Latest / Today’s AI News / The $1.8B Solo Founder Is Here, OpenAI Buys a Media Company, Google Goes Open Source

Transcript

The $1.8B Solo Founder Is Here, OpenAI Buys a Media Company, Google Goes Open Source · Today’s AI News

• 0:00
  You know, building a billion dollar company used to be, well,
• 0:03
  like constructing a massive skyscraper with an absolute army
• 0:07
  of specialized workers. But today AI has basically given
• 0:11
  one guy the master control panel to build that exact same
• 0:15
  skyscraper just sitting right there on his living room sofa.
• 0:19
  We have officially entered the era of the solo billion dollar
• 0:22
  founder, so welcome back to today's AI news brought to you
• 0:26
  by 9X Productions. It is.
• 0:29
  It's going to be a really good deep dive today.
• 0:31
  Yeah. For today's Deep Drive, our
• 0:32
  mission is to explore how these, you know, theoretical promises
• 0:36
  of AI, from massive corporate strategy down to individual
• 0:40
  everyday leverage, are suddenly becoming concrete, undeniable
• 0:44
  realities. Right, we are moving way past
• 0:46
  the whiteboard phase. Exactly.
• 0:47
  And we are skipping the fluff today.
• 0:48
  We're diving immediately into the latest developments, hitting
• 0:51
  you with some rapid fire updates, and then breaking down
• 0:54
  a community workflow you can use literally right now.
• 0:56
  Sounds perfect. So let's unpack that opening
• 0:59
  analogy because we have been hearing tech leaders predict the
• 1:02
  solo billion dollar company for a long time.
• 1:05
  I mean, Sam Altman specifically called this back in 2024.
• 1:08
  You did, Yeah. And now we finally have the
• 1:11
  proof. In Wild, I want to break down
• 1:13
  the story of Matthew Gallagher and his startup Med V.
• 1:16
  The numbers on this are just, well, they're staggering.
• 1:18
  They really are. It took Gallagher just two
• 1:21
  months and a $20,000 initial investment to stand this entire
• 1:26
  operation up. Wow, right.
• 1:28
  And Fast forward a year and a half.
• 1:29
  The New York Times is reporting that Med V is projecting $1.8
• 1:33
  billion in annual sales. That is just wild.
• 1:37
  Yeah, and they sell GLP 1 weight loss drugs online, but here's
• 1:41
  where it gets really interesting.
• 1:42
  The only full time hire on his entire team is his brother.
• 1:46
  Which is, I mean, to really grasp the magnitude of this, you
• 1:49
  have to look at the traditional barrier to entry in
• 1:52
  pharmaceutical distribution or telehealth.
• 1:55
  Right. It's usually massive.
• 1:56
  Exactly. A company projecting nearly 2
• 1:58
  billion in sales would historically require so much
• 2:01
  overhead. You'd need floors of cubicles,
• 2:03
  an entire human resources department just to manage
• 2:06
  benefits for, you know, hundreds of customer service reps,
• 2:09
  developers, marketers. And he just bypassed all of
• 2:11
  that. Completely Gallagher bypassed
• 2:14
  all that legacy infrastructure by orchestrating A comprehensive
• 2:17
  AI stack to basically replace an entire corporate workforce.
• 2:22
  So walk us through exactly how that stack functions
• 2:25
  mechanically. Like how do you replace an
• 2:27
  entire executive suite and frontline workforce with just
• 2:31
  software? Well, he essentially
• 2:32
  compartmentalized the company into departments and then just
• 2:35
  assigned AI models to run them. OK, so for the engineering and
• 2:39
  coding back end, instead of hiring a massive team of
• 2:41
  developers to build the platform, he used a combination
• 2:44
  of ChatGPT, Claude, and Grok to write and debug the code.
• 2:49
  Just. Bouncing between them.
• 2:50
  Right. And then for the marketing
• 2:51
  department, which you know requires constant generation of
• 2:54
  ad creatives, he deployed mid journey and runway to produce
• 2:57
  these highly polished visual assets.
• 2:59
  That saves so much time. But the most impressive
• 3:01
  mechanism, honestly, is how he handled customer service.
• 3:05
  Right, because in telehealth, patient intake and support are
• 3:08
  massive bottlenecks. Exactly.
• 3:10
  So he built custom AI agents and integrated them with 11 labs for
• 3:14
  voice synthesis. This allowed the AI to handle
• 3:17
  patient inquiries in real time using natural language 24/7.
• 3:21
  Wait, let me stop you there and push back a little on this
• 3:23
  narrative. Sure, go ahead.
• 3:25
  Is he truly a solo founder if he's, you know, outsourcing the
• 3:29
  doctors, the medical prescriptions, and the actual
• 3:32
  physical shipping of these drugs to 3rd party telehealth
• 3:36
  platforms like Care Validate and Open Loop?
• 3:38
  That's a fair question. I mean he isn't physically
• 3:41
  mailing out GLP ones himself. Isn't he just playing the
• 3:43
  ultimate general contractor here?
• 3:46
  Like he's coordinating the digital side, but other human
• 3:48
  beings are still doing the heavy lifting out in the real world.
• 3:51
  That is a very valid distinction, but framing it that
• 3:54
  way actually helps redefine what a modern company fundamentally
• 3:57
  is. Well, the core operation, the
• 4:00
  strategy, the marketing engine and the execution, the business
• 4:02
  model that was all generated and managed solo.
• 4:05
  OK, I see. Relying on API driven real world
• 4:09
  infrastructure like Care, Validate or Open Loop.
• 4:12
  It's kind of the physical equivalent of a software company
• 4:15
  relying on Amazon Web Services to host their website.
• 4:18
  Oh right, you don't build your own server farm anymore, so why
• 4:21
  build your own shipping network if an API can trigger a third
• 4:23
  party to do it? Precisely.
• 4:25
  The breakthrough here isn't that Gallagher built some
• 4:29
  revolutionary new foundational AI model from scratch.
• 4:33
  It's that he aggressively applied existing AI tools to a
• 4:36
  traditional, highly regulated business model.
• 4:39
  He had really strong builder instincts, recognize the extreme
• 4:43
  leverage that AI and AP is provide when you stitch them
• 4:46
  together, and he just executed. Which perfectly aligns with the
• 4:49
  prediction of the solo Unicorn. Exactly, the first billion
• 4:52
  dollar solo founder isn't an AI researcher, they are a highly
• 4:56
  leveraged operator who knows exactly how to connect the dots.
• 4:59
  A highly leveraged operator. I love that that frames AI
• 5:02
  perfectly as a force multiplier rather than just, you know, a
• 5:05
  stand alone product. It really does.
• 5:07
  So we just talked about an individual using the AI to
• 5:10
  bypass traditional gatekeepers and achieve massive scale.
• 5:13
  Let's flip that on its head and look at a giant AI incumbent
• 5:17
  trying to bypass traditional media gatekeepers to control its
• 5:21
  own massive scale and public narrative.
• 5:23
  Open AI just made its very first media acquisition.
• 5:28
  The media play. This is such a fascinating pivot
• 5:30
  in corporate communications. Yeah, they are buying TBPN,
• 5:34
  which is a daily live tech talk show, for what is reportedly in
• 5:38
  the low hundreds of millions. And the growth of the show is
• 5:40
  nuts. It really took off.
• 5:42
  Co founders Jordy Hayes and John Coogan only debuted TBPN like 17
• 5:47
  months ago. Wow.
• 5:48
  And they've already scaled it to 70,000 daily viewers across
• 5:51
  YouTube and X. The company was reportedly on a
• 5:54
  $30 million revenue pace, and now Open AI owns them outright.
• 5:59
  So the logistics of this corporate integration tell a
• 6:01
  very compelling story about open AI priorities right now.
• 6:04
  Break that down for us well. The entire eleven person team at
• 6:08
  TVPN is being brought in house and they will immediately drop
• 6:11
  their ad business. Which makes perfect financial
• 6:13
  sense since open AI obviously doesn't need to monetize through
• 6:16
  mattress commercials. Right.
• 6:17
  Exactly. But the critical detail is that
• 6:20
  they will reportedly retain editorial independence on the
• 6:23
  show, and they'll be reporting directly to Open a Eyes chief of
• 6:27
  global affairs Chris Lafane. Interesting.
• 6:29
  Yeah, and Fiddy Simo, who sits on Open a Eyes board, offered a
• 6:34
  really revealing quote about this.
• 6:36
  She stated that because Open AI is driving such a massive
• 6:39
  technological shift, the standard communications playbook
• 6:43
  just doesn't apply to them anymore.
• 6:45
  I have an analogy for this, and I want you to tell me if I'm
• 6:47
  being, you know, too cynical here.
• 6:49
  Let's hear it. To me, this feels a bit like a
• 6:51
  politician buying the local newspaper but promising they
• 6:54
  won't read the articles before they print.
• 6:56
  I mean is open AI just buying high NPR here or are they
• 7:00
  genuinely trying to buy a community?
• 7:02
  That is a brilliant comparison honestly, and your skepticism is
• 7:05
  totally warranted. To understand the mechanism
• 7:08
  behind this acquisition, you really have to look at the
• 7:10
  current tech media landscape and open a eyes.
• 7:13
  Recent history because. They've had a rough year.
• 7:15
  Exactly. Over the past year, Open a Eyes
• 7:18
  public perception is taking some significant hits.
• 7:21
  Executive departures, safety controversies, regulatory
• 7:25
  scrutiny, all of it. And traditional public
• 7:28
  relations, you know, just sending out a press release to
• 7:31
  mainstream journalists that leaves the company vulnerable to
• 7:35
  how those journalists interpret and frame the narrative.
• 7:37
  And they lose control. Right, so by buying TBPN, Open
• 7:41
  AI is acquiring a direct, unfiltered channel to the
• 7:44
  cultural tech vibes. Basically, they are buying the
• 7:47
  audience. They are buying a highly
• 7:49
  concentrated, incredibly influential audience.
• 7:51
  I mean that 70,000 daily viewership isn't just a random
• 7:54
  sample of the public. TBPN is the go to show for
• 7:57
  Silicon Valley CEO's, venture capitalists, and top tier
• 8:00
  developers. Those are the exact people
• 8:03
  building products on top of Open AI as API.
• 8:05
  So by acquiring the show, Open AI embeds its narrative directly
• 8:10
  into the daily conversation of the tech elite, completely
• 8:13
  bypassing that traditional media filter.
• 8:15
  So. Smart it is, but whether that
• 8:17
  promised editorial independence actually holds up the next time
• 8:21
  Open AI faces a major controversy, well, that's the
• 8:24
  real stress test we will all be watching.
• 8:26
  Yeah, we'll see how independent they really are.
• 8:29
  But Speaking of, you know, editing the final product and
• 8:32
  tweaking reality, let's move from media narratives to graphic
• 8:35
  design. OK, a big shift.
• 8:36
  Everyday users are suddenly getting unprecedented granular
• 8:40
  control over AI generations, and it is going to completely change
• 8:44
  how you create visual content. Canva just rolled out a new
• 8:47
  feature called Magic Layers. Oh, I saw this.
• 8:49
  This is one of those features that sounds like a simple
• 8:52
  quality of life update, but mechanically it completely
• 8:55
  eliminates the biggest bottleneck in AI image
• 8:58
  generation. Yeah, so here's how it works.
• 9:00
  You take any flat AI generated image, and we all know the
• 9:03
  struggle of getting a stunning AI image, but there's a
• 9:06
  misspelled word on a sign or some weird deformed artifact
• 9:09
  floating in the background. With magic layers, Canva reads
• 9:12
  the flat image and splits the background, the text, and the
• 9:16
  distinct objects into individual, fully editable
• 9:19
  layers in about 30 to 60 seconds.
• 9:22
  Amazing. It's literally really like
• 9:23
  unbaking a cake to change the flavor of the flower.
• 9:27
  You don't have to reroll the prompt roulette wheel just to
• 9:29
  fix 1 typo anymore. And that prompt roulette is
• 9:32
  exactly the friction point. I mean, if you have ever spent
• 9:35
  an hour generating the perfect lighting and composition only to
• 9:39
  realize the AI mangled the text, you know the pain it's.
• 9:42
  The worst? Because before this, if you
• 9:44
  change your text prompt to fix the spelling, the AI would
• 9:48
  generate an entirely new image based on a new mathematical
• 9:51
  seed. Right.
• 9:52
  The lighting changes, the subject moves, the whole
• 9:54
  aesthetic is ruined. Exactly so mechanically, Magic
• 9:58
  layer solves this using advanced semantic segmentation.
• 10:01
  Break that down for us. What is semantic segmentation
• 10:04
  actually doing behind the scenes?
• 10:05
  So the AI isn't just cropping the image, it is mathematically
• 10:09
  recognizing the boundaries of discrete objects.
• 10:12
  It's identifying what is a car, what is a tree, what is text.
• 10:16
  It separates those vectors, and it actually infills the pixels
• 10:20
  behind the object so you don't have a blank hole in your
• 10:23
  background when you move things around it.
• 10:24
  Is so. Clever.
• 10:25
  It really bridges the massive gap between raw AI generation
• 10:29
  and a professional graph design workflow.
• 10:32
  You can swap out a date on a flyer, change a tagline, or drop
• 10:35
  in a real product photo without losing the original AI generated
• 10:39
  aesthetic. It effectively brings the
• 10:41
  granular control of a senior Photoshop editor to anyone who
• 10:45
  can just point and click. It's wild how fast consumer
• 10:48
  tools are maturing. It really is.
• 10:50
  But none of this, not the solo founders marketing materials,
• 10:53
  not the editable Canva images, none of it happens without the
• 10:57
  foundational open source models powering them.
• 11:00
  Very true. And there is a high stakes
• 11:01
  geopolitical chess game happening right now in that
• 11:04
  foundation layer. Google DeepMind just released
• 11:07
  the Gemma 4 family of open models.
• 11:08
  This is a major strategic play by Google.
• 11:11
  It shakes up the entire open source ecosystem.
• 11:13
  Yeah, these models are highly versatile.
• 11:16
  They range in size from something small enough to run
• 11:18
  natively on your smartphone all the way to massive computer
• 11:22
  scale models. They handle complex coding,
• 11:25
  computer vision, multi step agents and the smallest variants
• 11:28
  even process offline voice directly on your device and
• 11:31
  their performance is top tier. The benchmarks are really
• 11:34
  impressive. Yeah, the. 31 billion and 26
• 11:36
  billion parameter models are placing right next to
• 11:39
  heavyweights like Kimi K 2.5 and Alibaba's Kin 3.5 in terms of
• 11:43
  raw intelligence, but at a fraction of the computational
• 11:47
  size. Exactly.
• 11:48
  But wait, I have to jump in here and question the business logic
• 11:51
  because there is a massive detail regarding the license.
• 11:53
  The Apache license, yes. Google switched from their usual
• 11:56
  custom license to an Apache 2 Point O license for Gemma 4 for
• 12:00
  anyone who isn't a developer. And Apache 2 Point O license
• 12:03
  means you can modify the model, deploy it, and sell it
• 12:05
  commercially with 0 legal friction.
• 12:07
  You owe Google nothing, right? If Google is handing out models
• 12:11
  that rival the absolute best on the market, they are essentially
• 12:14
  arming their own competitors. What is the hidden catch here?
• 12:18
  Well, to understand the catch, you have to look at the broader
• 12:22
  frontier model landscape and the concept of an ecosystem Moat.
• 12:26
  OK, explain that. Over the last year, Chinese open
• 12:29
  source models, specifically the Quinn family from Alibaba, have
• 12:34
  heavily dominated the open source developer space.
• 12:37
  Developers flocked to them because they were powerful and
• 12:40
  very accessible. But recently we are seeing those
• 12:44
  Chinese AI labs pulling up the ladder, slowly moving toward
• 12:49
  closed proprietary systems so they can actually monetize their
• 12:53
  research. So they got everyone hooked and
• 12:55
  now they're closing the gates. Exactly.
• 12:58
  So Google sees an opening to capture that orphan developer
• 13:01
  base by dropping all commercial friction with an Apache 2 Point
• 13:05
  O license. Google along with other U.S.
• 13:08
  companies like RCAI who just dropped their Trinity Large
• 13:11
  model, they are heavily incentivizing the global
• 13:14
  developer community to build on their architecture.
• 13:16
  Oh, I see. Think of it like giving everyone
• 13:18
  free blueprints to build the city, knowing that you own the
• 13:21
  only power plant that can turn the lights on.
• 13:23
  That's a great way to put it. Google is intentionally
• 13:25
  commoditizing the model layer to win developer mindshare.
• 13:30
  If developers build their apps, their complex workflows, and
• 13:33
  their entire businesses around Google's free models, eventually
• 13:36
  those businesses will scale, right?
• 13:38
  And when they scale, they will inevitably need to buy Google
• 13:42
  Cloud Compute to run those applications.
• 13:44
  That ecosystem play makes total sense when the developer today
• 13:48
  own the enterprise cloud contracts tomorrow.
• 13:51
  Precisely. And that open source ecosystem
• 13:54
  play by companies like Google is exactly why we are seeing such
• 13:57
  explosive rapid fire breakthroughs in the broader
• 14:00
  development world right now. I mean, look at what developers
• 14:02
  are doing with these models just today.
• 14:04
  Let's do a speed run of the ripples hitting the tech world.
• 14:06
  Time for our quick hits. Let's do it.
• 14:08
  The pace of these modality updates is staggering.
• 14:10
  First up in Video Byte Dances, Sedan's two Point O AI video
• 14:14
  generator is now broadly available across major platforms
• 14:18
  and it has immediately taken the absolute top spot on the
• 14:21
  artificial analysis leader boards for visual quality and
• 14:24
  physics consistency. That's huge and over the coding
• 14:28
  world. Cursor just unveiled built
• 14:30
  Cursor 3. This is a completely rebuilt
• 14:32
  developer interface that allows users to run parallel fleets of
• 14:36
  local and cloud agents. Wow.
• 14:38
  Yeah, Mechanically this means instead of an AI testing one
• 14:41
  line of code at a time, it can simulate thousands of testing
• 14:45
  variations simultaneously across multiple repositories, all from
• 14:48
  1 workspace. Sticking with coding, Alibaba
• 14:51
  dropped Quinn 3.6 Plus. This model rivals Opus 4.5
• 14:56
  encoding benchmarks, but the big story is that it natively
• 14:58
  supports a massive 1,000,000 token context window.
• 15:02
  Which is incredible. Yeah, for context, 1,000,000
• 15:05
  tokens is roughly 750,000 words. That means a developer doesn't
• 15:09
  have to build complex search pipelines to query their
• 15:11
  software. The AI can literally hold an
• 15:13
  entire massive code base in its short term working memory all at
• 15:16
  once. It changes everything now in the
• 15:18
  audio space. Microsoft launched Mai
• 15:20
  Transcribe one in public preview.
• 15:22
  This is a new speech to text model that is currently topping
• 15:25
  accuracy benchmarks across 25 different languages, capturing
• 15:28
  nuance and dialects with near flawless precision.
• 15:31
  Finally, in the realm of persistent agents, Japanese AI
• 15:35
  startup Sakana AI has opened beta testing for Marlon.
• 15:39
  This is an autonomous AI assistant that doesn't just
• 15:42
  answer a prompt and go to sleep. It can work an 8 hour shift
• 15:45
  autonomously executing business related tasks across different
• 15:49
  applications. You know what is truly
• 15:51
  fascinating for anyone paying attention to these updates is
• 15:54
  the sheer convergence. Yeah, it's all coming together.
• 15:58
  We just jumped from video generation to parallel coding
• 16:01
  fleets, to massive working memory models to flawless audio
• 16:05
  transcription to persistent autonomous agents.
• 16:08
  All of these distinct modalities are accelerating at the exact
• 16:12
  same time. It's not just a chat box
• 16:13
  anymore. Exactly.
• 16:14
  It isn't just text on a screen. The models are learning to see,
• 16:18
  to hear, to reason over massive data sets, and to work
• 16:21
  independent shifts. And when you combine all those
• 16:23
  distinct modalities, just like Matthew Gallagher did to build
• 16:26
  his $1.8 billion company, the output becomes exponential,
• 16:30
  which brings us perfectly to our final segment.
• 16:32
  We love to tie all this high level billion dollar technology
• 16:35
  back down to your reality. Right, practical application.
• 16:38
  How are actual people using this to save time right now?
• 16:41
  Let's dive into community AI workflows.
• 16:44
  Seeing how individuals route these APIs to solve daily
• 16:48
  friction is always my favorite part of the day.
• 16:49
  Today's workflow comes from a listener, Philip H over in
• 16:52
  Barcelona, Spain. Philip uses a Claude Co work
• 16:55
  automation that he meticulously integrated with his HubSpot CRM,
• 16:59
  Slack, Gmail, Google Calendar, and fellow AI.
• 17:04
  That is a serious multi platform integration stack.
• 17:08
  Integrating A reasoning engine across that many data silos is
• 17:11
  incredibly powerful. Yeah, so he set it up so that it
• 17:13
  fires automatically at 8:00 AM every single work day.
• 17:17
  While he is having his coffee, the AI is reading his emails,
• 17:21
  checking his Slack mentions, cross referencing his CRM for
• 17:24
  client updates, and looking at his calendar.
• 17:26
  That's brilliant. It synthesizes all that
• 17:28
  scattered data to provide him with one unified briefing on any
• 17:31
  missed items, and it lays out exactly what his priorities are
• 17:34
  for the day. If you've ever felt overwhelmed
• 17:36
  in the morning trying to piece together what happened while you
• 17:39
  were asleep, you know the friction Philip just solved.
• 17:41
  He has taken the immense cognitive load of triage, you
• 17:45
  know, the mental energy spent just figuring out what to do,
• 17:48
  and he completely outsourced it to an AI agent.
• 17:51
  The big take away here for you listening is that Philip is
• 17:54
  essentially acting like his own solo founder.
• 17:56
  He didn't hire a human executive assistant to brief him every
• 18:00
  morning. He built his own out of AP.
• 18:02
  Is it is the micro version of the billion dollar startup we
• 18:05
  talked about at the top of the show?
• 18:07
  The barrier to entry for achieving extreme leverage has
• 18:10
  filed into near 0. You no longer need venture
• 18:14
  capital or a massive hiring budget.
• 18:16
  You just need curiosity and a bit of time to connect the AP.
• 18:19
  Is we covered a massive amount of ground today, from individual
• 18:22
  workflow automation to the commoditization of the global
• 18:25
  tech stack. We really did, and I want to
• 18:27
  leave you with a thought to Mull over as you go about your day.
• 18:31
  If Matthew Gallagher can build a $1.8 billion company from his
• 18:35
  living room by orchestrating AI models, and Philip in Barcelona
• 18:40
  is already automating the cognitive load of his morning
• 18:42
  executive prep, in five years, will the standard unit of a
• 18:46
  company be measured by its employee headcount, or will it
• 18:50
  be measured purely by its compute power?
• 18:52
  Compute power as the new headcount that completely shifts
• 18:55
  how we have to think about business scale and human
• 18:57
  leverage. I love that.
• 18:59
  Well, that is all the time we have for today's deep dive.
• 19:01
  If you want to stay updated on all these rapid shifts and learn
• 19:04
  how to apply them, make sure you hit that subscribe button.
• 19:06
  And if you enjoyed the conversation, please leave us a
• 19:09
  five star rating. It really helps the show.
• 19:11
  Thank you so much for joining us, and we'll see you back for
• 19:13
  tomorrow's deep dive.