← pool

🤖 Next year's foundation model commoditizes this year's AI app-layer startups

Sacks' value-capture question about the unicorn-valued AI customer-support startups: if the foundation models keep advancing at this rate, a startup of a few engineers can take next year's model, train it on the same freely available support corpus, and rebuild the whole product - so a thin app-layer wrapper has no durable moat and those valuations do not hold. Chamath's receipt is that 8090 deliberately refused to build in customer service for exactly this reason ('commoditized and run over by these foundational models within a year'), and Jason's version is that it eventually runs locally on a laptop. The value therefore accrues to the model layer and the companies that own it rather than to the application startups chasing the disruption.

0 CONVICTION
peaked 56.3 WATCH
DORMANT
band · contested
▲ LONG MSFT
expression · bullish tech
PARTIAL
outcome · R +19.9% · α +1.7%
2025-09-20
window closed
⚠ CONFLICTING IDEAS ON THE SAME TICKER — the board is arguing with itself; net it before trading (ticker view →)
GOOGL · Blue cities ban or license-cap robotaxis SHORT EMBER 28.2 ← stronger than this idea (5.7)
GOOGL · Tort floodgates open on social media SHORT DORMANT 14.6 ← stronger than this idea (5.7)
NVDA · Nvidia risk: special-purpose silicon + Huawei SHORT DORMANT 10.4 ← stronger than this idea (5.7)
MSFT · ChatGPT takes Google search share and the market starts pricing the decay SHORT DORMANT 9.3 ← stronger than this idea (5.7)
GOOGL · ChatGPT takes Google search share and the market starts pricing the decay SHORT DORMANT 9.3 ← stronger than this idea (5.7)

also touching these tickers, same direction: AI compute capex supercycle has 12-24 months of runway 87 (NVDA) · SaaSpocalypse is overdone for compliance-moat enterprise software 71 (MSFT) · Google wins consumer AI via distribution 61 (GOOGL) · Nvidia + Hugging Face makes NVDA the open-source AI champion 58 (NVDA) · Anthropic IPO (~Oct 2026) is attractively priced at $1.5-2T 54 (GOOGL) · Personal AI agents: Google's data moat wins the form factor 29 (GOOGL) · Frontier AI labs keep durable premium pricing (not commoditized) 27 (MSFT) · Frontier AI labs keep durable premium pricing (not commoditized) 27 (GOOGL) · Own AWS, Azure and GCP and nothing else — the clouds capture the AI application dollars 21 (MSFT) · Own AWS, Azure and GCP and nothing else — the clouds capture the AI application dollars 21 (GOOGL) · Independent frontier-model labs lose to hyperscaler capital 14 (MSFT) · Independent frontier-model labs lose to hyperscaler capital 14 (GOOGL) · Willow's error-correction scaling puts encryption on a two-to-five-year clock 9 (GOOGL) · Meta's AI assistant takes ten points of search share from Google 7 (GOOGL)

⚖ Why this verdict

fully deterministic — evaluate.py replays this from daily closes; nothing below is editable or hand-set

  1. Window: 2024-09-20 → 2025-09-20 — first mention + 12-month horizon, then the window locks.
  2. The call: ▲ LONG MSFT (primary play). MSFT rose 19.9% over the window → direction-adjusted R = +19.9% (the call made money).
  3. Benchmark: holding SPY over the same window returned +18.2%α = +19.9 − (+18.2) = +1.7% — what this call made or lost against just owning the index. This is the number the verdict uses.
  4. Rule fired:
    · HIT — R ≥ +10% AND α ≥ +5
    ▶ PARTIAL — R ≥ +5% OR α ≥ 0
    · MISS — everything else
  5. Credit: supporters of a PARTIAL earn 0.5 each, opposers the inverse — this feeds the scoreboard weights. supported: Chamath, Jason, Sacks | opposed: Gavin Baker | proxy-sensitive: GOOGL→HIT (+56.4%); NVDA→HIT (+52.3%)

Conviction timeline

bands: green ≥ 65 · watch ≥ 45 · ember ≥ 15

Plays vs SPY · % since first mention (2024-09-20)

Plays

expressionsymbolkindrelevancerationale
▲ LONG MSFTstock PRIMARY OpenAI's commercial vehicle and the closest listed proxy for the model layer capturing app-layer value
▲ LONG GOOGLstock adjacent owns a frontier model plus distribution, so it captures the same shift
▲ LONG NVDAstock adjacent value concentrating in ever-better models keeps the training and inference bid intact

Mention log

Sacks
Sacks tech w=0.77 · n=53 · E196 (2024-09-20) · explicit_prediction · strength 2 ▶ 25:26 SUPPORT · horizon 12mo
“if the foundation models are advancing at such a rate, like a year from now, why couldn't a developer, just a startup of a few guys, take next year's model and train it”
Chamath
Chamath tech w=0.90 · n=83 · E196 (2024-09-20) · sentiment · strength 2 ▶ 27:19 SUPPORT
“You'll be able to deploy these. It's just too easy.”
Jason
Jason tech w=1.05 · n=56 · E196 (2024-09-20) · sentiment · strength 1 ▶ 27:22 SUPPORT
“You'll be able to do it on a local computer. I mean, you'll just download the entire database of every call on a Macbook with an M3.”
MA
Mark Cuban (guest ×0.5) no tech track record · E198 (2024-10-03) · explicit_prediction · strength 2 ▶ 1:45:02 SUPPORT
“I think the problem, and this happens with all new technologies, is we're seeing the gold rush right now where everybody calls everything AI, particularly with agents. And I think you can put all these vertical agents together to do all these different things, but agents are just going to be a feature, not a product.”
Gavin Baker
Gavin Baker (regular guest ×1.0) tech w=1.46 · n=13 · E206 (2024-12-07) · explicit_prediction · strength 2 ▶ 1:05:55 OPPOSE
“there are lots of companies that are just these thin wrappers over a foundation model, and they go from zero to 40 million instantaneously”

Who built this conviction

each voice's total force on the score — supports and opposes from every mention, weighted exactly as the replay applied them · share = % of all mention-driven movement

Sacks
Sacks w=0.77
1 scoring event · 36% of moves
+27.1 → net +27.1
Gavin Baker
Gavin Baker w=1.46 regular guest ×1.0
1 scoring event · 23% of moves
+0.0 / -17.5 → net -17.5
Chamath
Chamath w=0.90
1 scoring event · 20% of moves
+14.8 → net +14.8
Jason
Jason w=1.05
1 scoring event · 14% of moves
+11.0 → net +11.0
MA
Mark Cuban guest ×0.5
1 scoring event · 7% of moves
+5.5 → net +5.5

⏳ decay drained -35.2 over the idea's life — that's time passing, attributed to no one

Score events

episodekindΔafternote
E196 2024-09-20 init +27.1 27.1 E196 born by Sacks (explicit_prediction x2) [w=0.77]
E196 2024-09-20 reinforce +14.8 41.9 E196 Chamath support x2 (new voice) [w=0.90]
E196 2024-09-20 reinforce +11.0 52.9 E196 Jason support x1 (new voice) [w=1.05]
E197 2024-09-27 decay -2.1 50.8 E197 silent
E198 2024-10-03 reinforce +5.5 56.3 E198 Mark Cuban support x2 (new voice) [w=0.50]
E199 2024-10-11 decay -2.3 54.1 E199 silent
E200 2024-10-18 decay -2.2 51.9 E200 silent
E201 2024-10-25 decay -2.1 49.8 E201 silent
E202 2024-11-01 decay -2.0 47.9 E202 silent
E203 2024-11-08 decay -1.9 45.9 E203 silent
E204 2024-11-16 decay -1.8 44.1 E204 silent
E205 2024-11-23 decay -1.8 42.3 E205 silent
E206 2024-12-07 oppose -17.5 24.8 E206 Gavin Baker opposes x2 [w=1.46]
E207 2024-12-13 decay -1.0 23.8 E207 silent
E208 2024-12-20 decay -1.0 22.9 E208 silent
E209 2025-01-04 decay -0.9 21.9 E209 silent
E210 2025-01-11 decay -0.9 21.1 E210 silent
E211 2025-01-18 decay -0.8 20.2 E211 silent
E212 2025-01-25 decay -0.8 19.4 E212 silent
E213 2025-01-31 decay -0.8 18.6 E213 silent
E214 2025-02-07 decay -0.7 17.9 E214 silent
E215 2025-02-15 decay -0.7 17.2 E215 silent
E216 2025-02-21 decay -0.7 16.5 E216 silent
E217 2025-03-01 decay -0.7 15.8 E217 silent
E218 2025-03-08 decay -0.6 15.2 E218 silent
E219 2025-03-15 decay -0.6 14.6 E219 silent
E220 2025-03-22 decay -0.6 14.0 E220 silent
E221 2025-03-29 decay -0.6 13.4 E221 silent
E222 2025-04-05 decay -0.5 12.9 E222 silent
E223 2025-04-11 decay -0.5 12.4 E223 silent
E224 2025-04-19 decay -0.5 11.9 E224 silent
E225 2025-04-26 decay -0.5 11.4 E225 silent
E226 2025-05-02 decay -0.5 11.0 E226 silent
E227 2025-05-09 decay -0.4 10.5 E227 silent
E228 2025-05-17 decay -0.4 10.1 E228 silent
E229 2025-05-24 decay -0.4 9.7 E229 silent
E231 2025-06-13 decay -0.4 9.3 E231 silent
E232 2025-06-21 decay -0.4 8.9 E232 silent
E233 2025-06-28 decay -0.4 8.6 E233 silent
E234 2025-07-04 decay -0.3 8.2 E234 silent
E235 2025-07-11 decay -0.3 7.9 E235 silent
E236 2025-07-19 decay -0.3 7.6 E236 silent
E237 2025-08-01 decay -0.3 7.3 E237 silent
E238 2025-08-09 decay -0.3 7.0 E238 silent
E239 2025-08-15 decay -0.3 6.7 E239 silent
E240 2025-08-22 decay -0.3 6.4 E240 silent
E241 2025-08-29 decay -0.3 6.2 E241 silent
E242 2025-09-07 decay -0.2 5.9 E242 silent
E243 2025-09-19 decay -0.2 5.7 E243 silent