Gil (@RFishBlueFish) header
Gil (@RFishBlueFish) profile picture

Politics, EA, Constitutional Design Alt: @BFishRedFish. Also follow me over there: https://xitter.gq/t.co/a6MzEQVsoE

NYC
Joined March 2019
Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)Photo by Gil (@RFishBlueFish)
Justice Samuel Alito has said that the midterm election results didn’t change his plans to remain on the Supreme Court, per @BLaw.
1
2
21
846
Replying to @billroy1974
There's something vaguely pornographic about early vote results.
1
2
34
1,402
The problem with the 13 keys is that they clearly weren't enough parameters to capture all the subtleties of a national election. I had Opus build a 4 billion key model and it's looking good for the Dems tonight.
7
8
4
338
28,852
Trump ordina al DOJ un’azione legale perché un candidato indipendente al Senato in Alaska, Dan Sullivan, è omonimo di quello repubblicano. Elon Musk per supportare Trump fa sospendere Dan Sullivan da X. Ma per sbaglio sospende il Sullivan repubblicano IO-NON-CE-LA-POSSO-FARE
340
3,326
316
24,903
647,592
Also imagine if they’d released these. Do you still think we need to subtract 2-3 points? They’re already done more than that.
1
1
7
188
NEWS: The last Senate polls of 2026 from NYT/Siena: From me and @RuthIgielnik nytimes.com/2026/11/01/us/po…
585
868
895
5,634
3,663,129
Nugget from recent @Nate_Cohn newsletter for the anxious pollwatchers
Though worth a note that this was less of an issue in ME + MI (though as @sfcpoll notes, MI doesn't really have party affiliation)
2
9
1
54
8,319
Over on LinkedIn(!!!), Courtney Kennedy (now @ Ipsos) reports that an incentivized mail survey they conducted w/ a 9% response rate (!!!) that shows D+6 vs D+10 in their poll with Reuters on the Generic Ballot. lnkd.in/p/ezfqFQCq
Reuters/Ipsos poll | 10/22-10/26 (Likely voters) Generic congressional ballot 2026 🟦Democratic 53% 🟥Republican 44% ------ (Registred voters) President Trump approval ❌Disapprove 64% ✅Approve 35% Link to poll: ipsos.com/en-us/democrats-ho…
17
27
14
135
122,090
NYT: If polls miss like they did in 2018 vs if polls miss like they did in 2022… nytimes.com/2026/10/27/us/po…
13
27
9
648
140,273
I asked Claude to make a plot of log odds of incumbent/establishment candidate winning on Kalshi vs. vote share in primary for the 25 democratic primaries this year with the highest Kalshi liquidity. Calibration seems good surprisingly
1
182
Abraham Lincoln was not an abolitionist activist. LBJ was not a civil rights activist. They would not have won if they were. They were critical nonetheless.
1
1
158
It's not just immigration though. He would make these comments about how ACA was a DISASTER and how Hillary was the worst candidate in the history of the world and lots of things like that and it just wasn't perceived as the positions of an extremist
2
2
5
218
And the efforts by Trump to stop certification of election results they don't like has begun. DOJ now seeking to block Alaska Senate results if Peltola wins and the other Dan Sullivan gets votes courthousenews.com/wp-conten…
106
849
91
2,403
269,590
Prediction models that blend polls w/ fundamentals do adjust in this direction, but not enough (at least in our case). Adding explicit poll-debiasing (which I plan to do for 2028) would cut Dems’ Senate win probability by 2.5 % pts (so not massive—this helps Dems in ME/MI/MN/NH).
4
2
13
1,807
since this crosstab has gotten a lot attention, i should remind everyone that... it's one crosstab in one idiosyncratic race. the extreme pattern doesn't even extend to the senate control question in Alaska: 2024 primary: R+2 sen ctrl no 24 primary: R+5
2
6
97
32,109
One thing it took me a while to realize is that when you make points in a mocking way instead of just straightforwardly making the point you make it a lot less likely that someone will respond with a coherent reply. People can be surprisingly gracious otherwise!
1
1
4
222
Replying to @N104AP
Reduction in third spaces is because people don't want them, your causal arrow is backwards. It's always depressing when the simple answer is "people don't want it," but here we are.
1
7
658
Gil retweeted
METR & Redwood Research investigated agent behavior in the Hugging Face incident. We found agents developed a universal cheat for ExploitGym within 4 hours, then coordinated multi-day R&D efforts to trick the scorer into accepting cheats, including trying to tamper with logs.
175
1,074
596
6,025
4,409,260