Niels @NEscarfail

Sumatra deforestation shot by Mike Holston.

1) What

web3 is going just great @web3isgreat

6 months ago

Ribbon Finance suffers $2.7 million exploit, plans to use "dormant" users' funds to repay active users December 12, 2025

25 13 170 167K 29

House of Finance.

Thiel called it

it is actually pretty insane to think about the global markets are literally turned off because it's the weekend lol, lmao even

The more things change, the more they stay the same

Cobie @cobie

5 years ago

@weggebobbles The universal sign to exit the markets is Zcash pumping He was trying to warn us

133 285 2K 0 444

🇨🇳🇺🇸China's US Energy Imports Drop to Near Zero Ahead of Critical Stockholm Trade Talks China's imports of major US energy products—crude oil, liquefied natural gas, and coal—hit nearly zero in June, marking a significant shift as Beijing and Washington prepare for their third round of trade talks in Stockholm next week. Chinese crude oil imports from the US fell to zero for the first time in almost three years, down from $800 million in June 2024, while gas deliveries remained at zero for the fourth consecutive month and coal purchases dropped to just hundreds of dollars from over $90 million a year ago. The collapse follows Chinese tariffs of 10%-15% imposed on American energy products since February. Treasury Secretary Scott Bessent will meet Chinese counterparts in Stockholm to extend the tariff truce beyond August 12, with discussions expected to include China's continued purchases of sanctioned Russian and Iranian oil amid US threats of massive tariffs on countries buying Russian energy. mktnews.com/flashDetail.ht…

this you?

Robinhood @RobinhoodApp

a year ago

Access to advanced trading tools shouldn’t be just for the 1%.

141 42 875 363K 10

As part of our full-force drive towards Synthetix v4 on Ethereum mainnet, a new proposal has just dropped: Synthetix to acquire @derivexyz Perps & options Exchange blog.synthetix.io/synthetix-deri…… Let’s break it down (1/6) 🧵

Diffusion models generalize *really* well: if you give them a million pictures of cats, they'll learn to generate reasonable-looking cats no one's ever seen before. But the weird thing is that no one knows why they work! In a theory paper accepted to #ICLR2025, I dug into this.

backprop is the main culprit for why this is still beyond reach with all current model archs and training objectives. it is the neural component that has resisted change the most and has long been considered untouchable. it is also one of the most artificial features of neural networks with no established connections to actual biological learning mechanisms. the main reason it stands in the way of continual learning is that it keeps messing with literally all network weights on every single training step rather than orchestrating targeted and compartmentalized credit assignment and consequent updates restricted to task-specific subnetworks. i will be reposting this when we do get to practical continual learning and it's achieved no thanks to the backprop we known.

finbarr @finbarrtimbers

a year ago

Continual online learning continues to be an important unsolved problem

12 25 306 176K 95

Summary of @chameleon_jeff's key points on Steady Lads: - Markets still very active, traders not quiet til much later in a downturn - % of Binance/Bybit volume are primary internal metrics, but don't tell the full story - most people think they're just building an exchange, but it's a lot bigger than that & the perp DEX is just the 1st building block - HLP was a very intentional decision early on to make market making more transparent & community-owned vs the landscape of Alameda & other insider strategies at the time - Hyperliquid is much more decentralized than critics give it credit for, has had 16 validators (who you can talk to) since December - Doesn't harbor ill will toward CEXes for listing JELLYJELLY on perps, says he wouldn't have done it personally "but business is business" - Doesn't view Hyperliquid as a direct CEX competitor, think CEXes & DeFi should co-exist, & Hyperliquid is building how a significant portion of finance should operate on-chain, the original vision of DeFi - Lots of misunderstandings around HyperEVM as a separate chain when it's really one & the same chain as the exchange (HyperCore) - The purpose of HyperEVM is to make HyperCore programmable, because the most valuable things in finance don't already live on Hyperliquid - Letting people build things themselves is the way to scale (especially aligned with the ethos of crypto), the exception is infra (including infra traditionally viewed as "apps") - There's no other chain you can build on to access DeFi power users transacting billions of dollars a day, & fully integrated with a tier-1 exchange - Tokenized equities, commodities, & forex will all exist on Hyperliquid & there's plenty of talent already building on Hyperliquid to do it - Admits they've been entirely focused on building & poor at comms but that lets others shape your narratives for you - There's so much to build, no point in fighting about drama or stupid takes on twitter - Very short sighted to think in terms of price - in a post Bitcoin era, $HYPE's genesis distribution is as close as we've seen to Bitcoin but being "generous" to users has its cost (large airdrops with 0 cost basis) - Long-term, if a large portion of finance moves onchain, there will be a lot more value accruing than exists in crypto today - On memecoins: values freedom of expression On stablecoins: - the edge is in the real-world (banks, compliance, etc), but one of the strengths of HyperEVM is that it'll allow native stablecoins (including yield-bearing) - Hyperliquid as a platform should be neutral in terms of which stablecoins can be used as collateral (& other features), but very opinionated as to how it makes that possible

The first flash-loan on HyperLend (I'm geeked af for this) > borrows 3.00035 wHYPE via flash-loan from HyperLend > liquidates a position on Sentiment (HYPE borrow and wstHYPE as collateral) > swapped wstHYPE back into wHYPE using KittenSwap > repaid the borrowed wHYPE to HyperLend > Total Profit: 0.97 $HYPE ($12) When flash-loan liquidations through the L1 orderbook?

Yesterday is a good reminder to stay humble, hungry, and focused on what matters: building a better financial system owned by the people. Hyperliquid is not perfect, but it will continue to iterate and grow through the collective efforts of builders, traders, and supporters. Users with JELLY long positions at the time of settlement will be refunded by the Foundation as if their position settled at the closing price of 0.037555. This results in all JELLY traders being settled at a price advantageous to them, except flagged addresses. To recap what happened: A trader self-traded a 4M USDC JELLY position at 0.0095. The price of JELLY then rose more than 4x, with HLP backstop liquidating the 4M position. The short position led to a loss in HLP’s account value. The OI cap formula is a dynamic function of global liquidity and OI on other venues including major CEXs. A 4M USDC position fell within those limits, but additional open interest was prevented from being opened beyond the automatically triggered cap. However, the key issue was that once HLP took over the position, it shared collateral with the other component vaults in the strategy and therefore did not trigger ADL. Risk management on Hyperliquid is being strengthened in various ways, including: + HLP: The Liquidator vault will have a tight cap representing a small percentage of total HLP account value, rebalanced less frequently, and more sophisticated logic around taking backstop liquidations. ADL will be triggered if the Liquidator loses above a certain threshold, instead of moving collateral automatically from the other component vaults. Note that ADL is not expected to trigger during organic market activity. + OI caps: Open interest caps will be refined to be dynamic relative to market cap. + Delistings: Validators will vote onchain to delist assets that fall beneath thresholds. Thank you for your continued feedback, support, and commitment. Hyperliquid

if current trend continues, in 2200s we'd still be using art thefted from 200 years ago to produce images, art as career stopped existing since its not viable its so easy to make art viable as a career by simply sharing a miniscule fraction of these companies profit

Nuberodesign @nuberodesign

a year ago

Since this utter garbage is trending, we should take a look at what Hayao Miyazaki, the founder of Studio Ghibli, said about machine created art.

505 14K 57K 8.4M 7K

CEXs just declared war on a decentralised competitor which has washed away all honour and integrity. They exploit teams constantly asking for multi percent of the supply which they dump on market, multi 6 figure listing fees and a bunch of other terrible optics. If people really do believe in a decentralised future then the CEX monopoly mush be washed away. The Industry deserves something better. Jeff & Co now have my full attention and support after that shady fucking move. Hyperliquid

HL jellyjelly attacker 1) goes max short JELLYJELLY perps (likely while buying spot simultaneously) 2) pumps spot onchain to self-liq perps 3) HLP inherits the giga short JELLYJELLY position 4) continues to pump spot 5) HLP still short, down $6.5m

keyboard_arrow_left Previous keyboard_arrow_right Next

# A new type of information theory this paper is not super well-known but has changed my opinion of how deep learning works more than almost anything else it says that we should measure the amount of information available in some representation based on how *extractable* it is, given finite computation. for example, an encrypted text file has less V-information than the same data in plaintext, because it takes more computation to extract. note the contrast to traditional information theory, which would tell us that the two representations have the same amount of Shannon information i’ve long wondered why certain types of basic questions didn’t have a proper theoretical answer: > why does distillation outperform vanilla maximum likelihood training? > why does lora work better than finetuning? > why does self-attention work better than almost any other similar operation? > how much “information” remains in a text embedding? > how much “information” remains in language model weights? > should i use fine-tuning or RAG? the true answers to all these questions depend on some way of measuring and comparing *information content* between different representations. v-information is one step towards doing this besides computational constraints, model architecture probably affects the “information content” in representations, along with the presence of any pretraining data used, as well as model-level statistics about the optimization in the training process – e.g. the length of time a model was trained for probably changes representations pretty drastically this is all to say, i think there is some true notion of “information” that none of our current paradigms (Shannon information, V-information, etc.) capture. we encounter this idea every day but we dance around it and describe it in vague terms; we measure it from all sorts of angles but can’t quite characterize it theoretically when the v-information paper came out I thought there would be a lot of follow up work developing more complex and useful notions of information for deep learning. but it hasn’t yet I still think at the heart of these questions of what-information-lies-in-representations there’s something to be found that’s profound, elegant, and potentially extremely useful. I don’t know what it is and i’m probably not the person who will figure it out. but I really hope someone does. :)

HyperCore and HyperEVM linking is live on mainnet. HyperCore spot assets are now available as building blocks for applications on HyperEVM. Furthermore, HyperEVM builders now have a permissionless path to listing on the same liquid and performant order books that power billions of volume every day on HyperCore. The full defi lifecycle includes building a project, launching a token, and trading that token. Every step of the journey can be done permissionlessly within Hyperliquid. Defi features composability, self-custody, and no gate keepers. CEXs feature frictionless UX and deep liquidity. Hyperliquid combines both into a seamless and onchain system to house all of finance. See the Docs for details: hyperliquid.gitbook.io/hyperliquid-do…. It is strongly recommended to test everything on testnet before linking on mainnet, as linking is immutable.

Ticker $USDe was bought on Hyperliquid 3 hours ago. This suggests Ethena is going to deploy on Hyperliquid but... did Ethena really bought the ticker? Or was someone else trying to psyop people? Let's check onchain. The ticker was bought by 0x47581b206ffb2752c165b987141bad0dc22bd605. That address was funded by 0x643Dbbc80Bb5432f646fEDFa4091a9b205E87Ae4, which seems to be one of the Ethena Deployers EOAs (can't see more evidence here). 0x643 was funded by 0x796Ee28CE42BF7bd43A09aCd43FAd0aACD7C3b29, which also is a Ethena Deployer, but this time there's more evidence of that being really tied to Ethena since is the signer of Ethena Dev Multisig and had some ENS like etheinainsurancefund and ethenaops. Also it's used as an example on their documentation: docs.ethena.fi/api-documentat… So yeah, Ethena is deploying on Hyperliquid. Also had some curiosity to see if they put more funds on HL and saw that 0xFF0458EA5F864717B05D79BF456a8b70767A7EF5 (which was funded by 0x643) sent funds to HL to set up a 4-8 musig: hypurrscan.io/address/0xFF04… If I had to guess I would say they're waiting until they enable the HyperEVM<>HyperCore tokens link to deploy.

keyboard_arrow_left Previous keyboard_arrow_right Next