Latest / AI Fire Daily / 🎙️ EP 396: Anthropic’s $1.4T IPO Roadshow & Inkbox AI Agents on iMessage!

Transcript

🎙️ EP 396: Anthropic’s $1.4T IPO Roadshow & Inkbox AI Agents on iMessage! · AI Fire Daily

• 0:00
  $900 billion. Yeah. $900 billion in future computing
• 0:04
  commitments. It is, it's just a massive number.
• 0:09
  And a $1 .4 trillion IPO valuation. We are really
• 0:16
  looking at a truly historic financial moment
• 0:18
  right now. We definitely are. Artificial intelligence,
• 0:21
  you know, it is no longer just a software tool.
• 0:23
  It's really the core engine of the global economy.
• 0:26
  Yeah, it's a massive paradigm shift. I mean.
• 0:29
  Welcome to today's Deep Dive. We're really glad
• 0:31
  you're joining us. Today, we are unpacking Anthropic's
• 0:34
  completely unprecedented market move. We're also
• 0:38
  exploring this wild explosion of independent
• 0:40
  AI agents. Right, the digital workers. Exactly,
• 0:43
  digital workers with their own actual phone numbers.
• 0:47
  And finally, we'll look at some new real -world
• 0:49
  regulations. Because New York is, they're rushing
• 0:52
  to contain these highly capable systems. We really
• 0:55
  have a lot of fascinating ground to cover. We
• 0:57
  do. Let's just start with this anthropic IPO
• 1:00
  roadshow. Yeah, let's get into it. They're officially
• 1:02
  targeting a $1 .4 trillion valuation, which,
• 1:06
  you know, that number is just kind of hard to
• 1:08
  comprehend. It really is. I mean, it potentially
• 1:10
  makes it the biggest IPO in history. Right. They're
• 1:13
  offering 825 million shares, and those are priced
• 1:18
  between $70 and $78. So that pricing raises roughly
• 1:22
  $64 .4 billion. Yeah. And that is just the public
• 1:26
  money. Wow. Right. They're getting another $26
• 1:28
  billion from major players. So Google, Nvidia,
• 1:32
  Microsoft, and AMD are all actively chipping
• 1:35
  in. We should see their NASDAQ debut on November
• 1:37
  19th. They're going to trade under the ticker
• 1:40
  symbol ANT. But the financials backing this up
• 1:42
  are just fascinating. Yeah, the revenue figures
• 1:45
  are growing at a breakneck pace. I mean, as of
• 1:48
  October 31st, they hit this staggering number,
• 1:51
  an $81 billion annualized revenue run rate. But
• 1:56
  here's what I... am really stuck on. Okay. They
• 1:59
  have nearly 900 billion in future computing commitments.
• 2:04
  900 billion. Yeah. Why do they need almost a
• 2:08
  trillion dollars in compute? Well, because they're
• 2:11
  securing future server time and chips. They're
• 2:14
  essentially buying the raw fuel for artificial
• 2:17
  intelligence. Like stacking Lego blocks of data?
• 2:20
  Right, exactly. They're stockpiling silicon.
• 2:23
  Like it's a strategic reserve because they know
• 2:26
  these models are getting exponentially larger.
• 2:29
  At the exact same time, Anthropic announced new
• 2:31
  workplace tools. The Claude Office platform.
• 2:34
  Yeah. We're seeing tools like Claude Whiteboards.
• 2:37
  Which, you know, for collaborative visual plans.
• 2:40
  Right. And they also launched Claude Sheets to
• 2:43
  edit spreadsheets. Yeah. And there's Claude Data
• 2:45
  for building connected business dashboards. And
• 2:48
  then there is Claude Meetings. Which. That one
• 2:50
  just sounds incredibly futuristic. It's wild.
• 2:53
  It actively joins your calls. It works on tasks
• 2:57
  while teams are actually talking. It's listening,
• 3:00
  processing, and executing work in real time.
• 3:02
  Imagine a meeting where you just discuss a project
• 3:04
  plan, and the AI just builds the JIRA tickets
• 3:07
  while you speak. Microsoft shares actually dropped
• 3:09
  over 2 % right after this reveal. Oh, wow. Yeah.
• 3:12
  Wall Street clearly sees this as a direct threat.
• 3:15
  It directly challenges the dominance of Microsoft
• 3:17
  Office. Anthropic is really expanding Claude
• 3:20
  into a fully integrated workplace. They are.
• 3:22
  Oh, and they also just gave some users a huge
• 3:24
  unexpected bonus. Right, the reset. Yep. Pro,
• 3:27
  max, ultra, and team users got a free full reset.
• 3:31
  Yeah, across models and limits. You just have
• 3:33
  to claim it in your settings. Make sure you grab
• 3:35
  that before December 10th. So before launch,
• 3:38
  Anthropic also tested Claude Fable and Mythos
• 3:41
  5 .2. Right. And the resulting demo videos just
• 3:43
  went completely viral online. People were asking
• 3:47
  if it was just, you know, a language update,
• 3:49
  but then they watched the demos and realized
• 3:51
  the huge logic leap. Yeah. I mean, these models
• 3:55
  are tackling incredibly ambitious coding projects.
• 3:57
  It's amazing because I still wrestle with prompt
• 4:00
  drift myself. Oh, yeah. You know, when the AI
• 4:02
  slowly forgets my initial instructions. Right.
• 4:05
  Just kind of wanders off. Exactly. But these
• 4:07
  new models are literally running entire corporate
• 4:10
  boardrooms. They operate completely autonomously.
• 4:13
  during complex executive meetings. Right. And
• 4:15
  that completely changes how we value this company.
• 4:17
  It really does. So I am struggling to wrap my
• 4:20
  head around this valuation, though. OK. An $81
• 4:23
  billion run rate is massive, sure. But how does
• 4:26
  that actually justify a $1 .4 trillion valuation?
• 4:30
  Well, investors are looking past the traditional
• 4:32
  software market entirely. They're basically pricing
• 4:35
  in the total replacement of corporate human labor.
• 4:38
  So they aren't selling software, they are selling
• 4:41
  digital employees. Exactly. Anthropic is building
• 4:44
  the digital office of the future, but somebody
• 4:48
  else has to build the actual workforce. Right,
• 4:50
  which means the underlying infrastructure has
• 4:52
  to completely change. Yeah. We're moving from
• 4:54
  standard chatbots to entirely independent AI
• 4:57
  agents. Which brings us to Inkbox. They just
• 5:00
  raised $5 .3 million in seed funding. Led by
• 5:04
  a peak X5 partners, right? Yeah, they led that
• 5:06
  massive funding run. Because Inkbox is building
• 5:08
  what we call AI agents, which is basically software
• 5:11
  that acts on its own to complete a specific goal.
• 5:14
  Right. And they need a totally different architecture
• 5:17
  for that. Inkbox gives these agents their own
• 5:20
  permanent digital identities. They get real phone
• 5:23
  numbers and dedicated email addresses. They even
• 5:26
  get their own persistent cloud computers. This
• 5:29
  is where it gets really interesting. Every agent
• 5:31
  has dedicated access to standard communication
• 5:33
  platforms. Right. They integrate perfectly with
• 5:36
  iMessage, WhatsApp, and Slack. So they do not
• 5:39
  just live inside a browser window. No. They exist
• 5:43
  out in the real digital world. These agents automatically
• 5:46
  wake up whenever a message arrives. Let's talk
• 5:49
  about that specific sleep and wake cycle. How
• 5:52
  does that actually work on the back end? It's
• 5:54
  a brilliant technical solution to a huge problem
• 5:56
  because running a cloud computer constantly is
• 5:58
  just incredibly expensive. Right. You pay for
• 6:01
  server time by the second. Exactly. So the agent
• 6:04
  basically pauses its state into memory. It goes
• 6:07
  to sleep when it's completely idle. And it just
• 6:10
  wakes up instantly when it receives a ping. Yeah.
• 6:12
  And it responds using deeply saved context from
• 6:15
  previous interactions. Each agent gets a cloud
• 6:17
  computer that retains important files. It remembers
• 6:20
  your complete conversation history perfectly.
• 6:22
  Wow. It also manages long -term scheduled tasks
• 6:25
  entirely in the background. This is a fundamental
• 6:28
  shift in computing architecture. It really is.
• 6:30
  You're no longer opening an app to perform tasks.
• 6:33
  You're texting a digital colleague who handles
• 6:35
  it for you. And Inkbox says it is already processed
• 6:38
  to millions of automated messages. They attracted
• 6:41
  tens of thousands of users in just four months.
• 6:44
  That is all since their Y Combinator launch this
• 6:46
  year. Yeah. Businesses are rapidly adopting this
• 6:49
  specific infrastructure for daily operations.
• 6:52
  They use it for scalable, complex sales outreach.
• 6:55
  People are even deploying them as dedicated personal
• 6:59
  administrative assistants. Whoa. Imagine scaling
• 7:02
  to a billion queries. Yeah. Millions of independent
• 7:05
  cloud computers waking up and going to sleep
• 7:07
  autonomously. The infrastructure required to
• 7:10
  support that network is absolutely staggering.
• 7:12
  And the major legacy players are definitely paying
• 7:14
  close attention. Oh, for sure. ChatGPT is getting
• 7:17
  a little more Google -like lately. Yeah, OpenAI
• 7:20
  just introduced a brand new dedicated business
• 7:23
  hub. It lets brands verify their profiles and
• 7:26
  track clicks. Companies can see mentions and
• 7:28
  deploy custom agents quickly. Right. We're also
• 7:31
  seeing massive moves from international tech
• 7:33
  giants. Like Alibaba. They're rolling out Quen4
• 7:36
  with stronger native coding capabilities. It
• 7:39
  features deeper knowledge retrieval and noticeably
• 7:41
  better agentic workflows. And Quen4 Max is also
• 7:45
  currently in preview across Alibaba's internal
• 7:47
  tools. AI agents are rapidly becoming highly
• 7:51
  independent digital teammates. They have their
• 7:53
  own communication channels and cloud computers.
• 7:56
  Right. They possess the unique ability to resume
• 7:58
  work instantly. But running millions of virtual
• 8:02
  machines sounds incredibly expensive. How does
• 8:05
  a business manage the computing costs of an always
• 8:07
  -on AI? Well, like we mentioned, the system automatically
• 8:10
  hibernates the agent's cloud computer when it's
• 8:12
  idle. The system hibernates when idle, saving
• 8:15
  power until needed. Exactly. It makes the underlying
• 8:18
  economics of autonomous agents viable. Sponsor
• 8:20
  read. So because these AI agents are now interacting
• 8:23
  autonomously with the public, consequences are
• 8:26
  arriving. Real world consequences. Exactly. We
• 8:29
  are seeing heavy real world regulations stepping
• 8:31
  in globally. Stakes are getting incredibly high
• 8:34
  in heavily regulated industries. New York's new
• 8:37
  AI safety law officially kicks in January 1st.
• 8:39
  This is a massive compliance hurdle for major
• 8:42
  AI developers. Huge. Companies like OpenAI and
• 8:45
  Anthropic must officially register their models.
• 8:48
  They're legally required to publish detailed
• 8:51
  safety transparency reports. And crucially, they
• 8:53
  must report critical incidents within 72 hours.
• 8:57
  That specific reporting window is going to be
• 8:59
  incredibly difficult. Yeah. Because when an autonomous
• 9:02
  agent makes a critical mistake, it happens instantly.
• 9:05
  Right. Tracing the exact logic path of that failure
• 9:08
  is complex. Exactly. It takes engineers weeks
• 9:11
  to figure out why an AI hallucinated. But the
• 9:14
  law is demanding strict accountability from foundational
• 9:17
  creators anyway. Yeah. We're also seeing AI push
• 9:20
  aggressively into the medical field. There was
• 9:22
  a massive acquisition announced in HealthTech
• 9:24
  today. Aura, the smart ring company, is buying
• 9:28
  an AI startup. Right. They're acquiring Council
• 9:31
  Health for $625 million. This is a huge pivot
• 9:35
  from simply tracking your health. The ultimate
• 9:37
  goal is turning wearable data into actual medical
• 9:40
  care. They want to integrate AI medical guidance
• 9:43
  directly into the Aura app. And they're pairing
• 9:46
  that AI guidance with licensed human doctors.
• 9:49
  Let's walk through how this hybrid model actually
• 9:51
  works. Okay. Say your wearable device detects
• 9:54
  a physical anomaly in your biometrics. Maybe
• 9:57
  your temperature spikes and your heart rate elevates
• 10:00
  simultaneously. Right. The AI agent immediately
• 10:03
  analyzes that data against your history. Then
• 10:06
  it drafts a preliminary triage report for a physician.
• 10:09
  And then the physician reviews the AI's findings
• 10:12
  and contacts you directly. It streamlines the
• 10:15
  medical diagnostic process in a profound way.
• 10:17
  It really does. But it also introduces massive
• 10:20
  liability questions moving forward. Oh, absolutely.
• 10:23
  If the AI incorrectly triages a patient, who
• 10:26
  is legally responsible? Is it the ringmaker?
• 10:28
  Or is it the algorithm developer? Or is it the
• 10:30
  doctor who quickly reviewed the automated notes?
• 10:33
  Right. Medical advice is not an area where you
• 10:36
  tolerate prompt drift. Definitely not. That's
• 10:39
  exactly why New York is pushing strict safety
• 10:41
  laws. Governments realize these systems are moving
• 10:43
  into high stakes life or death scenarios. But
• 10:46
  I keep thinking about the new New York law. Is
• 10:50
  a 72 hour reporting window actually fast enough
• 10:53
  to contain an incident? Honestly, no. There is
• 10:56
  massive friction right now between slow regulation
• 10:58
  and instant technology. The law moves at walking
• 11:02
  pace while algorithms sprint ahead. That is the
• 11:04
  perfect way to describe the regulatory landscape.
• 11:07
  Yeah. And underneath these new laws, the foundational
• 11:10
  models themselves are rapidly evolving. That
• 11:13
  brings us to the open weight arms race happening
• 11:15
  right now. The core models powering everything
• 11:18
  are becoming cheaper and highly specialized.
• 11:21
  Google just released their absolute strongest
• 11:23
  foundational model to date. Gemini 4 Carbon.
• 11:26
  Yeah, and it's now officially available for ultra
• 11:29
  -tier users. It's also rolling out to developers
• 11:31
  and enterprise clients. Google is integrating
• 11:34
  it deeply across their entire suite. But closed
• 11:37
  models from Google face intense open -source
• 11:40
  competition today. Like Minimax. They just launched
• 11:42
  a brand new image generation model. It's called
• 11:45
  Minimax H3 Image, and it features native 4K support.
• 11:49
  Right. And because Minimax is open weight, which
• 11:52
  basically means the core code is free for developers
• 11:54
  to download and use, they iterate wildly fast.
• 11:57
  Yeah, this model is already live on platforms
• 12:00
  like Hugging Face. Developers can access it through
• 12:03
  standard application programming interfaces easily.
• 12:06
  The quality of open weight image generation is
• 12:09
  getting truly spectacular. It really is. I mean,
• 12:12
  native 4K support means the outputs are incredibly
• 12:15
  sharp and usable. And we're also seeing new specialized
• 12:18
  consumer tools emerge quickly. Like Effect Builder.
• 12:21
  Yeah, a free web tool called Effect Builder just
• 12:24
  launched recently. It allows users to turn simple
• 12:27
  text prompts into Instagram effects. U .S. users
• 12:30
  can actually publish these effects directly to
• 12:32
  the platform. Though they do have to pass an
• 12:34
  automated safety review process first. Which
• 12:36
  makes sense. It democratizes the creation of
• 12:39
  augmented reality filters for everyday users.
• 12:41
  You no longer need to know complex 3D rendering
• 12:44
  software. You just type a sentence and the AI
• 12:46
  generates the effect. It's amazing. This rapid
• 12:49
  deployment of specialized tools changes the creative
• 12:51
  landscape entirely. The sheer speed of development
• 12:55
  puts immense pressure on corporate labs. How
• 12:58
  do open -weight models like Minimax actually
• 13:00
  compete against closed giants like Gemini for
• 13:03
  Carbon? They rely entirely on the decentralized,
• 13:06
  massive speed of global community development.
• 13:09
  The open community just iterates faster than
• 13:11
  single corporate giants. Yeah. 10 ,000 independent
• 13:13
  developers tinkering simultaneously is a powerful
• 13:17
  force. Let's take a moment to synthesize everything
• 13:19
  we have covered today. Good idea. The fundamental
• 13:22
  architecture of the Internet is experiencing
• 13:24
  a tectonic shift. We are no longer simply talking
• 13:27
  directly to AI systems ourselves. Right. Artificial
• 13:30
  intelligence is now actively talking to others
• 13:33
  on our behalf. We see this with Anthropic's $900
• 13:35
  billion computing chest. They are building the
• 13:38
  digital office where AI works continuously. We
• 13:40
  see it with Inkbox giving AI agents their own
• 13:43
  phones. And we definitely see it with ATA providing
• 13:45
  automated medical triage. Yeah. The internet
• 13:48
  is shifting from a human -to -human network to
• 13:50
  an agent -to -agent network. Our digital avatars
• 13:53
  will soon negotiate and collaborate with other
• 13:56
  avatars. Humans will simply oversee the final
• 13:58
  outputs of those interactions. It makes you wonder
• 14:01
  about the long -term implications of this shift.
• 14:03
  For sure. If an AI agent now has its own permanent
• 14:06
  phone number. Right, and its own cloud computer
• 14:08
  that permanently retains all files. And it can
• 14:11
  independently conduct complex sales outreach
• 14:14
  or medical triage. At what point do we have to
• 14:17
  start treating an algorithm as a legal entity?
• 14:19
  That is the kind of profound question we will
• 14:21
  face very soon. Thank you so much for joining
• 14:24
  us on today's Deep Dive. We hope this breakdown
• 14:26
  helped you navigate the rapidly changing technological
• 14:29
  landscape. Definitely. Keep questioning the data
• 14:31
  and stay curious about the systems shaping our
• 14:33
  world.