long degeneracy , hype WILL win , ยถ
Hyperliquid
Joined October 2018
- Tweets1.8K
- Following1.9K
- Followers479
- Likes3.4K
bliquidc.hl ๐ retweeted
as papertrade reaches its first month live, what started as a whitepaper continues to be tested at scale in production. today's update is another step towards what i think is the house of the future. i want to extend my sincerest thanks to everyone whos participation has helped make this new primitive possible.
while we continue to take each day one at a time, incentivized composable liquidity is a very exciting design arena. some things we're working on are probably obvious - like the option to pay gas manually to avoid onchain congestion. others well just put out when they're ready.
good luck to us.
Five contract level upgrades have been pushed to the timelock. The non-atomic nature of hyperliquid core and EVM blocks, along with the still broadly un-battletested precompiles necessitates extensive, heavy testing as the design patterns papertrade employs are unusual to see in solidity. Papertrade, by its nature is a risk on protocol. But when it comes to striking a balance between speed and security, security will always remain paramount, particularly when it comes to production contract changes.
Core lending:
The Exchange will now be able to supply idle USDC to HyperCore's native lending market, with the accruing interest being routed to stakers.
Collectibles:
A factory plus per-collection ERC-721 contracts for exchange collectibles and themes.
Staked PAPER transfer module:
PaperStaking gains a whitelist of transfer enabled contracts, along with contract level yield piping. This code is unused for now but will be necessary for future updates.
Sweep timing fix:
HyperEVM big blocks share a timestamp with a preceding small block. Whilst never triggered in production this could cause an ill-timed deposit sweep to fail and so the contract level timing checks are switching to block numbers rather than timestamps.
Direct LP donation:
A new admin locked donateToLp call moves USDC from a designated donor account into the LP's cash, providing a clean path for regulating externally held USDC back into the LP.
At some point we will likely write up a blog on our retrospective of the first month of papertrade, but we are few and there are only so many hours in a day or hats a man can wear.
A final quick thank you to those who helped out with adversarial testing, game theory and the general math that went into last month's upgrade, you know who you are. To those who are still confused -- often boisterously so, we thank you for sticking around for long enough to show you care.
good luck to us.
bliquidc.hl ๐ retweeted
Pearl doesn't mine hashes. It mines matrix multiplication.
That's the same math every LLM uses to train and to answer a prompt, per Evan Van Ness on The Rollup.
"Pearl is mining via arbitrary matrix multiplication, which is what every AI model, every LLM, every machine learningโฆ uses."
The pitch: turn the work GPUs already do for AI into the work that secures a chain.
FT @evan_van_ness @prlnet @therollupco @andyyy
bliquidc.hl ๐ retweeted
@HyperliquidX perps are live on @Bitso.
Bitso users can trade 200+ Hyperliquid perpetual markets from the Bitso Alpha interface, tapping into Hyperliquid and funded straight from their Bitso account.
Crypto, the S&P 500, gold and FX. Long or short, up to 50x depending on the market. 24/7, no expiry, and every order, fill and liquidation verifiable on Hyperliquid's onchain order book.
Live today.
#BitsoPerps
bliquidc.hl ๐ retweeted
what if on:
> robinhood chain you could access robinhood options/spot liquidity
> binance chain you could access binance perps/spot liquidity
> base chain you could access coinbase perp/spot liquidity
the above is fundamentally, technically, and legally impossible, therefore:
On Elysium, being built on top of Hyperliquid, fixes this entirely
bliquidc.hl ๐ retweeted
I explain why Pearl @prlnet is the only GPU-mined coin with a real shot at outgrowing Bitcoin
"Overhead is the entire ballgame"
"There are plenty of other GPU mined coins, but typically what's happening is that your GPU is using all of its computing resources to mine, and that's still the only revenue stream. What Pearl is building here is making GPUs more efficient"
"What separates Pearl from other GPU mine coins is that it's still the only one that has a shot at getting adopted by AI labs and inference clusters. Do any of those GPU mine coins have an actual shot at building out a larger electrical footprint than Bitcoin? No way"
"If they actually get penetration into the largest CapEx build out in human history and the fastest growing industry in human history, then yeah, it has actually a credible shot at becoming a larger network than Bitcoin. And that's why it's an interesting investment even without the useful work"
Why Pearl will Win
โโโโโโโโโโโโโโโโโโ
01:14 BTC is dying
03:24 Breaking up with my highschool girlfriend
06:51 Why it can work
09:35 Pearl's power footprint
12:45 AI capex > asic capex
15:00 FP8
19:00 Negative overhead
22:04 Elite team
23:06 Hashrate = hinge profile height
28:01 AI lab adoption
30:48 I want my MTV
32:31 Mark Knopfler shreds on guitar
35:52 Useful work vs Duplex work
38:32 Cheap verification is required for money
41:20 Can Pearl close the loop?
45:17 Addressing fundraise FUD
51:21 Risks
1:06:27 Pearl wins by becoming multi-denominational religion
bliquidc.hl ๐ retweeted
Elysium mainnet is now live.
elysium.kinetiq.xyz
bliquidc.hl ๐ retweeted
what if there's only one option?
teapawt.net/therollupco/status/211โฆ
NEW: Evan Van Ness On Why Pearl Could Be Bigger Than Bitcoin
Evan breaks down how useful-work mining could change data center economics, the flywheel behind cheaper inference, and why he believes the biggest opportunity is still ahead.
@evan_van_ness @andyyy @robbieklages
00:00 Intro
01:21 Pearl Could Be Bigger Than Bitcoin
03:47 Bitcoin's Ossification And Security Budget Problem
06:36 Everyone Wanted Proof Of Work Useful
08:51 Pearl Mines Through Matrix Multiplication
11:00 Near Zero Overhead Mining Alongside Inference
15:41 Pearl's Useful Work Is Under 1%
17:49 Everything Useless Must Become Useful Work
20:20 A Pearl Flywheel For Cheaper Inference
24:21 Midterms Threaten Data Center Buildouts
27:40 Frontier Labs Already Know About Pearl
31:00 Pearl At 3.7B FDV Versus Bitcoin
35:20 A Pearl Fork Aimed At Payments
bliquidc.hl ๐ retweeted
Ansem shares his long-term thesis on Pearl ($PEARL), dubbed as "AI Bitcoin," and why it could hit escape velocity
โPearl high risk for sure, but what they're trying to do is be a form of hard money in the same way that Zcash ($ZEC) is trying to be a form of hard money.โ
โBut their entire thesis is that instead of just doing proof of work where the work that you're doing isn't worth anything, it's just electricity cost, compute cost.โ
"The compute that you use to mine Pearl can also be used for all the AI inference matrix multiplications that you need whenever you're doing inference.โ
โThey call it 'useful work' when you're mining Pearl, where you can also make money by serving that inference to people.โ
โAnd that does make sense as a thesis if you think how much compute we're going to be using over time is going to continue to go up.โ
โPearl could hit escape velocity as being a hard form of money, 'AI Bitcoin' is what they call it.โ
bliquidc.hl ๐ retweeted
Since this post: the FP8 upgrade's proof code is now on its branch.
On Nov 3 and 4 @prlnet merged three pull requests into fp8-scheme, the branch that carries the upgrade: a cap on how much work a verifier can be asked to do, a smaller verifier circuit (2^18 to 2^17 gates, per its author), and simpler tile selection.
It is the first change to that branch since Oct 5. Three steps remain before a fork: the branch merged into the main code, a pearl-2:native mainnet block height, and a release.
โWhen FP8?โ comes up in most pearl-2:native threads.
Where it stands on Oct 28:
The FP8 upgrade (certificate V4, a hard fork) sits in pull request #311, open since Sep 9, with no mainnet block height in the code.
On Sep 25 @prlnet CEO said it was on testnet and expected to launch โwithin weeksโ. Work on its proof system is still in progress, with prover commits as recent as Oct 25. The signal to watch is a fork height for mainnet.
bliquidc.hl ๐ retweeted
RAPIDFIRE $PEARL FUD DISPULSION
VC FUNDRAISE SKEM?
Pearl Research Labs is an independent lab developing Pearl that raised funding. No Pearl tokens were purchased by VCs; the only way to acquire pearl is to mine or buy from a miner. There is no token<>equity conflict, and the Labs entity is well capitalized to build on top of the network.
PEARL WILL DUMP FROM GIGA EMISSIONS
There are no vesting schedules or free Pearl tokens to be dumped on market. BTC / ZEC / XMR have signaled credible interest in PoW tokens, and Pearl needs not go through the same consolidation cycle. Do you think the market will give everyone a free entry when emissions cool down, or price in what the network is worth in the future?
ATTACKERS CAN RENT GPUS AND 51% ATTACK PEARL
There are not enough gpus for rent on the market to do this. dpmo
EARLY PEARL MINING
Pearl had a reduced mining difficulty period early on to emit more Pearl into circulation. This allowed early adopter miners and believers to secure meaningful supply. Increasing emissions protected against 51% attacks in the first week before they became infeasible.
HYPEROPTIMIZED โPEARL GPUSโ WILL OUTPERFORM DUPLEX MINERS
Compute is already being hyperoptimized for inference in the same way ASICs were optimized to mine BTC. Duplex miners (compute serving inference AND mining Pearl) are incentivized to adopt the same GPU optimization; Pearl is on the bleeding edge with labs like Flapping Airplanes to do exactly this.
bliquidc.hl ๐ retweeted
๐ง๐ต๐ฒ $๐ฒ.๐ฏ๐ ๐๐ถ๐ป๐ฎ๐ป๐ฐ๐ฒ ๐ต๐ฎ๐ฐ๐ธ ๐ฑ๐ถ๐ฑ๐ป'๐ ๐๐๐ฎ๐ฟ๐ ๐๐ถ๐๐ต ๐ฎ ๐ด๐ฒ๐ป๐ถ๐๐ ๐ฒ๐
๐ฝ๐น๐ผ๐ถ๐.
It started with a job interview and a "coding assignment." 5 rules could've stopped it ๐
๐ฃ ๐๐ข๐ช ๐ง๐๐๐ฌ ๐๐ข๐ง ๐๐ก
1. Fake recruiter reached a Binance engineer on a legit job platform
2. Two interview rounds. Video call with "HR", then a chat with the "CEO" of a company that doesn't exist
3. They even shared personal social accounts to build trust
4. Round 2: "download this assignment"
5. It swapped a real dev tool (pyenv) for malware. Antivirus didn't catch it
6. From there they used an internal AI tool to reach 2 more engineers on the wallet team
๐ก๏ธ ๐ ๐ฌ ๐ฑ ๐ฅ๐จ๐๐๐ฆ ๐๐ ๐ฌ๐ข๐จ ๐ช๐ข๐ฅ๐ ๐๐ก ๐๐ฅ๐ฌ๐ฃ๐ง๐ข
1. Never run an interview task on your work laptop. Ever
2. Run any take-home repo in a throwaway VM or cloud box, not your main machine
3. Read the setup scripts before you run them. Install steps are where the payload hides
4. Verify the company exists outside the chat: real site, real people, real employees on LinkedIn
5. Pushed to "just run it quick" before the call ends? That's the tell. Walk
โ ๏ธ ๐ฅ๐๐ ๐๐ ๐๐๐ฅ
Binance says it blocked 57 other suspected North Korea-linked attempts this year. Fake interviews, fake dev tools, bad packages, browser extensions.
If you list a crypto job on your profile, you're a target.
Send this to the dev friend who's job hunting right now ๐ซก
bliquidc.hl ๐ retweeted
@AnthropicAIโs impressive work on FlashProve is a real signal for the value of @prlnetโs technology in the toolbox of provable inference (PI).
FlashProve proves inference correctness post-hoc. Moving it to real-time proof-of-inference requires time-stamping, which is what blockchains were made for..
The reality is that AI verification often cares about what a datacenter DIDNโT run. Correct inference proofs donโt rule out parallel, unmonitored workloads. @prlnetโs Pearl-GEMM protocol anchors the output to computation (FLOPs). Paired with PI, it could help bound capacity available for unauthorized training, and provide a โmodel-tracingโ audit for agentic workloads.
Pearl is uniquely positioned to build the work-accounting layer of PI infrastructure.
anthropic.com/research/provaโฆ
bliquidc.hl ๐ retweeted
Big news from @AnthropicAI on verifiable LLM inference, with an exciting connection to Pearlโs proof of useful work.
Two orthogonal questions:
โข Did the model you paid for actually produce this response?
โข Was computational work performed?
Anthropic reports that generating inference proofs currently takes ~70โ95ร the GPU time of ordinary inference. Pearlโs PoUW, on the contrary, adds negligible overhead. These costs buy different guarantees. Proving correct inference doesnโt automatically establish proof of work and the other way around. But the two are deeply complementary. Verifiable inference could certify which model produced an answer, while Pearlืณs PoUW only certifies that a computation has been done.
As inference proofs get more practical, that combination could make AI easier to trust and more economical to provide.
