← pool

🤖 Gemini's ideological tuning breaks Google's AI product

Gemini refuses or falsifies answers on undisputed facts (black George Washington, 'diverse' founding fathers, no answer on IQ data or on Trump's or Biden's legal cases) because a diversity-and-inclusion tuning layer overrides the model, and Sacks' read is that this is not a rushed-launch accident but an accurate self-portrait of Google's monoculture - so the fix will be to make the bias subtler rather than remove it. The tradeable consequence is product quality rather than politics: an information-interpretation engine that will not answer straight loses consumer trust, users defect to whatever tells them the truth, and Google's culture makes it structurally slow to respond. Bearish Google's consumer AI and search franchise. Distinct from the 2023 trust-and-safety censorship debate, whose window closed six days before this episode aired.

0 CONVICTION
peaked 64.5 WATCH
DORMANT
band · contested
▼ SHORT GOOGL
expression · bearish tech
MISS
outcome · R -25.2% · α -44.9%
2025-02-23
window closed
⚠ CONFLICTING IDEAS ON THE SAME TICKER — the board is arguing with itself; net it before trading (ticker view →)
GOOGL · Google wins consumer AI via distribution LONG WATCH 61.2 ← stronger than this idea (0.0)
GOOGL · Anthropic IPO (~Oct 2026) is attractively priced at $1.5-2T LONG WATCH 53.9 ← stronger than this idea (0.0)
GOOGL · Personal AI agents: Google's data moat wins the form factor LONG EMBER 29.3 ← stronger than this idea (0.0)
GOOGL · Frontier AI labs keep durable premium pricing (not commoditized) LONG EMBER 26.6 ← stronger than this idea (0.0)
GOOGL · Own AWS, Azure and GCP and nothing else — the clouds capture the AI application dollars LONG EMBER 20.6 ← stronger than this idea (0.0)
GOOGL · Independent frontier-model labs lose to hyperscaler capital LONG DORMANT 14.0 ← stronger than this idea (0.0)
GOOGL · Willow's error-correction scaling puts encryption on a two-to-five-year clock LONG DORMANT 8.8 ← stronger than this idea (0.0)
GOOGL · Meta's AI assistant takes ten points of search share from Google LONG DORMANT 7.4 ← stronger than this idea (0.0)

also touching these tickers, same direction: Blue cities ban or license-cap robotaxis 28 (GOOGL) · Tort floodgates open on social media 15 (GOOGL) · ChatGPT takes Google search share and the market starts pricing the decay 9 (GOOGL)

⚖ Why this verdict

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

  1. Window: 2024-02-23 → 2025-02-23 — first mention + 12-month horizon, then the window locks.
  2. The call: ▼ SHORT GOOGL (primary play). GOOGL rose 25.2% over the window → direction-adjusted R = -25.2% (the call lost money).
  3. Benchmark: holding SPY over the same window returned +19.6%α = -25.2 − (+19.6) = -44.9% — what this call made or lost against just owning the index. This is the number the verdict uses.
    Stock-selection read: -5.6% — did GOOGL move the predicted way relative to the market. The two only differ on a SHORT: this call was right about direction relative to the index but still cost money versus holding it.
  4. Rule fired:
    · HIT — R ≥ +10% AND α ≥ +5
    · PARTIAL — R ≥ +5% OR α ≥ 0
    ▶ MISS — everything else
  5. Credit: supporters of a MISS earn 0 each, opposers the inverse — this feeds the scoreboard weights. supported: Sacks | opposed: Chamath, Friedberg, Jason

Conviction timeline

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

Plays vs SPY · % since first mention (2024-02-23)

Plays

expressionsymbolkindrelevancerationale
▼ SHORT GOOGLstock PRIMARY the damaged product and franchise is Google's

Mention log

Sacks
Sacks tech w=0.77 · n=53 · E167 (2024-02-23) · explicit_prediction · strength 3 ▶ 54:58 SUPPORT
“This idea that Gemini turned out this way by accident or because they rushed it out, I don't really believe that. I believe that what happened is Gemini accurately reflects the biases of the people who created it ... I think they're simply going to dial down the bias to be less obvious.”
Chamath
Chamath tech w=0.90 · n=83 · E167 (2024-02-23) · explicit_prediction · strength 2 ▶ 59:07 SUPPORT
“Too many of these judgments, I think, will make this product very poor quality and consumers will just go to the thing that tells it the truth.”
Friedberg
Friedberg tech w=1.08 · n=65 · E167 (2024-02-23) · sentiment · strength 2 ▶ 1:14:19 SUPPORT
“Google could be going down the wrong path here in a way that they will lose users and lose consumers and someone else will be there eagerly to sweep up with a better product.”
Friedberg
Friedberg tech w=1.08 · n=65 · E175 (2024-04-19) · sentiment · strength 2 ▶ 16:35 SUPPORT
“there may have been this kind of entitlement culture where anytime Google employees ask for stuff, they get it... That's so much of the orientation of being an employee at Google is all about what Google can do for me and how I can get more.”
Jason
Jason tech w=1.05 · n=56 · E175 (2024-04-19) · sentiment · strength 2 ▶ 21:38 SUPPORT
“which we talked about with the Gemini issues and this stuff bleeding over into product... are people actually focused on products at Google anymore?”
Chamath
Chamath tech w=0.90 · n=83 · E179 (2024-05-17) · explicit_prediction · strength 2 ▶ 1:24:57 OPPOSE
“So I think that the reports of their death were premature and exaggerated.”
Sacks
Sacks tech w=0.77 · n=53 · E179 (2024-05-17) · explicit_prediction · strength 2 ▶ 1:34:34 REVERSAL
“Black George Washington. Doesn't matter. They can make 10 mistakes, but their monopoly is so strong that they can finally get it right by copying the innovator.”
Sacks
Sacks tech w=0.77 · n=53 · E202 (2024-11-01) · explicit_prediction · strength 3 ▶ 55:49 SUPPORT
“I think you've got to, at this point, have your head buried really deep in the sand, not to think that Google is incredibly biased in this election against Trump.”
Jason
Jason tech w=1.05 · n=56 · E202 (2024-11-01) · explicit_prediction · strength 3 ▶ 1:03:59 OPPOSE
“Yeah, it's not the search engine. It's just the corpus of media that's being indexed.”
Friedberg
Friedberg tech w=1.08 · n=65 · E202 (2024-11-01) · sentiment · strength 1 ▶ 59:51 OPPOSE
“it may not necessarily have been kind of overt action by Google, but just the way that the system is set up, that anyone can flag and if enough people flag, you get the sort of trigger.”

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
3 scoring events · 48% of moves
+45.4 / -27.7 → net +17.6
Friedberg
Friedberg w=1.08
3 scoring events · 21% of moves
+22.0 / -10.3 → net +11.7
Chamath
Chamath w=0.90
2 scoring events · 16% of moves
+13.7 / -10.8 → net +2.9
Jason
Jason w=1.05
2 scoring events · 14% of moves
+11.0 / -10.1 → net +1.0

⏳ decay drained -33.1 over the idea's life — that's time passing, attributed to no one

Score events

episodekindΔafternote
E167 2024-02-23 init +32.5 32.5 E167 born by Sacks (explicit_prediction x3) [w=0.77]
E167 2024-02-23 reinforce +13.7 46.2 E167 Chamath support x2 (new voice) [w=0.90]
E167 2024-02-23 reinforce +13.0 59.2 E167 Friedberg support x2 (new voice) [w=1.08]
E168 2024-03-01 decay -2.4 56.9 E168 silent
E169 2024-03-08 decay -2.3 54.6 E169 silent
E170 2024-03-15 decay -2.2 52.4 E170 silent
E171 2024-03-22 decay -2.1 50.3 E171 silent
E172 2024-03-29 decay -2.0 48.3 E172 silent
E173 2024-04-05 decay -1.9 46.4 E173 silent
E174 2024-04-12 decay -1.9 44.5 E174 silent
E175 2024-04-19 reinforce +9.0 53.5 E175 Friedberg support x2 [w=1.08]
E175 2024-04-19 reinforce +11.0 64.5 E175 Jason support x2 (new voice) [w=1.05]
E176 2024-04-26 decay -2.6 61.9 E176 silent
E177 2024-05-03 decay -2.5 59.5 E177 silent
E178 2024-05-10 decay -2.4 57.1 E178 silent
E179 2024-05-17 oppose -10.8 46.2 E179 Chamath opposes x2 [w=0.90]
E179 2024-05-17 reversal -27.7 18.5 E179 Sacks flips
E180 2024-05-24 decay -0.7 17.8 E180 silent
E181 2024-05-31 decay -0.7 17.0 E181 silent
E182 2024-06-07 decay -0.7 16.4 E182 silent
E183 2024-06-14 decay -0.7 15.7 E183 silent
E184 2024-06-20 decay -0.6 15.1 E184 silent
E185 2024-06-29 decay -0.6 14.5 E185 silent
E186 2024-07-04 decay -0.6 13.9 E186 silent
E187 2024-07-12 decay -0.6 13.3 E187 silent
E188 2024-07-19 decay -0.5 12.8 E188 silent
E189 2024-07-26 decay -0.5 12.3 E189 silent
E190 2024-08-02 decay -0.5 11.8 E190 silent
E191 2024-08-09 decay -0.5 11.3 E191 silent
E192 2024-08-16 decay -0.5 10.9 E192 silent
E193 2024-08-23 decay -0.4 10.4 E193 silent
E194 2024-08-30 decay -0.4 10.0 E194 silent
E195 2024-09-06 decay -0.4 9.6 E195 silent
E196 2024-09-20 decay -0.4 9.2 E196 silent
E197 2024-09-27 decay -0.4 8.9 E197 silent
E198 2024-10-03 decay -0.4 8.5 E198 silent
E199 2024-10-11 decay -0.3 8.2 E199 silent
E200 2024-10-18 decay -0.3 7.8 E200 silent
E201 2024-10-25 decay -0.3 7.5 E201 silent
E202 2024-11-01 reinforce +12.9 20.4 E202 Sacks support x3 (flipped from reversal) [w=0.77]
E202 2024-11-01 oppose -10.3 10.1 E202 Friedberg opposes x1 [w=1.08]
E202 2024-11-01 oppose -10.1 0.0 E202 Jason opposes x3 [w=1.05]
E203 2024-11-08 decay · 0.0 E203 silent
E204 2024-11-16 decay · 0.0 E204 silent
E205 2024-11-23 decay · 0.0 E205 silent
E206 2024-12-07 decay · 0.0 E206 silent
E207 2024-12-13 decay · 0.0 E207 silent
E208 2024-12-20 decay · 0.0 E208 silent
E209 2025-01-04 decay · 0.0 E209 silent
E210 2025-01-11 decay · 0.0 E210 silent
E211 2025-01-18 decay · 0.0 E211 silent
E212 2025-01-25 decay · 0.0 E212 silent
E213 2025-01-31 decay · 0.0 E213 silent
E214 2025-02-07 decay · 0.0 E214 silent
E215 2025-02-15 decay · 0.0 E215 silent
E216 2025-02-21 decay · 0.0 E216 silent