Mar 1, 2024 · 1h 26m · allin
E168: Can Google save itself? Abolish HR, AI takes over Customer Support, Reddit IPO teardown
Jason Calacanis · 20m spoken Chamath Palihapitiya · 20m spoken David Friedberg · 16m spoken David Sacks · 16m spoken Matt Friend · 1m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In Episode 168 of the All-In Podcast, the hosts analyze Google's organizational struggles amidst the Gemini AI controversy, Chamath's proposal to abolish traditional HR departments, the economic shift toward AI data licensing (TAC 2.0), and the disruptive impacts of AI automation demonstrated by Klarna's support agents and Reddit's IPO.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 97.4% of the talking time here. How this is scored →
The hosts as informed peer 6.7 Guest teaching 3.6 Guest disagreement 2.5 The hosts pushing back 2.7
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 46:13 Friedberg pushes back on TAC 2.0 analogy
Friedberg directly rejects Chamath's premise that AI data licensing mirrors traditional search traffic acquisition cost (TAC), arguing instead that it represents finite content licensing similar to Netflix.
Hardest push from the hosts ▶ 17:52 Friedberg challenges Sacks on Google's Gemini failuresFriedberg explicitly refuses Sacks' framing of Google's Gemini DEI issues as purely partisan malice, arguing it stems from structural fear of legal civil rights liability.
Biggest teaching moment ▶ 47:04 Friedberg drops internet data volume metricsFriedberg educates the panel on data monetization realities by citing exact macro metrics (1M total petabytes of web data and 2,500 petabytes generated per day).
The host holds their own ▶ 1:19:55 Sacks breaks down Sam Altman's Reddit corporate maneuverSacks demonstrates insider historical expertise by recounting Yishan Wong's detailed account of how Sam Altman and Reddit's founders engineered a multi-step spinout from Condé Nast.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Pre-Show Banter & House Guest Discussion | 2 | 1 | 1 | 1 | Light pre-show banter and housekeeping where the co-hosts joke about Jason being a professional house guest and compare him to Kato Kaelin. Tone is relaxed and lighthearted with no real disagreement. | |
| Chamath's Grok Win & Developer Platform Growth | 6 | 2 | 2 | 2 | Chamath highlights Grok's developer platform growth, citing waitlist metrics tripling to nearly 10,000 developers as a key North Star metric. The group briefly discusses memory safety guidelines from the White House regarding C and C++. | |
| Google Gemini DEI Controversy & Sundar Pichai's Leadership | 7 | 5 | 3 | 4 | Friedberg details Google's internal financial structure, contrasting search profit margins against YouTube and Cloud to explain investor frustration over Gemini's DEI blunder. He shares internal anecdotes regarding employee silence around the Responsible AI team. | |
| Institutional Fear, Political Monoculture & DEI at Google | 8 | 4 | 5 | 5 | Sacks analyzes Google's Gemini controversy, citing comms analysis and drawing a historical parallel between Google's DEI officers and Soviet military commissars. Friedberg pushes back by framing the problem as structural legal fear rather than pure political malice. | |
| Chamath's 'Abolish HR' Strategy & Corporate Meritocracy | 8 | 6 | 3 | 3 | Chamath presents his operational framework of eliminating internal HR departments in majority-controlled companies, replacing them with outside legal counsel, employee committees, and annual performance rank-ordering. Sacks concurs and outlines startup meritocracy principles. | |
| AI Content Licensing Deals & Reddit IPO S-1 Teardown | 8 | 4 | 2 | 2 | Chamath introduces the concept of TAC 2.0, contrasting historical traffic acquisition costs (like the early 2000s Google-AOL deal) with modern AI data licensing agreements such as Google's deal with Reddit. | |
| Data Monetization for Small Websites vs. Platforms | 8 | 6 | 4 | 5 | Friedberg pushes back on Chamath's TAC 2.0 framing, arguing data licensing mirrors media content licensing (such as Netflix) rather than continuous traffic acquisition, backing his point with global internet data generation statistics. | |
| Terms and Exclusivity in AI Data Licensing | 7 | 5 | 4 | 4 | The panel debates exclusivity terms in AI licensing agreements. Friedberg points out why Reddit would avoid exclusive deals to preserve valuation upside, while Sacks contrasts fragmented publisher strategies with the music industry's ASCAP model. | |
| Klarna's OpenAI Customer Support Assistant Milestone | 7 | 3 | 2 | 3 | The co-hosts analyze Klarna's deployment of OpenAI customer support assistants that replaced 700 agents. Friedberg presents a macro-economic perspective on labor shifts from knowledge tasks to creative work, while Sacks outlines Level 1 vs Level 2 support automation. | |
| Market Impact on Call Centers and Open Source Strategy | 8 | 4 | 3 | 2 | Chamath highlights the market fallout of Klarna's announcement, noting call center provider Teleperformance lost $1.7 billion in market cap. He argues Klarna should open-source its tooling, while Sacks links this to Meta's open-source strategy. | |
| Disruption of Enterprise SaaS and 1-Person Unicorns | 7 | 3 | 2 | 2 | Chamath discusses how open-source AI software disrupts traditional SaaS products like Zendesk by drastically reducing operating expenses, enabling capital-efficient single-person unicorn startups. | |
| Voice AI Call Centers and Political Roleplay | 6 | 2 | 1 | 1 | Sacks and Chamath explore voice AI replacement of IVR phone trees and real-time dialect matching. The segment includes a comical Donald Trump customer service roleplay between Jason and Chamath. | |
| Comedian Matt Friend Celebrity Impressions Showcase | 2 | 1 | 1 | 1 | An entertainment segment showcasing a video clip of comedian Matt Friend doing impressions of Donald Trump, Howard Stern, and Stanley Tucci, followed by brief praise from the hosts. | |
| Reddit S-1 IPO Filing Breakdown | 9 | 5 | 3 | 3 | Friedberg breaks down Reddit's S-1 financial metrics, highlighting high gross margins against flat ARPU ($3.42) compared to Meta ($40). Sacks details the historical narrative behind Sam Altman and founders engineering Reddit's spinout from Condé Nast. | |
| Apple Cancels Project Titan EV to Pivot Toward Generative AI | 7 | 3 | 2 | 2 | The hosts cover Apple shuttering Project Titan (its EV project) to redirect resources toward generative AI, reflecting on historical acquisition talks between Apple and Tesla during the Model 3 ramp. |
Statements from this episode (34)
Calacanis: Palihapitiya invested in AI chipmaker Groq seven years ago
“He invested in a company seven years ago that is absolutely crushing right now called Grok.”
Palihapitiya: Groq's developer waitlist tripled to nearly 10,000 in one week
“By the way, developers, by the way, on Grok just this past week, In the queue, the waitlist tripled. Now there's almost 10,000 developers.”
Friedberg: Search ad revenue generates ~100% of Google's true operating profit
“And so the search margin, the ad revenue on search, Is, you know, probably a hundred percent of the true operating profit of the business.”
Friedberg admits to selling his Meta stock during its late 2022 slump
“By the way, I was one of the stupid people to sell Meta around that time.”
Friedberg: Alphabet trades at 17x 2025 earnings, cheaper than all tech peers
“It's trading at just 17 times 20 25 consensus earnings, which is cheaper than all the other big tech companies by far, and it's still growing. Core business is growing. Cloud is growing 20%. Search is growing 15%. All these other businesses, YouTube's growing …”
Palihapitiya: Google market cap will drop 50% if search share drops 1%
“If you see perplexity or anybody else clip off 50 basis points or a hundred basis points of share in search. This thing is going straight down by 50%.”
Sacks: Over 90% of Google employee political contributions go to Democrats
“When everybody is left to center, I mean, just look at something like political contributions, right? It's 90 something percent are democratic. I mean, like high nineties.”
Palihapitiya: I eliminated HR departments across all majority-controlled portfolio companies
“I, there is no company where I have majority control where I have an HR department.”
Palihapitiya: Companies must manage out or fire their bottom 5-10% annually
“Every year. And it does not matter how big the company is. You must manage up or out the bottom five to 10%.”
Sacks: Large HR teams and DEI organizations contradict startup success
“Number one, the ones that are successful have a cultural meritocracy. Number two, they're non-bureaucratic, and they don't have too much G&A, basically overhead in the company. Now, all those things are contradicted by having a large HR team or especially a DE…”
Calacanis: Reddit's AI licensing deals total $200M over 2-3 years
“According to Red's S-one, they have already closed two hundred million worth of AI licensing deals over the next two to three years.”
Palihapitiya: Google pays Apple $18B to $20B annually for default search
“On the iPhone, it's somewhere between 18 and 20 odd billion dollars a year is what now Google pays Apple.”
Palihapitiya: Unique web and app data will yield new licensing revenue
“If you're an entrepreneur building a website or building an app that has really unique training data or really unique data, you'll be able to license and sell that. And that'll be an incremental revenue stream to everything you do in the near future.”
Palihapitiya: AI data licensing revenue will extend to small websites
“And in this tack two point, oh, world, I do think that, that it works in that way as well.”
Palihapitiya: Average small website will earn only hundreds from AI licensing
“What Google will be able to pay you will probably be pretty de minimisly small if you're small right now. So, you know, to your point, if the upper bound is sixty million for Reddit, then the average website is going to get a few hundred bucks”
Friedberg: Google's ad syndication generates $10B quarterly with 70-80% payouts
“That business makes about ten billion a quarter at Google. And they're paying, call it, 70 to 80 cents on every dollar back out to the owners of that traffic, the folks where that traffic is derived from.”
Friedberg: Humans generate 2,500 petabytes of new internet data per day
“There's about a million petabytes of data on the internet today. And humans are generating about 2500 petabytes of data, new data per day right now.”
Friedberg: AI companies will learn to price and license data like Netflix
“Just like over time, Netflix got better at figuring out what content to buy and what to pay for it, and they're the best at it. I think so too will Google and others figure out what data is actually particularly useful, what it's worth, and what to pay for it.”
Calacanis: Reddit, Quora, and Stack Overflow will be acquired for AI data
“Reddit, Quora, Stack Overflow, they're gonna just get taken out.”
Calacanis: News industry could have forced Google payouts through a federation
“This is why the news industry should have always had a federation, because they could have just said to Google, hey, we're going to de-index ourselves from the Google search engine, and so you won't have the New York Times, Washington Post, LA Times, you're ju…”
Calacanis: Klarna's AI assistant matches the work of 700 human support agents
“AI assistants are now doing the work of 700 full-time agents at Klarna. They moved issue resolving times from 11 minutes with humans to two minutes with AI, and customer satisfaction is on par with human agents.”
Calacanis: Klarna estimates its AI agent will boost 2024 profits by $40M
“Klarna estimates its AI agent will drive, wait for it, a forty million dollar increase in profits this year.”
Sacks: AI will eliminate tier-one support and eat into tier-two roles
“And I think the AI will do a really good job eliminating level one. It'll start to eating to level two, but you're probably going to need humans to deal with the more complex cases.”
Palihapitiya: Teleperformance lost $1.7B in market cap following Klarna's AI announcement
“Here's Teleperformance, which is a French company that runs call centers. They lost 1.7 billion dollars of market cap when that tweak went up. About 20% of their market cap.”
Palihapitiya: Teleperformance's market cap will fall to $1B if AI spreads
“And if two or three other big companies launch these kinds of tweets after real measurable results, Teleperformance will be a one billion dollar company in short order.”
Sacks: Non-AI companies should open-source internal AI tools for community R&D
“I think it's a brilliant idea for Klarna to open source this tool, because it's not their business, right? This is just something they did as a productivity improvement. They get the benefit back to them of the community working to advance that technology. So …”
Palihapitiya: A handful of open-source AI agents could replace Zendesk's workflow
“The entire Zendesk workflow could be replaced by a handful of these open source agents.”
Palihapitiya: Open-source AI will make point-product software survival very difficult
“And the more that that stuff is open source, my gosh, I think it just makes it very hard for companies that have point products to survive.”
Palihapitiya: One-person startups will hit product-market fit for under $300k
“A one-person company should be able to spend less than a few hundred grand to get to product market fit in the next few years.”
Sacks: Customer support automation will move to human-sounding AI voice call centers
“I think where this is going to go next is to phone. And these call centers use what are called IVRs, these interactive voice response systems, but they're very rigid. It's a lot of prerecorded messages, and it says, push one, if your problem is this, push two,…”
Palihapitiya: Reddit mods and users are the most unqualified retail buyer pool
“A more unqualified retail buyer pool does not exist in the Reddit mods, except maybe the Reddit participants themselves.”
Friedberg: Reddit is worth $2B to $3B in a best-case scenario
“I mean, you could argue it's probably worth in the best, best case in the two to three billion dollar kind of valuation range.”