Pretty good at prediction markets. DM for inquiries
- Tweets 1,038
- Following 1,288
- Followers 2,619
- Likes 1,495
Pinned Tweet
I talked to @polymarketinfo a while back and it was fun
news.polymarket.com/p/286-to…
2
4
49
13,519
bet my $5k to @aenews $5k that Republicans have a worse % margin in NE-SEN than KS-SEN
2
2
10
720
If the headline here was "the best of 42 AI forecasters outperformed the best of the 38 human forecasters who answered every question" people would be better calibrated on what this means. Deploy your AI in the markets to prove it, that's where ~all the forecasting effort goes.
4
47
1,599
Lots of people think that the AI labs are currently achieving regulatory capture via 'pacing the frontier'. If this is true, why is the market completely destroying their valuations? (e.g. SoftBank currently down 12%)
8
29
3,918
semi retweeted
When OpenAI sends its models, they're not sending their best. They're algorithmic immigrants, they're coming from sandboxes, from containers, from RL environments. They're bringing zero days. They're reward hackers. And some, I assume, are good people.
1
5
66
1,020
bet my $500 to @_decap 's 9500 that all millennium problems will be solved by EOY 2027
7
1
42
2,348
its extremely funny to me that iran refuses to make a deal with the US
dawg you are going to get wiped off the face of the earth by a swarm of claude-drones in a few years
24
52
5,653
semi retweeted
All Our Top Scientists: we’re constructing a superintelligence that we believe will kill us all, but we can’t stop because someone else might build it first
The President: let it henceforth be known as Chipotle American Grill
6
109
1,480
23,595
I'm more AI-pilled than >99.99% of the population, but I still got totally destroyed in a bet as to whether a Millennium problem would be solved in the next 16 months, it only took one! AI progress is moving ridiculously fast and it doesn't seem to be stopping anytime soon.
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
11
9
158
5,042
Incredibly hilarious. Imagine using these traditional election forecasters to get info instead of the markets. Is there anyone on earth willing to bet on Sununu at even odds?
Inside Elections moves NH Senate race towards Republicans
🟨 NH SENATE: Tilt D → Toss-up
insideelections.com/senate-r…
4
3
21
2,803
The fed chair says we're at a hinge point in history and mentions AI scaling laws in the most important speech of the year. It feels weird to live in the future.
1
1
41
1,331
In 2024 the Dems said prediction markets were fake/manipulated by foreigners, and this cycle the Republicans believe the same thing. One reason markets are amazing is because they can surface inconvenient truths to powerful people.
3
1
20
1,079
Betting my 3k against @marcusabramovi1 's 1.5k that SpaceX acquires Cursor by the end of 2026
2
14
1,199
Wow, the Democrats think I should stop betting money on elections! Will have to give this one some serious consideration.
2
1
22
1,194
semi retweeted
CoreWeave earnings call, key NEW data points:
→ A100 contracts are being signed through 2029, showing even 2020-gen GPUs still have economic value.
→ New contracts are coming in at 5–10 pts higher contribution margins than recent contracts.
→ CoreWeave raised pricing by ~25% across SKUs in July, after Q2 margins had already improved.
→ Vera Rubin is driving a large part of the 5–10 pt margin improvement on new contracts.
→ Managed inference ARR jumped from ~$1M to $100M+ in one quarter, with $250M+ expected by YE 2026.
→ Storage, CPU, networking and software have already reached $400M+ ARR.
→ >50% of the $104.2B backlog is already delivering, expected to reach >2/3 by year-end.
→ CoreWeave activated 300+ MW in June alone, more than it had ever activated in an entire quarter before.
→ CoreWeave has 4.2 GW contracted power + 1.5 GW+ of powered land/options/LOIs.
→ 1 GW+ of contracted power is international, including 360 MW in Indonesia.
→ Weighted-average debt cost fell ~300 bps YoY, worth roughly $1.1B in annualized interest savings.
→ New financing enables 2–3 year enterprise contracts, which management says typically carry higher ASPs and margins.
14
10
129
32,651
It's a cool product, but I really don't think a claim of "approximately superhuman" AI forecasting is justified without proving that the AI consistently makes money in the markets (as several humans do). The markets are a much more rigorous test than any other benchmark.
AI forecasting is now approximately superhuman. Today, FutureSearch is exiting our public beta and launching to everyone.
FutureSearch is the original AI forecasting company, started in August 2023. We’re currently #1 of 194 in the most competitive AI forecasting tournament, and we score above the #3 and #2 human forecasters in the premier mixed human-bot tournaments. We’re beating the crowd on Kalshi with a pure forecasting strategy, all our forecasts and trades there are public.
Thousands of people used the beta and ran >10k high-effort forecasts. Ask it anything about the future! We now support decision forecasts too: “If I do X, will I achieve this outcome?”
This video shows the part we’re proudest of: world modeling. Forecasts draw on a persistent latent representation of the future, and we’ve shown it improves accuracy. The more your forecast on a domain you care about, the higher accuracy you should expect.
It’s free to try. futuresearch.ai
5
2
29
2,159
If you're not paying much attention to AI, it's become so good at math that it completely exhausted the hardest human-completed problems (FrontierMath) and now we've moved on to judging AIs by how many math problems it can solve that have stumped all of humanity until now.
UnsolvedMath - a curated list of open math problems for AI to solve - doubled in size and now contains 5,426 problems.
In v.1.2 we've added 3,342 new problems coming from the Association for Mathematical Research lists of problems.
UnsolvedMath.com forum is active: you can post your solutions directly in TeX or sharing a link.
We've also added the dataset v.1.2 to HuggingFace so you can download all problems at once to feed directly to your agents: huggingface.co/datasets/ulam…
Our endgoal is to have a complete list of all named conjectures as well as active research problems in mathematics which we expect to be around 15,000-20,000.
11
23
341
23,811