JD🕶⚡️ (@janndriessen) header
JD🕶⚡️ (@janndriessen) profile picture

JD🕶⚡️

@janndriessen

cooking 🕶️ prev @indexcoop @bloomcbt @freeletics

 Outer Rim
janndriessen.com
Joined May 2012
  • Tweets1.3K
  • Following3.1K
  • Followers582
187Media
Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)Photo by JD🕶⚡️ (@janndriessen)
Currently experimenting with a new open-position screen for my hyperliquid trading app.
JD🕶⚡️ (@janndriessen) profile pictureJD🕶⚡️@janndriessenOct 25
After building the Rust order book, I went back to the order book in my personal Hyperliquid iOS trading app applying some of those learnings. Live Hyperliquid depth, tap a level to set TP/SL, zoom $0.1 → $100. Entry, TP, SL and liq sit on the ladder.
1
2
91
Wondering what @alica_ and @chavis_ would do (different)? 🤔
1
1
26
⚡️
Image from JD🕶⚡️ (@janndriessen)
1
185
After building the Rust order book, I went back to the order book in my personal Hyperliquid iOS trading app applying some of those learnings. Live Hyperliquid depth, tap a level to set TP/SL, zoom $0.1 → $100. Entry, TP, SL and liq sit on the ladder.
JD🕶⚡️ (@janndriessen) profile pictureJD🕶⚡️@janndriessenOct 25
In a recent system design interview I was asked how I'd build an exchange-style order book, matching engine included. Naturally, I had to actually build one to learn how it really works. 🧵
2
1
5
284
What carried over: • Do the work once per update: rows are resolved into one Equatable model per book, so SwiftUI skips every row that didn't change • Stable identity: rows keyed by price, not index, so a best-bid tick never rebinds a row • Only flash what's actually new
47
In a recent system design interview I was asked how I'd build an exchange-style order book, matching engine included. Naturally, I had to actually build one to learn how it really works. 🧵
2
1
4
597
more replies
Biggest learning: measure first before optimizing. almost all of the lag is not the code. ~0.3 s is the source's own delay before data reaches the service. The engine adds tens of milliseconds (batching, waiting for all 7 feeds), the code itself under one.
2
67
Code, measurements and trade-offs: github.com/janndriessen/orde… If you've built or run a real matching engine or market data system, I'd love your feedback. What would you have done differently?
GitHub - janndriessen/orderbook: An exchange-style order book service and matching engine for a...

GitHub - janndriessen/orderbook: An exchange-style order book service and matching engine for a...

An exchange-style order book service and matching engine for a single market (BTC-USD), driven by order flow derived from Hyperliquid data. - janndriessen/orderbook

github.com
40
Trade-offs, on purpose: • public price levels instead of running a Hyperliquid node, so queues are synthetic • one engine, no journal or replicas • JSON over WebSocket, not a binary feed All listed in the README, next to how production systems do it.
14
Some numbers: • ~10M commands/s in the matcher (~100 ns each) • ~0.2 ms of compute per live update, end to end • ~83 ms behind Hyperliquid's own feed when run locally • identical to Hyperliquid's own book on 28 of 28 updates compared side by side
17
JD🕶⚡️ retweeted
Porsche (@Porsche) profile picture
Porsche
@PorscheOct 15
Mission S. Authentic. Consistent. Timeless. A glimpse into the future - shaped by our heritage. Mid-engine architecture. A focus on the driver and the driving experience. Call it a love letter.
Image from Porsche (@Porsche)
6,646
4,752
4,294
77,137
20,720,081
Great talk by @lil_dill youtube.com/watch?v=GLvFTMtw…
How to scale intent, quality, and artistry with Al | Katie Dill (Stripe)

How to scale intent, quality, and artistry with Al | Katie Dill (Stripe)

As AI makes it easier to build, how do we keep products from becomi...

youtube.com
3
124
JD🕶⚡️ retweeted
dax (@thdxr) profile picture
dax
@thdxrOct 6
a weird inversion with LLMs is the models improve faster than the tinkerers when i see people with custom workflows and setups they're all addressing problems that don't exist anymore the person naively using vanilla codex is more likely to be experiencing state of the art
657
654
297
13,361
1,009,556
No one can print more bitcoin. block.xyz/bitcoin/does
19
74
14
409
105,312
I'm usually not a hype guy but Sonnet 5.5: youtube.com/watch?v=N03Ho4VM…
Wolf of Wall Street meme - Holy F**king Sh*t - Green Screen

Wolf of Wall Street meme - Holy F**king Sh*t - Green Screen

Free to use Green Screen Meme Template of Mark Hanna (Matthew McCon...

youtube.com
Claude (@claudeai) profile pictureClaude
@claudeaiSep 28
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family. It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.
1
332
i think about this at least once a week
Image from drew coffman 𝕚𝕤 𝕠𝕟𝕝𝕚𝕟𝕖 🟢 (@drewcoffman)
27
15
12
372
16,448
JD🕶⚡️ retweeted
Heinrich (@arscontexta) profile picture
Heinrich
@arscontextaSep 23
me talking to claude in english as a german
53
88
30
2,751
512,233
Humanity is $COOKED. Unless we change course. We're live: a meme coin bringing attention to Al existential risk-and ways to act. Come for the memes. Stay for the evidence. Make your voice heard. The future isn't decided. humanityiscooked.com
27
2
28
433
JD🕶⚡️ retweeted
Thariq (@trq212) profile picture
Thariq
@trq212Sep 18
We're adding support for AGENTS.md to Claude Code. Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md. You can toggle this behavior in /config.
2,050
2,628
2,521
31,287
5,637,641
Load more