E135

E135: Wagner rebels, SCOTUS ends AA, AI M&A, startups gone bad, spacetime warps & more

2023-07-01 spoken.md · speaker-labeled ▶ watch ← E134 all episodes E136 →

ideas born

ideas moved

captures · 4 voices

dissenting

+154.8

conviction added

-118.6

decay · 115 silent

Every number here is replayed from score_events — the same ledger the pool ranks on. Decay is what the 115 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

▲ +14.4 🤖 Open-source models commoditize the model layer and move to the edge watch → green threshold 56.8 → 71.2 still watch — green gate not met

▼ -9.3 🤖 OpenAI is 2023's biggest winner — Microsoft deal inevitable green threshold → watch 66.9 → 57.6

▲ +53.2 🏛️ Affirmative-action ruling propagates into corporate DEI and ESG born at watch 53.2

▲ +23.4 📈 FTX's cap table is a who's-who of no diligence ember → watch 21.9 → 45.3

▲ +39.9 🤖 Data-infra platforms buy the AI tool chain — more M&A at premium multiples born at ember 39.9

Kill dates that landed since E134

1 hit · 1 partial · 3 miss — windows that closed after 2023-06-24 and up to 2023-07-01, auto-scored against price data and never hand-set. verdict · R · α

idea	verdict	R	α	closed
🤖 Bottoms-up SaaS hits a growth wall — only the scaled platforms escape	HIT	+55.2%	+35.8	2023-06-30
🏛️ Crypto collapse triggers a regulatory crackdown that kills the opportunity	MISS	-52.2%	-71.6	2023-06-30
🏦 Own the debt, not the equity, of the tech survivors	PARTIAL	+7.9%	-11.6	2023-06-30
📈 Private equity takes out the busted post-IPO cohort	MISS	-16.2%	-35.6	2023-06-30
📈 Stock-based comp and Evergreen dilution get repriced	MISS	-28.3%	-47.7	2023-06-30

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

Chamath

9 captures · 38% of movement · 1 idea born

+73.2 / -10.8 → net +62.3

Jason

6 captures · 24% of movement

+40.3 / -12.6 → net +27.6

Sacks

8 captures · 21% of movement

+36.4 / -9.3 → net +27.1

Friedberg

1 capture · 17% of movement · 1 idea born

+37.7 → net +37.7

What got argued (13 ideas)

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

NEW ▼ ESGU 🏛️ Affirmative-action ruling propagates into corporate DEI and ESG closed 42 ▲ +53.2 -0.0 → 53.2

Chamath's first- and second-order derivative call on the June 2023 SCOTUS decision: legacy and athletic admissions get sued next, then the race-based hiring and recruiting programs at large private employers get challenged on the same civil-rights grounds, and downstream some ESG checkbox requirements become outright illegal. Changes are slow and then fast, and they land on private enterprise, not just universities.

plays ESGU ·primary DSI ESGV evals 2024-07-01

Chamath support ×3 explicit_prediction ▶ 29:01

And frankly, downstream, how ESG works, because all these ESG check boxes now, some of them will actually become illegal, right?

Jason support ×1 sentiment ▶ 31:19

DEI, to Chamath's point, it is illegal to hire people based on race, gender, any of those criteria, obviously.

NEW ▲ SNOW 🤖 Data-infra platforms buy the AI tool chain — more M&A at premium multiples closed 4 CONTESTED ▲ +39.9 0.0 → 39.9

Off Databricks/MosaicML ($1.3B) and Snowflake/Neeva ($150M): data infrastructure becomes a commodity unless it levels up with model tooling, so the big enterprise infra companies race to assemble an end-to-end AI tool chain (capture, label, store, grab an open-source model, train, infer) by acquisition. Friedberg and Sacks predict more deals at high valuations justified by strategic value rather than revenue; Chamath takes the other side, arguing peak M&A froth comes at the start of a cycle and it is marginally negative for the acquirer's existing shareholders.

plays SNOW ·primary IGV MDB evals 2024-07-01

Friedberg support ×2 explicit_prediction ▶ 55:58

which means that there are more acquisitions still to come.

Sacks support ×3 explicit_prediction ▶ 1:04:45

I think for sure there's going to be more M&A. And I think the valuations will be high, not because these companies have a lot of revenue yet, but because it's very strategic for these big infra companies to assemble the end-to-end tool chain.

Chamath oppose ×2 explicit_prediction ▶ 1:05:47

Peak M&A froth happens at the beginning of the cycle when hype is at maximum and facts are at the minimum.

▼ IPO 📈 FTX's cap table is a who's-who of no diligence closed 25 ▲ +23.4 21.9 → 45.3

Chamath support ×3 sentiment ▶ 1:10:53

they approached us for financing, and when we asked for a data room, we got a Google Doc link to a spreadsheet.

Jason support ×3 sentiment ▶ 1:14:21

The fact that SoftBank did this, that incredible evaluation, and the person who did this deal never checked that the customers were real makes you unfit to serve in the job.

▲ META 🤖 Open-source models commoditize the model layer and move to the edge closed 83 ▲ +14.4 56.8 → 71.2

Jason support ×2 explicit_prediction ▶ 56:58

they're trialing, Friedberg, on average, six, seven, eight language models before they pick one, and they're not picking OpenAI every time.

Sacks support ×3 explicit_prediction ▶ 1:07:53

That's where all the action is right now is customizing these open source models that then leads to basically be able to get the right inferences.

▲ MSFT 🤖 OpenAI is 2023's biggest winner — Microsoft deal inevitable closed 32 CONTESTED ▼ -9.3 66.9 → 57.6

Sacks oppose ×2 sentiment ▶ 1:07:53

But they do not want to share their data with OpenAI. That's very clear. They want to roll out their own models.

▲ MSFT 🤖 AI collapses the cost of knowledge work to ten cents on the dollar closed 28 CONTESTED ▲ +8.5 46.3 → 54.8

Jason support ×2 positioning ▶ 56:58

I'm looking at businesses as an investor, what I'm looking at right now is businesses that were previously not economically viable before this technology, and then that are now economically viable

▲ ITA 🏛️ Nobody's incentives point at de-escalation — the war economy ratchets up closed 50 CONTESTED ▲ +8.3 52.5 → 60.8

Sacks support ×3 explicit_prediction 3mo horizon ▶ 20:00

And I think you could see now over the next few months a full mobilization in Russia.

▲ NVDA 🤖 Moore's law never ended - it moved to GPUs and expert systems go next level closed 55 CONTESTED ▲ +7.5 68.0 → 75.6

Sacks support ×2 explicit_prediction 24mo horizon ▶ 54:21

We have a GPU shortage, and it's probably not going to get better for a year or two, if that.

Chamath support ×2 sentiment ▶ 1:01:48

the list price of an H100 is about 30,000, but the street price, so it's very hard to get it. So if you go to eBay and try to buy an H100, it's like 40 or 45,000.

▼ ARKK 🤖 A decade of zero rates under-trained a generation of operators closed 62 ▲ +6.9 76.6 → 83.5

Chamath support ×3 sentiment ▶ 1:10:53

there aren't enough checks and balances, and there are fundamentally people who are deeply inexperienced, who are in the wrong job.

Jason support ×2 sentiment ▶ 1:14:21

there is a generation of venture capitalists who were added during the boom, who were operators, but they've never been taught to have the discipline of capital allocators.

▲ MSFT 🤖 Foundation models are a big-tech oligopoly - the substrate gets given away closed 30 CONTESTED ▼ -5.6 60.6 → 55.0

Jason oppose ×2 explicit_prediction ▶ 56:58

they're trialing, Friedberg, on average, six, seven, eight language models before they pick one, and they're not picking OpenAI every time.

Chamath support ×2 sentiment ▶ 1:03:18

So whenever I see a chip maker and a cloud provider come together to put in a lot of money, it's essentially round tripping cash.

▲ NVDA 🤖 Generative AI is the next Silicon Valley bubble cycle closed 24 CONTESTED ▲ +4.0 84.5 → 88.5

Sacks support ×2 sentiment ▶ 54:21

Now, I don't think that's necessarily a rational behavior. I think that's more of evidence of a mania going on.

Chamath support ×3 explicit_prediction ▶ 1:10:53

That will play out in AI as well. It's just that we're at the beginning of the hype cycle, right?

▼ IPO 📈 Startup mass extinction event in late 2023 and 2024 closed 35 ▲ +3.6 87.3 → 90.9

Sacks support ×3 explicit_prediction 18mo horizon ▶ 1:23:03

It's only going to get worse over the next 18 months.

Chamath support ×3 explicit_prediction ▶ 1:23:26

I think 60% go to zero.

▲ BX 📈 Private capital concentrates in a few mega-GPs that take themselves public closed 41 CONTESTED ▼ +0.0 0.0 → 0.0

Chamath reversal ×2 explicit_prediction ▶ 1:20:35

And all of a sudden, the profits don't matter, which means the outcomes don't matter, which means the diligence is perfunctory.

Sacks oppose ×3 explicit_prediction ▶ 1:22:20

So Insight was trying to raise a $10 billion fund and they've only been able to raise two, according to this article.

Episode digest

written during extraction and stored in data/extractions/ep135.json — the auditable source of truth, including everything market-adjacent that did not earn a capture

Friedberg guest-moderated (Jason's voice was shot) — labels are clean, the top-talker inversion is innocent. Wagner: Sacks's only forward call is escalation, not settlement (full Russian mobilization within months), which lands on the war-economy ratchet rather than the detente idea, whose window closed a week earlier. Chamath's SCOTUS take is the one with a market channel — the ruling's second derivative hits corporate DEI hiring programs and makes some ESG checkboxes illegal (new idea). The AI M&A block was dense: Sacks and Friedberg predict more data-infra tool-chain acquisitions at strategic-not-revenue multiples (new idea) with Chamath opposing on 'peak M&A froth comes first, valuations go down'; Sacks brings a Craft market map showing enterprises grab open-source models off Hugging Face and fine-tune on their own data because 'they do not want to share their data with OpenAI', and Jason says startups now trial six-to-eight models and don't pick OpenAI every time — both reinforce open-source commoditization and cut against the oligopoly/OpenAI-ascendance side, while Chamath's Inflection math ('round tripping cash', all a pass-through to NVIDIA) argues the incumbents capture it anyway. The venture segment was a bloodbath of receipts: IRL's 95% fake users and Byju's 75% markdown feed diligence-failure and zero-rate-generation, Sacks says zombiecorns get worse for 18 more months and Chamath says 60% go to zero, and the Tiger/Insight fundraise collapse (80-90% smaller funds) plus Chamath's own 'fund sizes should go down, not up' is a reversal on his 2022 mega-GP consolidation call. Science corner (NanoGrav gravitational-wave background) produced nothing tradeable — Jason asked 'why does it matter' and the answer was 'it deepens our understanding' — so nothing captured from it.