any sufficiently advanced technology is indistinguishable from magic. senior software engineer @coinbase
Joined December 2019
- Tweets 3,459
- Following 829
- Followers 5,249
- Likes 10,729
780 Photos and videos
Pinned Tweet
Jan 18
0/ i analyzed every single trade on kalshi from 2021 to 2025.
i found a systematic wealth transfer where "takers" pay a massive premium for affirmative outcomes, and "makers" harvest the edge without needing to predict the future.
here is the data
11
9
148
32,515
Feb 10
> what can i do with this data?
make sick visualizations, perform deep data analysis, contribute to prediction market research, ...
Feb 10
added polymarket data to the public dataset.
400m trades going back to 2020.
36gb compressed.
MIT licensed, free to download via @Cloudflare R2.
17
19
861
177,636
Feb 10
* data is trade-weighted, not contract weighted
spikes in miscalibration in 2024 are from presidential election markets
37
5,736
Feb 10
added polymarket data to the public dataset.
400m trades going back to 2020.
36gb compressed.
MIT licensed, free to download via @Cloudflare R2.
130
239
3,986
762,580
Feb 10
6
16
359
45,620
Jan 28
prediction markets are remarkably accurate
Jan 28
Prediction markets are the ultimate form of truth seeking. When there’s skin in the game, the output is far more reliable. Everything else is biased by someone's agenda. I think we’ll look back at prediction markets as a breakthrough in how we discover truth in the world.
Prediction markets are now fully live in the U.S. on Coinbase, you can trade what’s happening in sports, politics, culture, news, and more.
Update your app and look for the "Predict" tab at the bottom.
1
1
22
2,756
Jan 25
started collecting all polymarket data as well for a comparative accuracy analysis
Jan 18
0/ i analyzed every single trade on kalshi from 2021 to 2025.
i found a systematic wealth transfer where "takers" pay a massive premium for affirmative outcomes, and "makers" harvest the edge without needing to predict the future.
here is the data
3
7
1,669
Jan 19
2
1,084
Jan 18
0/ i analyzed every single trade on kalshi from 2021 to 2025.
i found a systematic wealth transfer where "takers" pay a massive premium for affirmative outcomes, and "makers" harvest the edge without needing to predict the future.
here is the data
11
9
148
32,515
Jan 18
2/ if interested in the dataset, you can find it here
github.com/jon-becker/predic…
2
1
13
3,475
Jan 18
3/ that's all for now! all feedback is appreciated <3
let me know if i've missed anything, got anything wrong, etc. it helps me improve and learn! 🦦
1
6
2,566
Jan 6
crypto traders wake up at 9am, sports bettors stay up until midnight, and finance bros only show up when jpow speaks
1
18
1,152
Jan 6
note: this is a single chart from my upcoming paper. everything will be open-source
data sourced from kalshi's public API
4
733
jon becker retweeted
23 Dec 2025
Yes!! I've been thinking about this since some time:
Given large verified contract datasets (from @SourcifyEth and @verifalliance), can we build better decompilers?
Looking forward to what comes out of this 👀
23 Dec 2025
4/ next up: pulling thousands of verified contracts from sourcify.dev to build a massive evaluation dataset
these larger evals might not feasibly run on every commit, but will likely be a weekly long-running job to detect regressions
2
3
1,377
23 Dec 2025
0/ how do you measure if a decompiler is actually good?
string matching? nope. variable names are gone. arg names don't exist. comments are dust.
you need something that understands 𝘴𝘦𝘮𝘢𝘯𝘵𝘪𝘤 𝘦𝘲𝘶𝘪𝘷𝘢𝘭𝘦𝘯𝘤𝘦
4
1
18
1,639
23 Dec 2025
4/ next up: pulling thousands of verified contracts from sourcify.dev to build a massive evaluation dataset
these larger evals might not feasibly run on every commit, but will likely be a weekly long-running job to detect regressions
2
1
9
2,071
23 Dec 2025
5/ if interested, check out the evaluation repo here:
(very minimal, for now)
github.com/Jon-Becker/heimda…
10
668