
Aug 13, 2026 · 2h 12m
AI tests software moats as surveillance and compute reshape business
SaaSpocalypse Revisited, Singer x Louis Vuitton, Aman vs Ryan Walker | Igor Babuschkin, Brannin McBee, Garret Langley, Sonya Huang, Sean Cole
The episode connects software disruption, frontier-model competition, commercial surveillance, and privacy accountability into one argument about where durable power now sits.
- 1AI threatens application software unevenly, while infrastructure, data, network effects, and embedded customer relationships remain harder to displace.
- 2Frontier labs are competing through talent, compute, specialized capabilities, and acquisitions as AI moves from research into ordinary products.
- 3Surveillance systems create accountability questions that technology alone cannot resolve, leaving privacy protections dependent on audits and democratic oversight.
Don't miss
Igor Babuschkin explains how River AI thinks about game benchmarks, reinforcement learning, compute planning, and custom silicon while building a new lab.
The brief
TBPN revisits its SaaSpocalypse prediction and finds a more selective reckoning: vulnerable application layers face pressure, while infrastructure and embedded relationships still matter.
Twilio illustrates the distinction, with carrier relationships and operational complexity protecting it from easy replacement even as vibe coding lowers barriers to building software.
The conversation widens from luxury excess and a North Korean remote-worker operation to an Aman resort dispute and McDonald’s predictive customer dossier.
Igor Babuschkin discusses River AI’s fundraise, video-game benchmarks, reinforcement learning, compute planning, custom silicon, and the challenge of building a new AI lab.
Interviews on Flock Safety, biological computing, and AI acquisition rumors turn the episode toward privacy, specialized capabilities, and the institutional stakes of diffusion.
What was said on this episode
58 statements · 36 positive · 14 negative · 1 mixed · 7 neutral
The SaaSpocalypse erased $2 trillion in technology software market capitalization.
“$2 trillion wiped out, gone.”
Listen at 1:53
The iShares technology and software ETF rose 30% over six months.
“The iShares ETF that tracks tech and software is up 30% just over the last six months.”
Listen at 2:00
Roblox’s network infrastructure and distribution should retain future value.
“having all the Roblox network infrastructure distribution, that should be valuable in the future.”
Listen at 6:00
Marketplace dynamics, network effects, and strong go-to-market organizations resist AI disruption.
“Marketplace dynamics, network effects, strong go-to-market organizations.”
Listen at 6:14
Cybersecurity has become more important as AI adoption expands.
“Cybersecurity is more important than ever.”
Listen at 7:00
Palo Alto Networks gained 121% and reached a $320 billion valuation.
“Palo Alto Networks over the past year is up 121%. It's a $320 billion company.”
Listen at 7:06
Palo Alto Networks rose 121% over the past year.
“Palo Alto Networks over the past year is up 121%.”
Listen at 7:06
CrowdStrike reached $230 billion after rising 107% over twelve months.
“CrowdStrike's a $230 billion company up 107% this year over the past 12 months.”
Listen at 7:12
Nikesh Arora’s Palo Alto Networks investment became worth $26 million.
“It's now worth $26 million.”
Listen at 7:41
Some SaaS companies will not recover without reinventing themselves because LLMs made them obsolete.
“there are SaaSpocalypse victims that have not come back, are probably not coming back, need to completely reinvent the business because they have been made obsolete by just base level LLM capabilities.”
Listen at 9:21
Chegg has not recovered from AI disruption.
“Chegg is the canonical example of Saspocalypse victim that has not made a comeback.”
Listen at 9:50
Chegg’s stock declined 99% over five years.
“The stock is down 99% over the past five years.”
Listen at 9:54
Canva is entering a situation similar to Chegg’s AI-driven decline.
“Canva is slipping into a similar situation”
Listen at 10:50
Image models can generate many designs that users previously made with Canva.
“a lot of Canva designs can be one shot by image models like ChatGPT images, Nana Banana Pro, Grok.”
Listen at 10:53
Infrastructure SaaS companies benefit from increased AI token consumption.
“there are so many other SaaS companies that are buoying the index because they are either in the token path”
Listen at 11:32
Databases and analytics companies benefit from increased AI data generation.
“those companies are doing really well on the back of that.”
Listen at 11:57
Twilio rose 150% over the past year.
“Twilio is doing incredibly well. I don't know if you've been tracking this. It's up 150% over the past year.”
Listen at 12:02
More senior departures from MSL will occur within one or two months.
“I don't think this will be the last of the exits that we see out of MSL over the next even one or two months.”
Listen at 16:57
Instagram’s new wordmark will become normalized within days.
“this is going to be normalized probably within days, and people will just forget about it.”
Listen at 21:01
The Louis Vuitton Porsche customization was technically difficult and well executed.
“What they've accomplished clearly was incredibly difficult. And they executed their plan seemingly very well.”
Listen at 29:25
North Korean IT workers apply for American jobs at enormous scale.
“The North Koreans apply for jobs at enormous scale.”
Listen at 34:27
North Korean IT workers could plant spyware after gaining system access.
“once you let a North Korean IT worker in your systems, they could be planting all sorts of spyware.”
Listen at 38:20
Ryan Walker’s Aman video is extremely misleading and should be deleted.
“The social media video is tight. He probably should have deleted it by now because it is extremely misleading.”
Listen at 39:56
Anthropic may acquire Descartes for $6 billion.
“Anthropic is apparently in talks to buy Descartes for $6 billion.”
Listen at 52:50
- Igor Babuschkinon River AIPositive56:42
River AI aims to let users own and customize their AI systems.
“We want to figure out how we can distribute AI to everybody in a way where you own it, you're able to shape your own AI systems.”
Listen at 56:42
- Igor Babuschkinon video games as AI benchmarksPositive58:54
Video games provide complex, measurable benchmarks for AI generalization.
“games are amazing because they have all kinds of complex things you might have to do, all kinds of problem solving you might have to develop on the fly.”
Listen at 58:54
GPU prices are steadily increasing.
“GPU prices are increasing steadily.”
Listen at 1:05:20
- Igor Babuschkinon custom AI model trainingPositive1:07:21
Demand for custom AI model training will continue increasing.
“demand is going to keep going up, we expect.”
Listen at 1:07:21
- Igor Babuschkinon custom AI siliconPositive1:08:22
Custom AI silicon adoption will increase.
“I think we'll see more and more custom silicon, obviously.”
Listen at 1:08:22
- Igor Babuschkinon Personal AI agentsPositive1:08:46
Personal AI agents will drive further growth in token demand.
“We expect that the demand for tokens will go up even further.”
Listen at 1:08:46
- Igor Babuschkinon personal AI agentsPositive1:08:47
Personal AI agents will drive further growth in token demand.
“the demand for tokens will go up even further.”
Listen at 1:08:47
The creation of new AI labs is likely only beginning.
“Hopefully we're just getting started”
Listen at 1:10:15
CoreWeave significantly exceeded expectations in Q2.
“Q2 was a phenomenal beat for us”
Listen at 1:11:45
CoreWeave’s business performance will continue improving through year-end.
“balance of the year is just going to keep on improving from here.”
Listen at 1:11:55
CoreWeave uniquely delivers the best-performing AI infrastructure.
“CoreWeave is singular in its ability to deliver the best-performing infrastructure out there.”
Listen at 1:12:28
CoreWeave’s AI infrastructure has an accurate six-year depreciable life.
“Six-year depreciable life is accurate for this infrastructure.”
Listen at 1:18:41
CoreWeave is not experiencing accelerated GPU burnout or unexpected error rates.
“We're not seeing any acceleration of burnout or unexpected rates of errors.”
Listen at 1:19:41
CoreWeave’s main power constraint is delivered data-center capacity, not electricity.
“the bottleneck is less electrons for us it's more of powered shell”
Listen at 1:20:38
- Igor Babuschkinon AI demandPositive1:23:45
AI demand is currently stronger than ever.
“the demand for AI is the strongest it has ever been.”
Listen at 1:23:45
- Igor Babuschkinon AI inferencePositive1:23:54
AI inference is profitable for CoreWeave’s clients.
“Inference is absolutely profitable for our client base”
Listen at 1:23:54
Vandalism of Flock cameras is a persistent problem.
“vandalism is real. It's always been real, though.”
Listen at 1:29:29
Lawmakers should decide national rules governing Flock-style surveillance.
“I think it should be decided by our lawmakers because it comes down to our fundamental rights as American citizens.”
Listen at 1:38:07
Nine state bills concerning Flock-related policy passed this year.
“we had nine state bills get passed this year.”
Listen at 1:39:48
AI application companies are increasingly building internal research labs.
“all the application companies are starting to build their own research capabilities, their own labs.”
Listen at 1:46:35
- Igor Babuschkinon small AI research teamsPositive1:52:37
Small AI research teams can achieve substantial results.
“Very, very small teams can get very far.”
Listen at 1:52:37
- Igor Babuschkinon Harvey research teamNeutral1:52:53
Harvey’s research team consists of seven people.
“their entire research team is seven people.”
Listen at 1:52:53
- Igor Babuschkinon companies’ AI strategyPositive1:54:59
Companies should retain highly capable internal leaders for AI strategy.
“Companies need to have... extremely smart people in charge of this in-house.”
Listen at 1:54:59
Falling AI prices can increase application companies’ gross margins as usage rises.
“their gross margins going up because at the same time as like AI usage is going way up”
Listen at 1:56:19
- Sean Coleon PrasmaPositive1:58:33
Prasma trains biological brain cells to perform computation.
“we are training brain cells for compute.”
Listen at 1:58:33
- Igor Babuschkinon Prasma token prediction systemNeutral1:59:08
Prasma uses approximately 200,000 brain cells for token prediction.
“We're using about 200,000 brain cells to do token prediction.”
Listen at 1:59:08
- Igor Babuschkinon Prasma biological computing systemPositive2:00:53
Prasma’s biological system achieved 78%, exceeding silicon on a constrained task.
“we got 78%. We beat silicon on this very constrained task”
Listen at 2:00:53
- Igor Babuschkinon Prasma cultured brain cellsNegative2:01:10
Prasma’s cultured brain cells currently last six months.
“So I think currently they last six months.”
Listen at 2:01:10
- Igor Babuschkinon human brains versus siliconPositive2:02:15
Human brains are more power-efficient than silicon computing.
“Human brains are extremely power efficient in comparison to silicon.”
Listen at 2:02:15
- Igor Babuschkinon Prasma automated laboratoryPositive2:05:30
Prasma’s automated lab could expand into drug testing.
“the automation aspect of our lab can easily be branched out into different things like drug testing.”
Listen at 2:05:30
Workday’s valuation reached approximately $50 billion.
“And so now a $50 billion company.”
Listen at 2:07:47
Workday may pursue AI transformation more easily as a private company.
“I can imagine there's a bunch of AI transformation that they'd feel like they'd be more easily able to do as a private company.”
Listen at 2:08:02
AI will become less discussed if it fully diffuses into ordinary products.
“If it diffuses fully, you stop talking about the underlying technology.”
Listen at 2:11:19
The AI investment race could affect the entire global economy.
“The entire global economy, potentially.”
Listen at 2:11:27
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.