen, you know, core code and core Co
• 9:27
  works, let's say, has already started doing that.
• 9:29
  I'm going to do an episode as a full breakdown, but if you
• 9:33
  really want to get into it and I would recommend people having a
• 9:35
  go, I'll give you a few tips here and also how to do it
• 9:38
  safely. So if you want to set it up,
• 9:54
  ideally you have clock code maybe running in an IDE like
• 9:58
  Visual Studio Codes. And then it makes it really,
• 10:00
  really simple because it will just help you get through it
• 10:03
  all. So if you want to do it, you can
• 10:04
  go and look for Open Claw on GitHub and obviously just pull
• 10:10
  it into whatever ID you're using.
• 10:12
  You can go and search for Open Claw, clone the repo, go and
• 10:17
  find the web URL, pull it into Claw code and you'll be able to
• 10:21
  get guided through it. Quite frankly, I pick a use
• 10:23
  case. It's quite simple, you know,
• 10:25
  daily updates through my emails, for example, all that triaging
• 10:29
  of emails. That's quite a nice one to get
• 10:30
  into. If you've not got claw code in
• 10:32
  like an ID or whatever, maybe go and check out Rep or cursor.
• 10:36
  It's really, really simple. It honestly takes just a few
• 10:38
  minutes if you haven't. The experience is a bit
• 10:40
  friendlier if you're not used to being in, you know, Visual
• 10:43
  Studio Code. And again, just find the repo,
• 10:47
  get it in there, and you'll get guided through the process.
• 10:50
  The one tip I'd really give for everyone is do not run this on
• 10:54
  your main machine. So do not use this on your main
• 10:57
  laptop. It's why people have been
• 10:59
  rushing out to buy Mac minis. I think in the future, Mac minis
• 11:02
  are going to be very, very, very popular because as we need 24/7
• 11:08
  agents, you need your local machine to be turned on for an
• 11:11
  agent to be doing work. And so I think just Mac Mini's
• 11:14
  running in the background all the time and going to just have
• 11:17
  agents. And it means I think we're going
• 11:19
  to have Mac mini's always on in the background just to enable
• 11:21
  our agents to do work for us. But anyway, don't set all of
• 11:25
  this up on your local machine because there's a massive
• 11:27
  security risk. You know, I think it was like a
• 11:29
  third of all public skills made had some form of malicious
• 11:35
  software or malicious prompt injecting hiddens inside it.
• 11:39
  So do just be super careful there.
• 11:41
  Once you give this access to your PC, it can literally go and
• 11:44
  access your personal e-mail, your banking details,
• 11:47
  everything. So I'm not just saying this to
• 11:48
  be, you know, fear mongering because it's very easy to be
• 11:51
  safe, but just make sure that you don't use your local main
• 11:55
  PC. So where does this leave us?
• 11:58
  Well, Jensen Huang called Open Claw probably the single most
• 12:01
  important release of software probably ever.
• 12:04
  And the claim is obviously incredibly self-serving The it's
• 12:08
  a trillion dollar thesis on how can we sell loads of computer
• 12:12
  chips because the world is consuming loads of tokens from
• 12:16
  large language models. I think it's probably correct or
• 12:19
  at least closer to correct than it may be first sounds.
• 12:22
  This is the greatest leap towards the dream of agents
• 12:25
  that, you know, people have talked about online for so long
• 12:29
  and every model provider is racing to add open clause
• 12:32
  features to their products. So I think very, very soon
• 12:36
  you're going to see these type of capabilities democratized and
• 12:40
  accessible to everyone. Anyway, thanks for listening and
• 12:43
  I'll see you next week.