flopipop

@kdotcrypto @bulktrade @junbug_sol Yeah it doesn't really matter for the testnet. Just thinking about mainnet, and the fact that this has happened with other apps so don't want it to be the case with Bulk as well...

@ipopflop @bulktrade @junbug_sol will speak with devs just curious, why do you want to use your ledger for a testnet? just singup with an email or hotwallet

In the darkest of times... A hero emerges Join the fight: early.bulk.trade

@kdotcrypto @bulktrade @junbug_sol It isn't. Just tried yesterday and it didn't work... There's a button that allows you to state you're using l*dg3r, but when you click it and try to sign it doesn't work...

@ipopflop @bulktrade @junbug_sol phantom ledger is supported

@WSJmarkets Hyperliquid

The owner of the S&P 500 index is licensing the world’s most tracked stock index for the launch of a derivative contract that trades around the clock on the crypto exchange Hyperliquid on.wsj.com/4bezxK0

@0xSunRun What are your favorite subnets?

There’s also going to be a lot of hate, seethe, cope that enters the zeitgeist as more of CT is forced to pay attention to Bittensor You logged into this world everyday and the golden ticket was staring you in the face for years and you still couldn’t muster up the agency to figure it out

IOW, there's no special sauce in the game design. That's centralized in the subnet owner. The special sauce is the incentives and permissionless participation in the game. Both of which could be achieved by centralized players who could offer orders of magnitudes more in incentives...

I'm not. I'm questioning Bittensor specifically. Why? Because Bittensor is currently incentivizing miners in the order of $1-$5M per year to compete in a certain game in areas where other centralized competitors have raised hundreds of millions or even billions of dollars (and could do something similar). How can this be competitive?

@Old_Samster This can happen in a centralized setting as well. You build a competitive game and allow anyone to participate. This already happens with HackerOne or Immunefi...

@ipopflop key is open collaboration/competition and aligned incentives

Latency doesn't matter if you don't need to adjust parameters within a block. If you do, it does matter. And if you don't it just means there wasn't enough volatility within the block window, and in that case MMs on a CEX wouldn't need to update their liquidity distribution either...

@ipopflop Latency doesn’t matter The point is that on chain liquidity is not subject to 400ms ticks

0/ PropAMMs are one of the most important innovations in market microstructure in years, possibly decades. A thread

@magicdhz @MattTheFinGuy @KyleSamani So what's the current "shred time"? Are there different guarantees between inclusion on a shred vs. block?

@MattTheFinGuy @KyleSamani @ipopflop depends on the scheduler, but most validators run schedulers that include and shred txs in increments that are smaller than the standard slot time so prop AMMs update quotes multiple times a slot

@KyleSamani @threadreaderapp unroll

19/ If propAMMs on Solana mainnet can already quote tighter than all CEXs without these improvements, just imagine how good they will become as these upgrades roll out {fin}

@markjeffrey It isn't even a better product. But even if it were it's an unbelievably tough uphill battle.

Agree. It is not enough to make a better product. You must then MARKET and SELL it better. Compu-history is replete with lesser products winning through sales, marketing and distribution cough Windows 98.

@jbrukh @tplr_ai Still no one seems to have been able to solve decentralized RL afaik, right?

Congrats @tplr_ai — progress in decentralized AI training! A 72B parameter AI model was trained here on a decentralized network, by using aggressive communication compression between nodes (SparseLoCo). In this setup, nodes are still quite powerful GPU stacks (8 Blackwells) but you can connect the dots on where this is going. Very soon all kinds of GPUs will be training models in swarms.

@ttrofel wen token ser?

We have been cooking a lot at Exponent behind the scenes over the past months, with many new products tailored for what’s next in the industry. Very excited to finally start sharing more in the coming weeks, starting with a new look for Exponent Time to build a generational company

@lui1of1 wen token brother?

2026 is going to be an insane year for Exponent So much cooking over the next month And it starts with a brand refresh that captures what we're all about Exciting times!

@markjeffrey What's the case for Targon. It has no competitive advantage against any hyperscaler which offers essentially the same product (in hardware terms) with better assurances (and certifications) for their users...

Proveably secure inference -- on Bittensor:

@jbrukh Are there even any useful "web3 agents" out there? Virtuals is all garbage...

I only care about one thing, and it's disgusting. On-chain metrics for web3 agents.

@jarxiao Lol. Yet users cannot log in using a HW wallet... Get the basics of the basics first before pontificating about anything brother.

No user will ever care, but building fully on-chain perps on Solana is really technically challenging. I respect all the teams who have attempted this Herculean task in the past: Drift, Zeta, Mango, Jupiter to name a few. Many of these teams have found success with real users, but I think it’s fair to say that all of them had to make product tradeoffs as a direct result of Solana’s programming model. We achieved: - Price time priority on user limit orders with built-in margin - Synchronous matching (no crank) - Lazily accumulating funding payments - Low compute quote updates for makers - Platform-wide cross and isolated margin This required innovation on the core system architecture. I usually don’t like talking about this because, in my eyes, these properties are table stakes for being competitive with any centralized exchange. Our users won’t care about the sick tech, but they will care about how good the product feels. We moved mountains to build @PhoenixTrade. No on-chain vault + off-chain matching. No compromises on microstructure. No vapor. All of the program code will be made source-available at general launch.