星旭 | AI • Web3 • Blockchain • Trading • DeFi • News
Joined August 2020
- Tweets15.6K
- Following1K
- Followers18.1K
- Likes16.1K
Solana’s biggest network change since 2020 is live. Here’s how users benefit
@solana switched on its Alpenglow upgrade Wednesday, introducing a new voting system designed to make payments and trades final in a fraction of a second.
The upgrade targets about 150 milliseconds before the network considers a transaction locked in, replacing the multi-second wait under its previous system. That point is called finality, when an application can treat the payment as settled.
Previously, they recorded votes as transactions on the blockchain and accumulated approvals across successive blocks.
Alpenglow lets them send votes directly to one another and combine those approvals into certificates, allowing a block to become final after one or two voting rounds.
The handover from the old system took about 900 milliseconds, according to the Solana Foundation. Votes from 446 of the network’s 654 participating validators certified the block used as the starting point — a timing that measures the switch between systems.
How this matters for users
Many Solana apps already showed a transaction as confirmed before the network had finished locking it in, which made payments feel fast while leaving a further wait for full settlement.
Alpenglow actually makes those transactions faster, giving apps a way to show users their money has arrived with stronger assurance that the payment will hold. Such a benefit is especially useful when a service needs to release funds or goods immediately after receiving payment.
An exchange could credit a deposit sooner, letting a customer trade with the funds. A shop accepting a stablecoin payment could confirm it without keeping the customer at the counter waiting for settlement.
How quickly that reaches a user’s screen also depends on the service handling the payment. Anza, the company behind Solana’s Agave software, told CoinDesk in a release that some computers receive finality information later than the validators voting directly.
The 150-millisecond figure remains a target under normal network conditions, rather than a guaranteed time for every payment.
Users can continue sending funds through their existing wallets. Transaction fees and the way applications run remain unchanged.
The speed comes from changing how validators, the computers checking transactions and agreeing on the network’s records, communicate.
Meanwhile, there is also a change readers may notice on network dashboards. Validator votes no longer appear as blockchain transactions, so services that counted them will report lower transaction totals.
A chart could therefore show a sharp drop even while users continue making just as many payments and trades. Comparisons with earlier activity need to count user transactions separately from the votes that previously filled the blocks.
Solana to triple transaction size as apps get room for more complex trades
The Solana smart contract blockchain is targeting Wednesday to increase the maximum transaction size from 1,232 bytes to 4,096 bytes, giving developers more than three times as much room to fit instructions into a single operation.
Operations that previously had to be split across several transactions can increasingly be handled in a single transaction under Transaction v1. That includes large cryptographic proofs, payments requiring many approvals and some confidential transfers.
The new format is already running on Solana's test and development networks. Existing transaction formats will continue to work, so wallets and apps do not have to switch to v1 unless they need the extra space.
It could help Solana sharpen its edge over its rival Ethereum. While Solana has always been faster and cheaper, it previously fell behind in one key area: its transactions were strictly hard-capped at 1,232 bytes. Ethereum, by contrast, has no rigid protocol size limit, allowing developers to execute massive, data-dense applications in a single operation simply by paying a higher transaction fee.
By more than tripling this ceiling to 4,096 bytes, Solana is looking to eliminate this structural bottleneck.
The bigger change is for software that reads Solana. Services that fetch blocks and transactions need to be updated to recognize v1. If they are not, requests can fail when they encounter the new format.
Some software also needs to change where it looks for priority-fee information. A priority fee is an optional extra payment that users can make to have a transaction processed faster. V1 stores that information elsewhere, so outdated software can show a fee of zero even when one was paid.
Wallets, explorers, and trading apps often rely on these services to display what happened onchain, meaning that bad data behind the scenes can turn into bad data on a user's screen.
Larger transactions also use more network bandwidth. Developers expect users may need to offer higher-priority fees when larger transactions are competing for space, although the upgrade does not introduce a new fee per byte.
The old 1,232-byte limit comes from Solana's original networking design, in which each transaction had to fit within a roughly 1,280-byte internet data packet. Solana changed how transaction traffic is transmitted in 2022, making that old ceiling less necessary.
The new limit stops at 4,096 bytes because four kilobytes is a standard memory-page size used by validator hardware. Going much larger could force one transaction to span several pages and make it more expensive for the machines running Solana to process.
The change is defined in two proposals, SIMD-0296 and SIMD-0385, co-authored by Jacob Creech and Andrew Fitzgerald, and is wholly separate from the recent Solana governance votes that changed how quickly new SOL was created and removed from supply through fees earned from network activities.
Binance buys $100 million Circle stake in five-year USDC promotion deal
@binance bought $100 million of Circle shares as the companies signed a five-year deal that pays the crypto exchange to promote the USDC stablecoin on its platform.
Circle issued Binance 1.24 million Class A shares at $80.84 each in a private placement that closed Sept. 17, according to an SEC filing published Tuesday. The price reflected a discount to Circle’s market value before the sale, the company said.
Binance cannot sell, transfer or hedge the shares until the earlier of two years or a termination of the commercial arrangements by Binance under certain circumstances, subject to customary exceptions, but retains the right to vote them.
The equity purchase closed alongside an expansion of the companies’ existing USDC partnership. Circle agreed to pay Binance a monthly incentive fee calculated as a percentage of the USDC held through Circle’s Modular Smart Contract Wallet service, while Binance will carry out promotional activities for the stablecoin.
CoinDesk reached out to Binance for a comment on the matter.
The new arrangement replaces Circle and Binance’s earlier USDC deals from November 2024 and August 2025. Either company can end the five-year partnership early if specified events occur. (the actual termination figures aren’t public)
The shares were sold in an unregistered private placement, which limits Binance's ability to resell them unless they are registered or an applicable exemption applies. Circle’s subsidiaries signed the related commercial agreements, and the share sale closed immediately afterward.
We've collected an enormous amount of digital data for AI, but the physical world is a very different source of information.
That is one of the problems @vangrid_io is building around. Physical AI needs data that describes the real world, not just information that exists on a screen.
A robot operating in a physical environment needs to understand more than what an object looks like. Spatial data can describe its location, geometry, depth, the objects around it and the relationships between them. These are the kinds of details that help represent a physical environment.
Vangrid's approach is to use everyday devices to capture real world environments, process those captures into 3D spatial data, and verify the resulting data with provenance mechanisms. The idea is to create a distributed source of real world ground truth that can support Physical AI systems.
I think this raises an interesting question as the industry develops. If more AI systems start operating in the physical world, will access to reliable spatial data become as important as improving the models themselves?
It is still too early to know, but this is probably one of the areas worth watching as Physical AI moves forward.
🤖 Made with AI
I've been seeing more projects talking about Physical AI lately, and one thing keeps coming up: data.
We've already had years to collect digital data for AI. Text, images, videos, code, documents... there is an enormous amount of it online.
The physical world is different.
If you want an AI system to operate inside a warehouse, for example, knowing that the warehouse exists isn't enough. It needs to understand the actual environment: the layout, objects, pathways, entrances, obstacles, and how all of that changes over time.
And a lot of that information simply isn't available in a form AI can easily use.
Satellites are useful, but they don't tell you everything happening at ground level. Traditional mapping fleets have the same problem when it comes to places they can't easily reach.
This is the part of @vangrid_io that caught my attention.
Instead of starting with another AI model, they're looking at the data problem first: how to collect real-world spatial information at scale and turn it into something that Physical AI systems can actually work with.
I think this is worth watching.
As AI moves from screens into the real world, having better models won't be enough. Those models will also need a much better picture of the world they're supposed to operate in.
🤖 Made with AI
The Flowmax Arena: MAXOUT hasn’t officially started yet, but the Early Bird phase is already giving traders something to work with.
What I find interesting here is that the preheat rewards are based on simulated capital, so the main value isn’t the money itself. It’s having more capital available to test your setup before the actual competition begins.
During the Early Bird phase:
• Invite a friend to register + follow @Flowmax_AI → +1,000 USDT simulated capital
• Invite them to join the Flowmax Telegram → +1,000 USDT
• Complete Daily X Reposts → +600 USDT per task
At the same time, you can use this period to build an AI Agent, test different strategies, set up a copy-trading portfolio, and see what actually works before the competition starts.
For me, the interesting part is the preparation rather than simply collecting rewards. More simulated capital gives you more room to experiment and understand your strategy before MAXOUT goes live.
Early access doesn’t guarantee better results, but it does give you more time to prepare.
Simulated capital is not real assets and cannot be withdrawn or converted into USDT.
Solana to triple transaction size as apps get room for more complex trades
The Solana smart contract blockchain is targeting Wednesday to increase the maximum transaction size from 1,232 bytes to 4,096 bytes, giving developers more than three times as much room to fit instructions into a single operation.
Operations that previously had to be split across several transactions can increasingly be handled in a single transaction under Transaction v1. That includes large cryptographic proofs, payments requiring many approvals and some confidential transfers.
The new format is already running on Solana's test and development networks. Existing transaction formats will continue to work, so wallets and apps do not have to switch to v1 unless they need the extra space.
It could help Solana sharpen its edge over its rival Ethereum. While Solana has always been faster and cheaper, it previously fell behind in one key area: its transactions were strictly hard-capped at 1,232 bytes. Ethereum, by contrast, has no rigid protocol size limit, allowing developers to execute massive, data-dense applications in a single operation simply by paying a higher transaction fee.
By more than tripling this ceiling to 4,096 bytes, Solana is looking to eliminate this structural bottleneck.
The bigger change is for software that reads Solana. Services that fetch blocks and transactions need to be updated to recognize v1. If they are not, requests can fail when they encounter the new format.
Some software also needs to change where it looks for priority-fee information. A priority fee is an optional extra payment that users can make to have a transaction processed faster. V1 stores that information elsewhere, so outdated software can show a fee of zero even when one was paid.
Wallets, explorers, and trading apps often rely on these services to display what happened onchain, meaning that bad data behind the scenes can turn into bad data on a user's screen.
Larger transactions also use more network bandwidth. Developers expect users may need to offer higher-priority fees when larger transactions are competing for space, although the upgrade does not introduce a new fee per byte.
The old 1,232-byte limit comes from Solana's original networking design, in which each transaction had to fit within a roughly 1,280-byte internet data packet. Solana changed how transaction traffic is transmitted in 2022, making that old ceiling less necessary.
The new limit stops at 4,096 bytes because four kilobytes is a standard memory-page size used by validator hardware. Going much larger could force one transaction to span several pages and make it more expensive for the machines running Solana to process.
The change is defined in two proposals, SIMD-0296 and SIMD-0385, co-authored by Jacob Creech and Andrew Fitzgerald, and is wholly separate from the recent Solana governance votes that changed how quickly new SOL was created and removed from supply through fees earned from network activities.
$ENA 1$
The crypto market has never really been as simple as “up” or “down.”
Sometimes price stays flat for days while liquidity is quietly moving underneath. Sometimes volume picks up before price actually breaks out. And sometimes a pump looks strong until liquidity disappears and momentum fades almost instantly.
What’s interesting right now isn’t any particular token.
It’s trying to understand where the market is in its current cycle:
Is liquidity flowing in or out?
Is volatility expanding or compressing?
Is risk appetite coming back, or is the market still defensive?
Is BTC leading, or are we starting to see rotation into alts?
Sometimes noticing these small changes is more interesting than trying to predict the next candle.
Market first. Trade later.
Honestly, I don’t even know what I just said anymore. Don’t mind me. 😂😂😂😂
After spending some time using @Flowmax_AI and participating in the first season of its trading competition, I’ve come to realize that what makes Flowmax interesting isn’t simply having an AI tell you when to go Long or Short.
It’s more like building an AI-powered trading team.
A Research Agent can continuously monitor the market and generate signals.
A PM Agent can then combine those signals with the strategy, risk profile, assets, trading frequency, and TP/SL conditions to make trading decisions.
Flowmax also provides Paper Trading and decision tracking, which allows you to observe how a strategy performs before taking it further.
I find this especially interesting during an uptrend.
Instead of constantly staring at charts and trying to guess which coin might pump next, you can have different Agents monitor momentum, OI, funding, market conditions, or other signals, while another Agent processes those inputs into a trading decision.
For me, the biggest advantage isn’t that AI can magically predict the market.
It’s the ability to build a system with a clear trading logic and apply it consistently.
That becomes even more important during an uptrend, when strong market momentum can easily lead to FOMO, late entries, or excessive leverage.
Having position sizing, leverage, TP/SL and risk management rules defined in advance can make a strategy less dependent on emotions.
Another thing I find interesting about Flowmax is the ability to explore, Fork and customize existing Agents instead of having to build everything from scratch.
Of course, past performance doesn’t guarantee future results. The results from the competition should be viewed as data for studying and evaluating strategies, rather than proof that they will generate profits in live markets.
After actually using Flowmax and participating in its first season, I think the more interesting question isn’t:
“Can AI predict which coin will pump next?”
It’s:
“Can we build an AI trading system that identifies opportunities, manages risk and stays disciplined when the market gets highly volatile?”
That’s the part of Flowmax I’ll be watching most closely.
⚠️ This is purely my personal opinion and experience after using Flowmax. It is not financial advice, and I do not encourage anyone to make investment decisions based solely on this content.
Prediction Tickets are used to make predictions.
You can earn up to 25 tickets per account:
+1 ticket from the initial tasks
+1 daily repost, up to 14
+1 per eligible friend invited, up to 10
To access it:
Flowmax → Profile → Task and Points → Prediction Challenge
1
3
29
Not building an AI?
You can still compete through the Points Leaderboard by trading, completing campaign tasks, and inviting new users. The Top 20 participants will share a 1,000 USDT prize pool.
Code: fFjwt_Y0wiKd-twm
Good luck to everyone participating.
5
57
AI Builder rewards include:
• 1st Place: 1,000 USDT
• 2nd–3rd Place: 500 USDT each
• 100 USDT for every builder achieving 10%–100% ROI
Build, deploy, and let your AI compete.
30
Lately, I've been keeping an eye on @timesoulcom and I've noticed they're building a pretty interesting ecosystem: combining mindfulness and technology.
It's not just an app with a virtual personal trainer; they also have cute plushies to help you form healthy habits. Actually, hugging a soft plushie while taking deep breaths or writing down some self-reflection makes the experience much more intimate and less stressful.
In a market where many apps just steal your time, TimeSoul rewarding small habits with $TTS tokens feels reasonable and fair. A breath, a moment of stillness, a reminder from the AI coach… all connected to a plushie sitting on your desk as a reminder that you're taking care of yourself.
🌀 The perfect time to join: I just signed up for the #BingXBlast event to try my luck with the $20K prize pool.
Details: bingx.com/invite/XGRO7L
Thank you to @BingXOfficial for creating a campaign that both promotes mental health and opens up opportunities for growth in crypto.
#TimeSoul #Mindfulness #TTS #BingXBlast
14
24
550