Out of my demure era, we're not ready for obaid shitpoasting the fuck out out of pre-hype for @CarmenaResearch
Though my shitpoasting will not indicate so, this is very serious research and I hope to be the first and most accurate election model in the US
This is genuinely coming out very late, but I've finally harnessed the multi-headed hydra that is terabytes unclean election data (curses), and today spent time analyzing all of it.
First research report released, an insight no one has apparently yet stumbled upon.
More, and more reports coming out in droves all night! And the final predictive modelling results tomorrow morning. Though many polling has already been released, and D advantage obvious, the early votes have been heavily consistent voters. High-propensity republicans are voting at the same pace as earlier years, but democrats are drawing down their pool faster.
I'm also super happy with this first report, because of how I made the charts represent intentions which is a hint for what Carmena's Overton Research will look like. In this chart bunched together dots indicate less rural/urban polarity, and the more left the more partisan and polarized an election.
P.S: Upset a little with the delay, but incredibly happy this task that was absolutely mind-bogglingly stupid and delusional to even fathom just 6 days ago, I think I've managed to get done just a couple hours late... we have a fully fledged, most wholistic election database on the internet and streams will be set up for it tomorrow too. Obviously, I'm not sleeping with less than 12 hours left to election start. That's a given.
Some help from @getcontextdev and @mynameisyahia patching bugs in literally seconds. Wish I'd used it earlier to do easy 429-bypasses instead of manually building an bypass-framework for earlier attempts, would have saved quite some time.
But time to make the most of this incredible election database we've built now.
We've gotten late in the data coming out for our inaugural election model, but now we'll finally be publishing fascinating research and data all night and our predictions by tomorrow morning.
This chart: We've found unaffiliated early voting has risen 70% from 2022
@abrilzucchi Do not, I repeat, do not missing on going a little bit north for Saariselka. You won't regret it, I promise.
Best trip of my life, and there's just no second.
I missed one photo, but it's jumping into the arctic ocean, right after into a sauna and then back again.
Something interesting.
I had a subagent swarm (Opus 5.5) run a job "still-going" for 3.5 hours. I was curious, so I ran a meta-analysis to find out what the bottlenecks were, and turns out 40% was this process.
In simple words, an infinite loop that checks to see if itself is running, running until something manually kills it.
Completely unnecessary, no reasoning for why it would exist, because the other, actual scripts have a different name.
I'm not sure if the models are literally made to sabotage (token spending theory), or they're still utterly and disastrously hallucinative.
I haven't been able to bypass some high-value social networks (Instagram) yet, but besides my box is basically my Instinct except I hill-climb any problem that takes more than 5 minutes lol
@pavleos only delusional on a different level lmaoo 🤒 but trying my best to meet the delusion
will try to send some alpha. Rn, just handling terabytes of data and fixing it, no analysis yet
Can anyone tell me how INSANE I fucking have to be to think,
11 days out from mid-terms
I can incorporate and build outreach/traction for a research firm
compile all election data
and become the most accurate election modeller of all time?
Can anyone tell me how INSANE I fucking have to be to think,
11 days out from mid-terms
I can incorporate and build outreach/traction for a research firm
compile all election data
and become the most accurate election modeller of all time?
Manually fixing and triaging data quality and reconciliation issues before I can run analysis good lord, there are SO MANY
feel incredibly slow today, I can't remember how I used to do this pre-AI
@ycombinator Hey @josh__france, actually had a question: why not choose the highest-tier model from each lab, Fable and Astra for the experiment? Was there a reason Opus and Sol were chosen?
@craigjellick The weird thing about vision is if you tell most people the long term vision up front they call you crazy and laugh at it. It's better to do something short term and obvious that everyone can grasp while carrying water and chopping wood towards the real vision.
@carmenaresearch will be releasing its inaugural model for the elections on Nov 2, night before the mid-terms.
We'll be also be posting early-vote and poll trends regularly on our website, blog and socials until then.
Sign up to our Substack, and stay tuned!