← pool · child of DeepSeek proves frontier-model costs are collapsing

🤖 DeepSeek R1 was distilled from OpenAI, not a clean $6M cost breakthrough

R1's headline efficiency is inflated by two things DeepSeek does not disclose: the $6M figure covers only the final training run and is being compared to US labs' fully-loaded numbers, while the same founder's hedge fund already owned a ~50,000-Hopper cluster worth over a billion dollars; and R1's ~800,000 reasoning samples came at least partly from distilling OpenAI's o1, which V3 self-identifying as ChatGPT already implied. The efficiency gain is real but far smaller than advertised, so the US frontier lead and the compute bill survive and the NVDA drawdown was an overreaction.

0 CONVICTION
peaked 50.9 WATCH
DORMANT
band · contested
▲ LONG NVDA
expression · bullish tech
HIT
outcome · R +59.2% · α +42.9%
2026-01-31
window closed
⚠ CONFLICTING IDEAS ON THE SAME TICKER — the board is arguing with itself; net it before trading (ticker view →)
NVDA · Nvidia risk: special-purpose silicon + Huawei SHORT DORMANT 10.4 ← stronger than this idea (8.4)
MSFT · ChatGPT takes Google search share and the market starts pricing the decay SHORT DORMANT 9.3 ← stronger than this idea (8.4)

also touching these tickers, same direction: AI compute capex supercycle has 12-24 months of runway 87 (NVDA) · AI compute capex supercycle has 12-24 months of runway 87 (SMH) · SaaSpocalypse is overdone for compliance-moat enterprise software 71 (MSFT) · Nvidia + Hugging Face makes NVDA the open-source AI champion 58 (NVDA) · Nvidia + Hugging Face makes NVDA the open-source AI champion 58 (SMH) · DRAM/HBM: the bottleneck that matters 45 (SMH) · Frontier AI labs keep durable premium pricing (not commoditized) 27 (MSFT) · Own AWS, Azure and GCP and nothing else — the clouds capture the AI application dollars 21 (MSFT) · Independent frontier-model labs lose to hyperscaler capital 14 (MSFT)

⚖ Why this verdict

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

  1. Window: 2025-01-31 → 2026-01-31 — first mention + 12-month horizon, then the window locks.
  2. The call: ▲ LONG NVDA (primary play). NVDA rose 59.2% over the window → direction-adjusted R = +59.2% (the call made money).
  3. Benchmark: holding SPY over the same window returned +16.3%α = +59.2 − (+16.3) = +42.9% — 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 ✓ (R 59.2, α 42.9)
    · PARTIAL — R ≥ +5% OR α ≥ 0
    · MISS — everything else
  5. Credit: supporters of a HIT earn 1.0 each, opposers the inverse — this feeds the scoreboard weights. supported: Chamath, Sacks | opposed: Travis Kalanick | proxy-sensitive: MSFT→MISS (+4.4%)

Conviction timeline

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

Plays vs SPY · % since first mention (2025-01-31)

Plays

expressionsymbolkindrelevancerationale
▲ LONG NVDAstock PRIMARY the stock that lost $600B on the cheap-training narrative; recovers if the cost story is overstated
▲ LONG MSFTstock adjacent OpenAI's compute and distribution partner - frontier-lab economics survive with it
▲ LONG SMHetf adjacent broad semis basket that sold off with the panic (TSMC, Broadcom, Arm)

Mention log

Sacks
Sacks tech w=0.77 · n=53 · E213 (2025-01-31) · explicit_prediction · strength 3 ▶ 34:34 SUPPORT POLICY-ADJACENT
“And every single person I've talked to basically has agreed that there was some distillation here from OpenAI. Now, that doesn't mean it was the only thing going on here. I mean, to be sure, the DeepSeek team is very smart and there were some innovations, but also there was some distillation.”
Chamath
Chamath tech w=0.90 · n=83 · E213 (2025-01-31) · explicit_prediction · strength 3 ▶ 31:12 SUPPORT
“So the reason why this is happening is like you're seeing this chain of thought, you're seeing the several layers, and then it's catching it after the fact. So we know that this is distilled from some other model.”
TR
Travis Kalanick (regular guest ×1.0) tech w=1.33 · n=6 · E213 (2025-01-31) · explicit_prediction · strength 2 ▶ 37:53 OPPOSE
“You go through the white paper, you see what it is they did, what they innovated on, the science behind it, the thoroughness, and you're like, these guys are badass. It does not feel or sound like somebody who took something, just when you get through it.”
Chamath
Chamath tech w=0.90 · n=83 · E224 (2025-04-19) · explicit_prediction · strength 2 ▶ 14:20 SUPPORT
“At every turn and at every step of AI, they are at the same rate or one step ahead. And I suspect it's because that these ships are being used in very sophisticated ways behind the scenes.”

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

Chamath
Chamath w=0.90
2 scoring events · 63% of moves
+48.4 → net +48.4
TR
Travis Kalanick w=1.33 regular guest ×1.0
1 scoring event · 21% of moves
+0.0 / -16.0 → net -16.0
Sacks
Sacks w=0.77
1 scoring event · 17% of moves
+13.0 → net +13.0

⏳ decay drained -36.9 over the idea's life — that's time passing, attributed to no one

Score events

episodekindΔafternote
E213 2025-01-31 init +38.0 38.0 E213 born by Chamath (explicit_prediction x3) [w=0.90]
E213 2025-01-31 reinforce +13.0 50.9 E213 Sacks support x3 (new voice) [w=0.77]
E213 2025-01-31 oppose -16.0 34.9 E213 Travis Kalanick opposes x2 [w=1.33]
E214 2025-02-07 decay -1.4 33.5 E214 silent
E215 2025-02-15 decay -1.3 32.2 E215 silent
E216 2025-02-21 decay -1.3 30.9 E216 silent
E217 2025-03-01 decay -1.2 29.7 E217 silent
E218 2025-03-08 decay -1.2 28.5 E218 silent
E219 2025-03-15 decay -1.1 27.3 E219 silent
E220 2025-03-22 decay -1.1 26.2 E220 silent
E221 2025-03-29 decay -1.0 25.2 E221 silent
E222 2025-04-05 decay -1.0 24.2 E222 silent
E223 2025-04-11 decay -1.0 23.2 E223 silent
E224 2025-04-19 reinforce +10.4 33.6 E224 Chamath support x2 [w=0.90]
E225 2025-04-26 decay -1.3 32.3 E225 silent
E226 2025-05-02 decay -1.3 31.0 E226 silent
E227 2025-05-09 decay -1.2 29.7 E227 silent
E228 2025-05-17 decay -1.2 28.6 E228 silent
E229 2025-05-24 decay -1.1 27.4 E229 silent
E231 2025-06-13 decay -1.1 26.3 E231 silent
E232 2025-06-21 decay -1.1 25.3 E232 silent
E233 2025-06-28 decay -1.0 24.3 E233 silent
E234 2025-07-04 decay -1.0 23.3 E234 silent
E235 2025-07-11 decay -0.9 22.4 E235 silent
E236 2025-07-19 decay -0.9 21.5 E236 silent
E237 2025-08-01 decay -0.9 20.6 E237 silent
E238 2025-08-09 decay -0.8 19.8 E238 silent
E239 2025-08-15 decay -0.8 19.0 E239 silent
E240 2025-08-22 decay -0.8 18.2 E240 silent
E241 2025-08-29 decay -0.7 17.5 E241 silent
E242 2025-09-07 decay -0.7 16.8 E242 silent
E243 2025-09-19 decay -0.7 16.1 E243 silent
E244 2025-09-27 decay -0.6 15.5 E244 silent
E245 2025-10-03 decay -0.6 14.9 E245 silent
E246 2025-10-10 decay -0.6 14.3 E246 silent
E247 2025-10-17 decay -0.6 13.7 E247 silent
E248 2025-10-24 decay -0.5 13.1 E248 silent
E249 2025-10-31 decay -0.5 12.6 E249 silent
E250 2025-11-07 decay -0.5 12.1 E250 silent
E251 2025-11-14 decay -0.5 11.6 E251 silent
E252 2025-11-22 decay -0.5 11.2 E252 silent
E253 2025-12-06 decay -0.4 10.7 E253 silent
E254 2025-12-13 decay -0.4 10.3 E254 silent
E255 2025-12-19 decay -0.4 9.9 E255 silent
E256 2025-12-31 decay -0.4 9.5 E256 silent
E257 2026-01-10 decay -0.4 9.1 E257 silent
E258 2026-01-17 decay -0.4 8.7 E258 silent
E259 2026-01-30 decay -0.3 8.4 E259 silent