E122: Is AI the next great computing platform? ChatGPT vs. Google, containing AGI & RESTRICT Act
2023-03-31 spoken.md · speaker-labeled ▶ watch ← E121 all episodes E123 →
Every number here is replayed from score_events — the same ledger the pool ranks on. Decay is what the 117 ideas nobody mentioned gave up this week; it applies only when an episode is processed.
Tier crossings
conviction thresholds crossed by this episode — 65 / 45 / 15 · ideas born here show where they landed. why 65 isn't always green
Kill dates that landed since E121
2 hit · 0 partial · 2 miss — windows that closed after 2023-03-24 and up to 2023-03-31, auto-scored against price data and never hand-set. verdict · R · α
| idea | verdict | R | α | closed |
|---|---|---|---|---|
| 📈 Late-stage private marks reset 30-60% — down-round IPOs become the norm | HIT | +37.9% | +49.7 | 2023-03-26 |
| 📈 Quality software grows through the de-rating — Snowflake | MISS | -42.1% | -30.4 | 2023-03-26 |
| 📈 Rate-arbitrage fintechs implode as their cost of capital rises | HIT | +38.0% | +49.7 | 2023-03-26 |
| 🤖 The ad duopoly ratchets prices and makes its numbers through the startup bust | MISS | -25.5% | -13.7 | 2023-03-26 |
Who moved the board
each voice's force on conviction this episode — supports and opposes, weighted exactly as the replay applied them · share = % of this episode's movement
What got argued (11 ideas)
ordered by how hard each idea moved · quotes are verbatim from the transcript, timestamps deep-link into the episode
Chamath's named call on where AI labor displacement shows up in the numbers before anywhere else: the code-for-hire scale shops - TCS, Accenture (~750k employees), Cognizant - have the most immediate and most shareholder-rewarded incentive to use LLM coding tools to squeeze opex, lift utilization rates and expand profitability, so they are the first firms to displace human labor at scale, ahead of law firms, accounting firms or any other white-collar vertical. AI eliminates unit testing and most routine coding, so the same or fewer engineers deliver more; it will be 'less visible in other companies'.
But what I would tell you, Jason, to answer the white collar question is I think there are a handful of companies you need to look at exclusively because they will be the first ones to really figure out how to displace human labor. And that is TCS, so Tata Consulting Services, Accenture, Cognizant.
And then when they make BARD a default, you know, a little snippet on your Google search return page or BARD is built into YouTube or Chrome or Android or the Play Store, they're going to roll right over ChatGPT because they have billions of users already.
Sacks' call the week OpenAI shipped ChatGPT plugins: ChatGPT is not just an intelligence API in the mold of Stripe, it is a destination site and the most important developer platform since the iPhone and iOS - arguably the biggest ever. Plugins let it intermediate apps, browsers, search and commerce, and the combination of hundreds of millions of users, the browsing plugin and the retrieval API (developers voluntarily handing over proprietary knowledge bases) compounds a lead that grows over the next year, until applications are first incentivized and then competitively forced to participate. The rest of the table takes the other side: Chamath sees no natural blocker and would stay patient, Friedberg doubts the plugin network effect is durable when a 700GB model can be forked.
So the point is, I think we've talked about this for a while, but all of these models will converge in the absence of highly unique data, right? What I've been calling these white truffles. So if you can hoard white truffles, your model will be better. Otherwise, your model will be the same as everybody else's model.
Episode digest
written during extraction and stored in data/extractions/ep122.json — the auditable source of truth, including everything market-adjacent that did not earn a capture
A dedicated AI episode built around OpenAI shipping ChatGPT plugins. Sacks table-pounds the biggest new call: plugins turn ChatGPT into a destination site and the most important developer platform since iOS ('maybe ever'), OpenAI's lead is substantial and grows over the next year, and it eats a chunk of search, browser and app usage - the other three all push back, Chamath on 'no natural blockers, stay patient, we are six weeks in', Friedberg on a forkable 700GB model with no durable network effect, Jason on Google bundling BARD into YouTube/Chrome/Android and rolling right over it. The Google pair got worked hard from both sides: Friedberg reinforced his own chat-disrupts-search thesis (plugins close the search-to-commerce loop Google can't, and explicitly walked back the 'search UI is defensible' half of his E115 inference-cost argument), while Chamath partially swung back toward his own bullish-Google thesis he had reversed at E115 - arguing incumbent distribution and $5-6T of market cap incentives suppress ChatGPT, though he refused to call a winner and warned Google is 'toast' if middle management doesn't move. Chamath's proprietary-data moat ('white truffles') got its strongest restatement yet, with Sacks' give-to-get blog post and Jason's Reddit/Stack Overflow/Quora M&A call on the same side; Jason's data-islands and Sacks' plugins-are-promiscuous argument was the episode's other live disagreement. New idea coined for Chamath's only concretely tradeable labor call: TCS, Accenture and Cognizant displace white-collar labor first because shareholders reward the opex squeeze. The 25-minute jobs debate mostly re-litigated Friedberg's creative-production thesis (support from Sacks and Jason, a strength-3 oppose from Chamath on AI supplanting human judgment in closed-loop systems - he disclosed one in front of the FDA now). Diarization CLEAN (Jason top talker at 89, all four present, addressed-by-name and fingerprint tests pass, no time gaps); one local bleed at 1:04:04 where a Sacks-labelled turn opens 'Now, wait a second, Sacks' - not used in any quote. The FLI pause-letter segment was almost entirely untradeable safety philosophy and the RESTRICT Act segment was civil-liberties analysis with no market channel, so only two low-strength TikTok-crackdown mentions came out of it.