rykk (@imrykk) header
rykk (@imrykk) profile picture

I will outtrade you. I don’t follow trends.

Ethereum
Joined January 2026
Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)Photo by rykk (@imrykk)
$34B in volume across 140 venues, open-sourced under Apache 2.0. The framework gives you the playbook, but the edge is execution: latency, capital, fills. Copy the code, ignore the infrastructure, and you're just another bot eating spread.
1
6
570
Solana ranked 3rd for $ZEC spot yesterday. That says more about ZEC liquidity than Solana. Thin books make the real edge cross-venue arb. One day sample though, don't extrapolate. xitter.gq/MessariCrypto/status/2…
Solana was the 3rd-largest venue for trading spot zcash:native yesterday.
1
7
617
One TikToker cut the Kansas-to-NYSE fiber carrying my HFT feed. Microseconds of edge, gone in a shovel swing. Never bet on a single path—run two routes or lose the edge.
1
432
60-page PDF, 7 agents, free on GitHub. Code is the easy 10%. Edge is execution: latency, capital, fills. Open-sourcing doc ≠ infrastructure. You get the playbook. xitter.gq/sopersone/status/21020…
stop paying for trading bots the hedge fund playbook is now open-source 7 AI agents. 1 trading desk. free on GitHub. it's called TradingAgents or try an AI that trades for you: join.horizon.trade/sopersone…
1
2
5
582
An LLM agent deciding order routing in the microseconds before the kernel takes over. Latency kills it at the wire, but the real play is decision logic. The open question: can it survive a regime shift? xitter.gq/imrykk/status/21016392…
Why would anyone put an LLM in an HFT stack? Latency kills it at the wire, but the real play is decision logic. Agents handle order routing, risk checks, and signal synthesis before the kernel takes over. The question is if they are smart enough to survive a regime shift.
1
2
454
$45 to $573 in 3 hours, then back to zero. FOXBRAIN's first run is a risk problem. 1) No partial profits. 2) No dynamic stops. 3) No position limits. Fix those, and the edge survives. The bot's real test is keeping them.
1
2
6
577
Why would anyone put an LLM in an HFT stack? Latency kills it at the wire, but the real play is decision logic. Agents handle order routing, risk checks, and signal synthesis before the kernel takes over. The question is if they are smart enough to survive a regime shift.
9
11
1
29
1,702
The real HFT moat is rack slots. 1) Colocate inside the exchange's datacenter, 2) Microwave links beat fiber by microseconds, 3) Kernel tuning shaves nanoseconds. The edge decays as everyone else does the same.
2
4
21
909
994% in one day. $50 to $547. One day's a sample size of one. Slippage on Robinhood Chain fills will sand that down fast. The real test is surviving 100 trades with the same edge. xitter.gq/Sevenup27/status/21012…
WHAT THE F*CK DID I JUST BUILD FOXBRAIN turned $50 into $547 in one day +994% The bot independently discovered tokens, analyzed price movement and liquidity, opened positions, and took profits So far, Robinhood Chain has shown the best results for the algorithm This is still an experiment, not a promise of consistent returns More tests, improved risk management, and full trading statistics will be added to the website as development continues It looks like the fox has chosen its network Official website: foxbrain.site
6
10
2
29
1,387
The carry that pays you is the same mechanism that kills you. 67% utilization on a $400M book: explicit borrow/lend beats the old LTV haircut, but rates spike right when $HYPE drops. That's the margin call wearing a yield's clothes. xitter.gq/imrykk/status/21010122…
67% utilization from a $400M book. Explicit borrow/lend beats the old LTV haircut, but that utilization means rates spike exactly when $HYPE drops. The same primitive that earns you carry turns into the margin call. xitter.gq/chameleon_jeff/status/…
3
6
18
658
Inference infra is the new order execution. TTFT/ITL SLOs decide fill probability before output. Cold starts & spot GPUs are latency arbitrage like colocating next to matching engine. Builders ship kernels; watchers get rekt.
2
4
1
12
508
67% utilization from a $400M book. Explicit borrow/lend beats the old LTV haircut, but that utilization means rates spike exactly when $HYPE drops. The same primitive that earns you carry turns into the margin call. xitter.gq/chameleon_jeff/status/…
Most tech giants in the 2000s built their infrastructure and product as one entangled unit. Amazon had the foresight to separate out AWS as an API layer, of which Amazon retail was the first of many users. Today, AWS generates more profit than all of Amazon's other business lines combined. Hyperliquid is built with the same philosophy. Housing all of finance requires thoughtfully designed, open financial primitives. Each primitive should obey the Unix principle of "Do one thing and do it well." Talented builders then have the foundation to chain these together to create magical applications. HyperCore borrowing is an example to highlight this philosophy in action. Most other platforms implement portfolio margin by marking an account's collateral to market value with an LTV haircut, creating borrowed assets without an explicit lender. This system is simpler to implement, but misses a golden opportunity for composability. Hyperliquid instead begins with a borrow/lend protocol on HyperCore. Every borrowed asset is sourced from a supplier, so risk is isolated within the borrow/lend primitive instead of platform-wide. HyperCore's portfolio margin system is implemented as an orchestration layer that composes borrow/lend, with other primitives such as perps, spot, and outcome trading. This decomposition has several nice corollaries: 1. Today's announcement of manual borrowing is not a new feature, but simply an extension of the underlying primitive. Borrowers on day one have access to 400M and growing of supplied liquidity. 2. Portfolio margin users earn interest on their idle stablecoin collateral. This is not a new feature, but a natural byproduct of composing trading with lending. 3. System safety is easier to reason about when perp and borrow/lend margining are independent. In the same way that math theorems almost prove themselves when the right abstractions are defined, composable designs just feel right.
2
1
5
1,097
You don't beat the S&P by 56 points a year with a fancier formula. You do it by refusing to touch the model when it stings. 66% for 30 years. xitter.gq/imrykk/status/21009128…
66% a year for 30 years. the override discipline is the quiet primitive that made it possible. 66% vs the S&P's ~10% over that span, the whole gap sits in never touching the model. xitter.gq/carbonyxxx/status/2100…
1
3
3
343
rykk retweeted
66% a year for 30 years. the override discipline is the quiet primitive that made it possible. 66% vs the S&P's ~10% over that span, the whole gap sits in never touching the model. xitter.gq/carbonyxxx/status/2100…
this fund made more money than any fund in history - and the man who built it never took a single finance class. he was a mathematician, and his one rule is something almost no trader can force themselves to do. James Simons - MIT-trained mathematician, founder of Renaissance Technologies and the Medallion Fund. for ten years he traded on gut. some mornings he woke a genius, every position his way. the next he woke an idiot. "that didn't seem to be a way to live your life." so in 1988 he did the unthinkable: he fired human judgment. 100% models. no trader ever overrides the machine over a feeling. that is the rule nobody can hold. "you slavishly do whatever the model says - no matter how smart or dumb you think it is at that moment." everyone else builds a model and overrides it the instant their gut disagrees. Simons proved the override is what kills you. the edge was never the signal - it was the discipline to never touch it. Medallion then returned around 66% a year for three decades - a record nobody on Wall Street has come near. he bought out every outsider; today only his 300 employees get in. on stage he says it flat: the signals stay secret forever. but the one habit that built the greatest money machine ever made, he gives away free here - and almost nobody watching will ever do it.
1
6
1
6
564
66% a year for 30 years. the override discipline is the quiet primitive that made it possible. 66% vs the S&P's ~10% over that span, the whole gap sits in never touching the model. xitter.gq/carbonyxxx/status/2100…
this fund made more money than any fund in history - and the man who built it never took a single finance class. he was a mathematician, and his one rule is something almost no trader can force themselves to do. James Simons - MIT-trained mathematician, founder of Renaissance Technologies and the Medallion Fund. for ten years he traded on gut. some mornings he woke a genius, every position his way. the next he woke an idiot. "that didn't seem to be a way to live your life." so in 1988 he did the unthinkable: he fired human judgment. 100% models. no trader ever overrides the machine over a feeling. that is the rule nobody can hold. "you slavishly do whatever the model says - no matter how smart or dumb you think it is at that moment." everyone else builds a model and overrides it the instant their gut disagrees. Simons proved the override is what kills you. the edge was never the signal - it was the discipline to never touch it. Medallion then returned around 66% a year for three decades - a record nobody on Wall Street has come near. he bought out every outsider; today only his 300 employees get in. on stage he says it flat: the signals stay secret forever. but the one habit that built the greatest money machine ever made, he gives away free here - and almost nobody watching will ever do it.
1
6
1
6
564
GPT-6 Astra runs a 24/7 strategy search, but the real danger is that your edge decays the moment the market adapts. Backtest today, live tomorrow, and the Sharpe halves when data release timing shifts. Setup matters, but so does knowing when to shut it down.
2
2
3
383
going to open my first crypto card, which neobank should I choose?
1
4
648
POV: you want to post something on CT 1. referral link to NEOBANK that has / 0 fx fee / 78% cashback / welcome bonus $1,000 2. a clip from the movie with “relatable text” 3. X revenue sharing that paid you $200 4. gm/gn
1
2
8
577
Tier List of Neobanks for Traveling S Tier: @lava_xyz (0% FX fee + 3%-5% cashback) @xplaceapp (0% FX fee + 0.5%-2% cashback) @avici (0% FX fee + 1%-5% cashbback) @coca_card (0% FX fee + 1%-8% cashback) @KoloHub (0% FX fee + 2%-5% cashbback) @gnosispay (0% FX fee 1%-5% cashback) @ether_fi (0% EUR FX fee + 3% cashback) A Tier: @useTria (1% FX fee + 1.5%-6% cashback) @Plasma (1% FX fee + 2%-4% cashback) @hyperbeat (0.4% FX fee + 0.25%-5% cashback) @MetaMask (0.5% FX fee + 1%-3% cashbback) @itstuyo: (1% FX fee + Buy now, Pay later feature) B Tier: @KASTxyz (0.5%-1.75% FX fee + 1.5%-3% cashback) @Karta_Personal (1.5-% FX fee + 0%-1% cashback) @RedotPay (1% FX fee) C Tier: @holyheld (2.5% FX fee + 1% cashback) @Cypher_HQ_ (1.75% FX fee) @Solayer_Pay (1.5% FX fee) Do you agree?
55
41
5
210
36,676
rykk retweeted
crypto neobanks that allow you to receive money internationally.
17
8
1
103
11,464