Sabitlenmiş Tweet
Wow, @ton_blockchain himself responded to my post😅
Celsius@celsius_135
Gn Do you use a sleep mask at night?🌙
English
Celsius
1.1K posts
Celsius
@celsius_135
Ambassador @1inch and @ston_fi Web3 Explorer | Blockchain Influencer 🌍 Your crypto info source. 📈 Helps projects scale & grow community. 🤝 Open to collabs.
Alpha Calls Katılım Aralık 2024
593 Takip Edilen866 Takipçiler
@Kozh_Crypto @Gradient_HQ Honestly, this is the first time I've heard of DSD, but it sounds like magic. I used to think that all these delays were just server glitches. Now I understand that there is a whole "guess-check" system working. The main thing is that the quality does not suffer at such a speed.
English
./ u all understand how large AI models like ChatGPT and others work, right?
they all have one drawback: waiting for the server to send the next word, since everything works over the network, which makes it slow.
from what I understand, dsd is a clever way to speed up this process by about 5 times, using "guess and check" instead of the usual
> "generate word → wait → generate next word."
- draft locally (the guess) - a small, fast model on the device generates several words without waiting for the network.
- confirm remotely (the check) - a large model doesn't rewrite the text, but checks the draft. It says "yes" or "no."
- adaptive logic - the system automatically adjusts the speed and rigor of the check depending on the topic:
> creativity → draft freedom
> precise topics → strict check.
dsd - thanks to it, the text is produced several times faster, but almost as accurately.
Parallax@tryParallax
we sped up distributed inference by up to 5x with decentralized speculative decoding. many don't realize that AI models normally generate text one single word at a time, waiting for the network after every word. speculative decoding changes this by using a "guess & confirm" system, similar to autocomplete. how it's done: 1. draft locally (the guess) instead of waiting for the network, a tiny, fast model on your device guesses the next few words instantly, without waiting for the network. 2. confirm remotely (the check) the massive remote model doesn't generate from scratch; it just checks the draft. it looks at the guesses in a batch and says "yes, yes, no." you get multiple words in the time it usually takes to get one. 3. adaptive logic dsd is smart. if the topic is creative, it lets the draft flow loose. if the topic is math or code, it checks more strictly. it balances speed and precision automatically so your inference almost feel instant. find out more: paper: arxiv.org/abs/2511.11733 blog: gradient.network/research/turni…
English
@GemResearch1 @PerceptronNTWK This is exactly what many people forget about at this noisy AI party. Everyone is chasing the architecture of the models, and the foundation - the data - often remains behind the scenes. If the guys really solve this issue, it will be a breakthrough.
English
While everyone's talking about AI models @PerceptronNTWK is quietly building something about the data these models need.
Real users, real traffic, real contributions to AI.
Not hype for the sake of hype but a clear mechanic. It will be interesting to see what the project will lead to.
#PerceptronNTWK #MindoAI #DeFI
English
Yesterday JBond made a list of the richest people in the world
But I'm going to repeat this more explicitly with crypto people
1. 🇦🇪 Changpeng Zhao CZ ~ $80B
2. 🇺🇸 Jed McCaleb ~ $27B
3. 🇮🇹 Giancarlo Devasini ~ $22B
4. 🇺🇸 Chris Larsen ~ $15B
5. 🇺🇸 Brian Armstrong ~ $12,5B
6. 🇨🇳 Li Xiaolai ~ $11,5B
7. 🇺🇸 Michael Saylor ~ $10B
8. 🇮🇹 Paolo Ardoino ~ $9,5B
9. 🇳🇱 Jean Louis van der Velde ~ $9,5B
10. 🇬🇩 Justin San ~ $8,5-12B
11. 🇺🇸 Fred Ehsram ~ $4,5B
12. 🇺🇸 Tyler Winklevoss ~ $3,5B
13. 🇺🇸 Cameron Winklevoss ~ $3,5B
14. 🇺🇸 Mike Novogratz ~ $3B
15. 🇺🇸 Matthew Roszak ~ $3B
16. 🇦🇺 Bijan Tehrani ~ $2,8B
17. 🇦🇺 Ed Craven ~ $2,5B
18. 🇺🇸 Tim Drapper ~ $2,2B
19. 🇨🇳 Jihan Wu ~ $2B
20. 🇨🇦 Vitalik Buterin ~ $1.2-2B
21. 🇪🇪 Rain Lohmus ~ $1,1B
22. 🇨🇦 Joseph Lubin ~ $1B
23. 🇮🇳 Sandeep Nailwal ~ $1B
24. 🇺🇸 Dan Morehead ~ $1B
25. 🇺🇸 Hayden Adams ~ $800M
26. 🇫🇮 Stani Kulechov ~ $700M
27. 🇬🇧 Gavin Wood ~ $500M
28. 🇰🇷 Do Kwon ~ $500M
29. 🇮🇹 Silvio Micali ~ $400M
30. 🇸🇬 Zhu Su ~ $300M
Now we have to work harder, friends, to be on this list in the future
JBond@jbondwagon
Here’s the Top 20 Richest People in the World. Number 14 will shock you 👀🤔 1. 🇺🇸 Elon Musk — $716.2B 2. 🇺🇸 Larry Page — $266.5B 3. 🇺🇸 Jeff Bezos — $256.1B 4. 🇺🇸 Sergey Brin — $245.9B 5. 🇺🇸 Larry Ellison — $238.2B 6. 🇺🇸 Mark Zuckerberg — $221.5B 7. 🇫🇷 Arnault family — $190.9B 8. 🇺🇸 Jensen Huang — $160.6B 9. 🇺🇸 Warren Buffett — $148.0B 10. 🇪🇸 Amancio Ortega — $146.2B 11. 🇺🇸 Steve Ballmer — $145.8B 12. 🇺🇸 Michael Dell — $134.2B 13. 🇺🇸 Rob Walton & family — $132.4B 14. 🏳️ Satoshi Nakamoto — $130B 15. 🇺🇸 Jim Walton & family — $129.7B 16. 🇺🇸 Alice Walton — $120.8B 17. 🇺🇸 Michael Bloomberg — $109.4B 18. 🇮🇳 Mukesh Ambani — $106.4B 19. 🇺🇸 Bill Gates — $104.3B 20. 🇲🇽 Carlos Slim Helú & family — $102.1B
English
@Planet_Founder @PolymarketTrade @Polymarket Yeah, from 13 million to 120 in half a day 😂
No, seriously, unless CZ himself posts with the hashtag #RANGER. Without this, the niche is narrow, there is no such money. I'm going to bet on NO, thanks for the tip.
English
Be extremely careful 🚨
Right now, the token sale for the RANGER project is ongoing on META DAO.
polymarket.com/event/total-co…
There are only 5 hours left until the end of the sale, and currently around $13–13.5 million has been committed (according to the official page). The final amount will most likely be significantly higher than $50–$60 million.
→ But will it exceed $120 million???
I’m not an expert in MetaDAO sales — the last sale raised approximately $100 million.
I find it hard to imagine that the amount will increase 9 times in just the remaining 5 hours.
That’s why my bet is:
“Total commitments for the Ranger public sale on MetaDAO >$120 mln” → NO!!
DYOR 🎰
English
@6mirra6 @magicblock @MagicIntern @fauxfire_ It's cool that DX doesn't suffer. Usually, a state compression is followed by a bunch of hacks, but if it works transparently for devs, then it's a game changer. Magicblock did a great job of extending this into the ER.
English
✦ Hidden State ✦
One of the most underrated shifts happening right now is how Light Protocol compressed accounts are starting to live inside Ephemeral Rollups.
After @magicblock December network upgrade, this quietly unlocked something important: compressed accounts can now be used inside ERs almost the same way as regular Solana accounts.
The result is simple, but powerful:
state storage becomes drastically cheaper,
developer experience stays clean,
execution remains gasless and low-latency,
and everything still settles back to Solana seamlessly.
This matters a lot for real-time systems - games with large inventories, fast-moving positions, social data, autonomous agents.
It’s not loud, it’s not flashy, and it’s easy to miss while everyone’s watching the more visible launches.
But this kind of infrastructure change is usually what shapes the next wave of on-chain apps.
English
@SergeyP77 @CodeXero_xyz the picture clearly shows a typical chill guy in crypto
English
Now we can relax and enjoy life.
After all, CodeXero @CodeXero_xyz makes it easy to implement our ideas for building dApps.
No complicated settings. No technical knowledge required.
Just the bright glow of a pure idea brought to life.
English
@bubonik_X @tread_fi Point 3 caught my attention
Entering 50k into a shitcoin without slippage is a dream, not a reality. If it really saves you from pump and dump when you enter, then the service is really worth using
English
🚨🚨Alpha Alert is here: How to trade like a fund, even if you only have $1,000 in your pocket 🧵
I finally got to thoroughly test @tread_fi and it's not just another terminal. It's an OEMS (order management system) that gives retail traders Tier-1 fund-level tools.
The main killer features that will change your approach:
1. Delta-neutral bot (Basis Trading) 📈
Forget about guessing on charts. The bot automatically opens Cash-and-Carry trades (spot + short futures). You are fully protected from price drops, and your profit is fat funding. In the current market, this is a stable 30-60% APR on full autopilot.
2. Perfect airdrop farming (MM Bot) 🚜
Want points from @pacifica_fi @paradex @nadoHQ or other DEXs? The Market Maker bot generates millions in volume by setting limits on both sides. You provide liquidity, spend almost nothing on commissions (only Maker orders!), and soar up the leaderboards.
3. Smart execution (TWAP/VWAP) ⚡️
Need to enter a $50k position in an illiquid shittcoin? Don't hit the order book! Use algo orders. The terminal will carefully split the position so you don't get caught in a 5% slippage.
4. Unified hub (CEX + DEX) 🌐
Binance, Bybit, OKX, Hyperliquid - all in one window. No more 20 open tabs and confusion with API keys.
Want to join? Use this link:
app.tread.fi/referral/V0PEF…
English
@GengarSH @PiSquared Let's say, but how will it work in real conditions under high load? It sounds interesting in theory, but I would like to see live examples of transactions.
English
Machine speed in @PiSquared is simply how the system operates under load.
Events are recorded and verified in tens of milliseconds. There is no waiting for a global state, no queues, no confirmations later. Each action is processed separately and immediately.
Nodes do not line up into a single order and do not wait for each other. Each node verifies what it is responsible for and records the result. Because of this, the system does not slow down when the number of operations grows. Parallelism is the base mode.
This approach is designed for scenarios where actions happen continuously. Requests, computations, data exchange, payments between processes. All of this happens automatically, without pauses or manual steps.
Machine speed here is a mode of operation where the system simply does not stop and does not switch into waiting.
Learn more: pi2.network
@ShannonSheriff | @LindsayCasale
English
@0x_aduka @inference_labs well, if I come across any interesting project again, I will immediately notify about it
English
I recently came across @inference_labs, which is now in its second week of earning points. There will be only 5 seasons of 14 days each. It is not yet clear how good the project is, but it is interesting that 0.5% has been allocated for Kaito users yaps.kaito.ai/inferencelabs.
English
@ceooflexus Honestly, if you expect a positive, comforting answer from me now, then maybe I will say no. Or rather, I'll say, I don't know. Because now everything has really changed, competition, X algorithms and so on. So now it's hard to say anything about all this..
English
@celsius_135 Im going back also ! It’s actually very hard to start from 0 but we will try our best , right ?
English
@Sadnoboy616 no, I'm a wolf
but honestly, I didn't understand the point of the question..
English
@MorphMoon you're right, real life should always come first, no matter what
English
@SergeyP77 i understand you, over the past 2 months, watching all this from the outside, I understand that many people have a similar problem now
English
@eduard_ko85716 @CodeXero_xyz if it is still in development, how long do we have to wait for its release?
English
gXero legends
It’s actually really cool to watch how @CodeXero_xyz
is developing. I’m talking more about the people who are using the product (like me) and are already creating their own dApps
Haven’t tried it?
These few weeks spent with CodeXero changed my perception of coding and vibe coding. And even at such an early stage I can see that people are using it, that it already works, that it already delivers results.
English
@SergeyP77 @alturax so, you're saying that Altura is significantly better than its competitors in the form of other similar projects?
English
It's 2026, and it's not about taking a break, but about continuing to do your thing.
I continue to follow and talk about Altura @alturax, because it is one of the few DeFi projects where everything fits together logically.
A few key things that are important to me at Altura:
• Real base yield, not figures from issues
• Full on-chain transparency: NAV, PPS, strategies
• Clear rules for users, no surprises
• An active team that constantly explains what works and how it works
The beginning of 2026 is the time to choose the products we will continue to work with.
Altura looks like just such a case.
GIF
English
I'll start the year with a post that I'll give a quote to in a year
My infofi statistics for 2 months of yapping / quaking
Infofi will live on guys, believe me, and it will reward quite a few people
Make a bookmarker of this post and compare my work in a year
Madi@madiweb3
GM CT ☕️ Last year was very interesting and eventful which we will not forget In turn, the new year will be several times better What are you plans in 1st day 2026? Today, the results of the nominations, where you identified the winner > continue study @0xPolygon > publish winners in nominations > new podcast with new guest > make smart replies ( it’s important) Thank you all so much for your support last year Now we have to transfer the same energy in the new year in 2x size Love you my family
English
@RektBarrister @magicblock well, I imagined Solana in turbo boots, I think they would be perfect
English
Why it matters & who uses it
@magicblock Edition
Why “no fragmentation” is a big deal
Other L2s are like building separate mini-cities with their own roads, money, and doors between them. Money and apps get split up, and moving between them is annoying.
MagicBlock stays inside the same big city (Solana). Tokens and NFTs never leave, and all the games and DeFi apps can still talk to each other easily.
For devs, it’s almost the same code – they just add a couple of extra “do it faster” calls. For users, it feels like normal Solana, just way smoother.
Also:
You don’t have to switch networks in your wallet.
MagicBlock’s “traffic controller” sends each transaction either to normal Solana or to the fast track (Ephemeral Rollup) for you.
Who is already using it:
Rush: A game where you bet on Bitcoin’s price every 10 seconds. It’s gasless and super fast, so it feels like a real-time trading game, not a slow dApp.
Flash Trade: A trading platform that uses the fast lanes for lots of quick trades with low cost.
Supersize: A fully on-chain multiplayer game that still feels snappy, because it runs with very low delay.
dTelecom: Voice and video calls that run in a decentralized way, but stay smooth thanks to low latency.
MagicBlock is like giving Solana turbo shoes - you still walk in the same world, with the same items and friends, but everything you do that needs speed suddenly feels instant
Alex@RektBarrister
Ephemeral Rollups ft. @magicblock Think of Solana as a big city, and Ephemeral Rollups as temporary race tracks you can open and close inside that city. How it works Delegation (opening the track): Your app says, “For this mini-game/trading session, use this special track,” and hands a few accounts over to the ER for a while. High‑speed mode: On that track, everything runs super fast and smooth, with almost no waiting – like cars racing on a closed circuit instead of busy city streets. Undelegation (closing the track): When the session ends, the ER packs up all the results and writes them back into the main city map (Solana L1) so everyone agrees on the final state. Cool powers No gas during the race: Users don’t pay fees while they’re playing on the track. Faster ticks: Games can update way faster than normal Solana block times. More tracks when needed: If lots of people show up, you can open more tracks in parallel so nothing jams. Extra tools: You can plug in things like instant randomness or super-fast price feeds to make games and trading feel real-time. Why it’s better than normal L2s Other L2s are like building a separate mini-city with its own money and bridges; moving between them is messy and splits people and assets. MagicBlock’s ERs stay inside the same city (Solana): Tokens and NFTs never leave Solana. Devs mostly just add a couple of extra calls instead of rebuilding everything. For users, it just feels like: “Same wallet, same network, but suddenly everything is way faster and smoother.”
English
English
🧑🎄Happy New Year 2026, @SeismicSys fam!
So grateful to be part of this amazing community, excited to see what the new year has in store for us.
Thank you for this amazing artwork as a gift for the holidays❤️Magnitude is the key
@SeismicsysRU
Seismic Russia 🇷🇺@SeismicsysRU
Поздравляем всю нашу @SeismicSys семью с наступающими праздниками 🧑🎄🎊🌲 Впереди нас ждут потрясающие события, 2026 будет просто 🚀 Спасибо нашим невероятным художникам за этот потрясающий, совместный сюрприз для нашего комьюнити. Это просто ванлав 🩷🩷🩷
English