E206

New SEC Chair, Bitcoin, xAI Supercomputer, UnitedHealth CEO murder, with Gavin Baker & Joe Lonsdale

2024-12-07 spoken.md · speaker-labeled ▶ watch ← E205 all episodes E207 →

4
ideas born
16
ideas moved
28
captures · 4 voices
6
dissenting
+288.2
conviction added
-118.7
decay · 129 silent

Every number here is replayed from score_events — the same ledger the pool ranks on. Decay is what the 129 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

▲ +21.4 🪙 Post-halving cycle plus ETF flows reprice Bitcoin against gold watch green threshold 57.0 → 78.4 still ember — green gate not met
▲ +27.7 🤖 Software without real lock-in loses its pricing power to AI-built clones ember green threshold 39.7 → 67.4 still watch — green gate not met
▲ +65.9 🤖 Pre-training scaling laws still hold — Colossus and Grok 3 are the proof born at green threshold 65.9 still ember — green gate not met
▲ +15.7 🏛️ Crypto owners become a swing bloc and force a pro-crypto framework ember watch 42.0 → 57.7
▲ +21.4 ⚡ China's nuclear build gives it a cheap-electricity manufacturing edge ember watch 35.8 → 57.2
▲ +56.3 📈 MicroStrategy's convert-funded Bitcoin flywheel breaks at scale born at watch 56.3
▲ +55.0 🏛️ US compute export controls keep China off the AI frontier from 2025 born at watch 55.0
▲ +25.1 ⚡ AI datacenter demand needs double-to-triple current US electricity generation ember watch 24.8 → 49.9
▲ +22.9 📈 Meta's profit engine plus buybacks makes its AI option value free ember watch 23.8 → 46.6
▲ +23.4 🏛️ Counter-drone defense becomes the new procurement line dormant ember 13.3 → 36.7
▲ +22.5 🏛️ Atkins SEC loosens accredited-investor gating and retail money reaches private markets born at ember 22.5
▼ -34.0 🤖 AI capex outruns app-layer revenue - Nvidia's terminal-value problem watch ember 52.3 → 18.4
▼ -21.0 🤖 OpenAI keeps the consumer AI lead through a developer network effect ember dormant 23.9 → 2.9

Kill dates that landed since E205

0 hit · 0 partial · 1 miss — windows that closed after 2024-11-23 and up to 2024-12-07, auto-scored against price data and never hand-set. verdict · R · α

ideaverdictRαclosed
📈 Record 2024 Atlantic hurricane season drives catastrophe losses MISS -20.6% -35.1 2024-12-07

Who moved the board

each voice's force on conviction this episodesupports and opposes, weighted exactly as the replay applied them · share = % of this episode's movement

Gavin Baker
Gavin Baker regular guest ×1.0
12 captures · 61% of movement · 3 ideas born
+220.6 / -67.6 → net +153.0
JO
Joe Lonsdale regular guest ×1.0
7 captures · 19% of movement · 1 idea born
+88.2 → net +88.2
Friedberg
Friedberg
8 captures · 16% of movement
+49.6 / -25.0 → net +24.6
Jason
Jason
1 capture · 5% of movement
+22.5 → net +22.5

What got argued (16 ideas)

ordered by how hard each idea moved · quotes are verbatim from the transcript, timestamps deep-link into the episode

NEW NVDA 🤖 Pre-training scaling laws still hold — Colossus and Grok 3 are the proof closed 34 CONTESTED ▲ +65.9 0.0 → 65.9

Gavin Baker argues the 'we're hitting a wall' debate had no data because nobody had built a cluster past ~32,000 H100s. Elon's Memphis Colossus made 100,000+ hoppers coherent (going to 200k, then a million), so Grok 3 is the first real test of training scaling laws since GPT-4 and he expects it to be a significant advance in the state of the art. Friedberg adds new scaling axes — networks of models, test-time compute, context window — so capability keeps buying out of compute and the compute demand curve keeps compounding.

plays NVDA ·primary AMD AVGO evals 2025-12-07
Gavin Baker
Gavin Baker support ×3 explicit_prediction guest ×1.0 ▶ 54:45
But you have not had a real test of scaling laws for training, arguably since GPT-4. And this will be the first test. And if scaling laws for training hold, Grok 3 should be a significant advance in the state of the art
Friedberg
Friedberg support ×2 sentiment ▶ 58:46
I think it's a new kind of scaling access arguably in terms of the potential set of applications. So networks of models, think time, context window. There are multiple dimensions upon which these tools ultimately kind of resolve to better performance
NEW MSTR 📈 MicroStrategy's convert-funded Bitcoin flywheel breaks at scale closed 10 CONTESTED ▲ +56.3 0.0 → 56.3

Gavin Baker's read on Saylor's convertible-note-funded Bitcoin accumulation: the debt stack gets too big relative to a ~$400M-revenue underlying business that has to service the interest, and fixed-income markets do not yet accept Bitcoin as collateral. Past a point he has to over-collateralize, and the 'magic money creation machine' being celebrated on X stops working.

plays MSTR ·primary COIN evals 2025-12-07
Gavin Baker
Gavin Baker support ×2 explicit_prediction guest ×1.0 ▶ 42:07
it will get to a point where it is, it is too big for the size of his company. And then yeah, maybe he can over collateralize it and, you know, have $10 in Bitcoin for every dollar of debt. But then like the magic money creation machine that, you know, I see discussed on X, breaks down
JO
Joe Lonsdale support ×1 sentiment guest ×1.0 ▶ 43:18
And I think the risks around how he's accessing this is actually very unusual and does scare me a bit
NEW NVDA 🏛️ US compute export controls keep China off the AI frontier from 2025 closed 12 CONTESTED ▲ +55.0 0.0 → 55.0

Gavin Baker: restricting China's access to advanced compute and networking is 'unfolding a sofa over China'. Chinese models have impressively stayed just behind the leading edge so far, but once Blackwell, new AMD parts and new Broadcom ASICs ship in 2025 the gap becomes uncloseable — an aggressive foreign policy that works, with unforeseeable consequences.

plays NVDA ·primary AVGO KWEB evals 2025-12-07
Gavin Baker
Gavin Baker support ×3 explicit_prediction guest ×1.0 12mo horizon ▶ 47:29
But you know, NVIDIA's Blackwell chip comes out next year. You're going to have new chips from AMD, new ASICs from Broadcom. And I think at that point, it is not going to be possible for them to keep up anymore
NVDA 🤖 AI capex outruns app-layer revenue - Nvidia's terminal-value problem closed 0 CONTESTED ▼ -34.0 52.3 → 18.4
Gavin Baker
Gavin Baker oppose ×3 explicit_prediction guest ×1.0 ▶ 1:04:33
And ROI has gone vertical since they ramped their capex on GPUs. It actually just started to level out in this latest quarter. So the ROI on AI has been very positive thus far. Just a fact
Friedberg
Friedberg oppose ×2 sentiment ▶ 1:19:27
And I think it really showcases the kind of the impressive impact that AI is having on workplace productivity
IGV 🤖 Software without real lock-in loses its pricing power to AI-built clones closed 60 CONTESTED ▲ +27.7 39.7 → 67.4
Friedberg
Friedberg support ×3 explicit_prediction ▶ 1:19:27
You don't need to buy SaaS tools. You don't need to have service providers do stuff for you
Gavin Baker
Gavin Baker support ×2 explicit_prediction guest ×1.0 12mo horizon ▶ 1:21:35
But I think next year, the human language will be the dominant programming language
XLU ⚡ AI datacenter demand needs double-to-triple current US electricity generation closed 81 ▲ +25.1 24.8 → 49.9
Friedberg
Friedberg support ×3 explicit_prediction ▶ 14:37
But for me, it has to be such a priority that we accelerate a nuclear energy rollout in the United States, because that's what China is doing
Gavin Baker
Gavin Baker support ×2 sentiment guest ×1.0 ▶ 16:13
I mean, more nuclear, more better
KTOS 🏛️ Counter-drone defense becomes the new procurement line closed 28 ▲ +23.4 13.3 → 36.7
JO
Joe Lonsdale support ×3 explicit_prediction guest ×1.0 ▶ 45:50
But on the margin, I'd much rather have ten thousand more smart drones above and below the water than like an incremental carrier
META 📈 Meta's profit engine plus buybacks makes its AI option value free closed 34 ▲ +22.9 23.8 → 46.6
Gavin Baker
Gavin Baker support ×2 sentiment guest ×1.0 ▶ 1:06:34
So outside of the ROI on AI that you're seeing in Google and Meta from, you know, using this across their businesses, you're seeing real ROI on AI from startups the same way they saw real ROI from cloud computing before anyone else
NEW BX 🏛️ Atkins SEC loosens accredited-investor gating and retail money reaches private markets closed 15 CONTESTED ▲ +22.5 0.0 → 22.5

Off a 2007 Paul Atkins speech attacking the accredited-investor premise, Jason and Joe Lonsdale argue the incoming SEC should let ordinary Americans into private funds via a sophistication test rather than a wealth test, and Gavin notes the big private-equity firms are already making a huge effort to reach Main Street. The read-through is a retail distribution channel for alternative managers. Friedberg takes the direct other side — private markets with weaker disclosure would be used to market junk securities to poor people — and Gavin agrees disclosure standards would have to be strengthened first.

plays BX ·primary APO KKR evals 2025-12-07
JO
Joe Lonsdale support ×2 sentiment guest ×1.0 ▶ 36:55
Imagine if Elon could raise for SpaceX from regular Americans. I'm sure he would have loved to do that if it wasn't crazy risky with the SEC
Friedberg
Friedberg oppose ×3 explicit_prediction ▶ 37:23
In fact, I think it's more likely than not that people are going to go market bullsh** securities and private markets and rip poor people off even worse
Gavin Baker
Gavin Baker oppose ×2 sentiment guest ×1.0 ▶ 38:17
And I do think if you were to allow ordinary Americans to buy private companies that are held to a lower standard of disclosure and reporting to public companies, like something would have to change
Jason
Jason support ×2 sentiment ▶ 39:08
I mean, there's such an easy solution to this. You just do a sophistication test. People take a five-hour course and they answer 50 questions. The end, like a driver's license
URA ⚡ China's nuclear build gives it a cheap-electricity manufacturing edge closed 52 CONTESTED ▲ +21.4 35.8 → 57.2
Friedberg
Friedberg support ×3 explicit_prediction ▶ 11:03
which is the increment in electricity production capacity in China compared to the United States. And so we are going from one terawatt to two terawatts. They're going from, I think, two to eight over the same time period, and they're doing it at a price
JO
Joe Lonsdale support ×2 sentiment guest ×1.0 ▶ 15:39
This is a key national security thing, too, because we need manufacturing here to be affordable and competitive in order for national security
cg:bitcoin 🪙 Post-halving cycle plus ETF flows reprice Bitcoin against gold closed 37 ▲ +21.4 57.0 → 78.4
Friedberg
Friedberg support ×2 explicit_prediction ▶ 29:40
It should be kind of store of value and alternative. It's just going to take over gold
Gavin Baker
Gavin Baker support ×2 explicit_prediction guest ×1.0 ▶ 31:31
But I do think Bitcoin will at some point be a serious threat to the US dollar. And that just is what it is
JO
Joe Lonsdale support ×2 sentiment guest ×1.0 ▶ 43:18
Listen, I am very bullish Bitcoin. I love all the energy of our society around it
MSFT 🤖 OpenAI keeps the consumer AI lead through a developer network effect closed 31 CONTESTED ▼ -21.0 23.9 → 2.9
Gavin Baker
Gavin Baker oppose ×3 explicit_prediction guest ×1.0 3mo horizon ▶ 1:00:43
Grok 3 should take the lead if scaling laws hold in January or February. I do think a lot of talent has left OpenAI
MSFT 🤖 Next year's foundation model commoditizes this year's AI app-layer startups closed 6 CONTESTED ▼ -17.5 42.3 → 24.8
Gavin Baker
Gavin Baker oppose ×2 explicit_prediction guest ×1.0 ▶ 1:05:55
there are lots of companies that are just these thin wrappers over a foundation model, and they go from zero to 40 million instantaneously
COIN 🏛️ Crypto owners become a swing bloc and force a pro-crypto framework closed 40 CONTESTED ▲ +15.7 42.0 → 57.7
JO
Joe Lonsdale support ×3 explicit_prediction guest ×1.0 ▶ 20:58
And the reason they're that angry at them is they were purposely not defining the rules in certain cases and then going after people after not having to find it in order to kind of like play a gotcha game. It was very dishonorable. And Paul sees all that and there's no way he's going to allow that
IWM 🏛️ A second Trump term reflates growth via deregulation and lower corporate tax closed 71 ▲ +3.2 94.1 → 97.3
Gavin Baker
Gavin Baker support ×3 explicit_prediction guest ×1.0 ▶ 4:43
So I just think deregulation and simplifying regulations and the tax code is going to lead to an immense amount of growth
JO
Joe Lonsdale support ×2 sentiment guest ×1.0 ▶ 7:28
I think almost all reasonable Americans can see that our growth is ridiculously constrained
TLT 🏦 Phase three of the 2023 crisis is a US government debt crisis closed 87 ▲ +0.5 96.1 → 96.5
Friedberg
Friedberg support ×2 explicit_prediction ▶ 11:03
And that leads to excessive spending, and that leads to the debt problem. And the debt problem creates this kind of arithmetic debt, debt spiral, which is something you don't want to find yourself in

Episode digest

written during extraction and stored in data/extractions/ep206.json — the auditable source of truth, including everything market-adjacent that did not earn a capture

Sacks and Chamath both off; Gavin Baker (Atreides) and Joe Lonsdale (8VC) sub in, with Jason recording from Sacks' house. Post-election framing is uniformly pro-growth: Gavin's Satya-at-Microsoft analogy for America (stop doing dumb things and you win) plus an explicit call that deregulation and tax simplification produce an immense amount of growth, and Friedberg re-runs his debt-spiral thread with China's electricity build (1->2 terawatts for the US against 2->8 for China) as the metric that forces a US nuclear acceleration. Gensler out / Atkins in is read as the end of enforcement-by-gotcha, and all three guests are bitcoin-constructive (Friedberg: it takes over gold and is the real threat to the dollar), but Gavin is the bear on Saylor's convertible-note flywheel — the underlying business only does ~$400M of revenue, fixed income does not accept BTC as collateral, and the 'magic money creation machine' breaks once the stack is too big. The meat is the AI segment: Gavin (an xAI investor, disclosed at 55:23) argues the scaling-law 'wall' debate had no data because nobody had built past ~32k H100s, Colossus's 100k coherent hoppers is the first real test since GPT-4, Grok 3 should take the frontier lead in January or February, and the AI-ROI debate is already settled by public-company ROIC going vertical since GPU capex ramped. He also expects Blackwell/AMD/Broadcom in 2025 to make the compute gap uncloseable for China. A 2007 Atkins speech on accredited-investor rules sets off a real four-way fight over retail access to private markets (Jason and Joe for, Friedberg hard against, Gavin siding with Friedberg on disclosure). The UnitedHealthcare CEO killing segment produced no tradeable directional view — Friedberg gave insurer MLR context (~85%) and both guests focused on the online celebration rather than on the stock or on regulation.