UKong
675 Posts
UKong
Square Verified知足常乐
265 Following
15.6K+ Followers
9.5K+ Liked
Posts
All
Quotes
Videos
Live
$MU
$SNDK $METAB
Last night, U.S. stocks continued to diverge. Tech stocks generally weakened: Micron fell by five percent, and SanDisk plunged more than fourteen percent. Yet the Dow still hit new highs despite the broader weakness.
Meta, which had previously dragged down the tech sector, came with another negative update: over the past four months, progress on AI agent development has been behind expectations. It wiped out a large portion of the gains from the prior day.
Now the market’s most sensitive point is very clear: big companies are aggressively pouring money into compute power and infrastructure—can they recoup costs, and can they keep generating profits?
Meta rents out idle compute capacity to cover downside from early investment. Infrastructure ramp-up always outpaces software monetization. If Meta can’t fully consume the capacity through its own AI business, it rents it out for cash flow—turning hardware that would otherwise be pure cost into a rotating asset and lifting overall returns.
But when the AI agents fail to deliver on the rollout, market perception shifts completely. If Meta’s own AI can run, then renting out compute looks like monetizing an asset. If Meta’s own business can’t make it work, then renting out becomes merely passive damage control—and investors are more confident that Meta is putting its compute into its own high-margin ad and social business, not earning meager rental income.
Separately, at Anthropic’s talks with Samsung about developing chips in-house, the core goal is to cut costs and reduce reliance on a single GPU.
The industry value chain is now being squeezed from both ends. Upstream players like Nvidia aren’t satisfied with just selling hardware; they push for revenue-sharing and subscriptions to lock cloud providers in. Downstream, Meta is moving into compute-rental—directly grabbing business from intermediary resellers. That kind of middleman has an increasingly shrinking space to survive.
The recent volatility isn’t because AI demand has collapsed. It’s a necessary adjustment as the industry shifts from mindless capacity expansion to making money more realistically.
Going forward, you shouldn’t focus on how aggressive the capital expenditures are. The key is the real investment return data.
The upstream hardware logic has also changed. Previously, expanding capacity alone could create a market tailwind. Now, companies that only stack more capacity without terminal-level moats will face ongoing pressure.
The industry is entering an era of “precision accounting.” If demand growth slows, pure capacity players will be the first to be eliminated.
$SNDK $METAB
Last night, U.S. stocks continued to diverge. Tech stocks generally weakened: Micron fell by five percent, and SanDisk plunged more than fourteen percent. Yet the Dow still hit new highs despite the broader weakness.
Meta, which had previously dragged down the tech sector, came with another negative update: over the past four months, progress on AI agent development has been behind expectations. It wiped out a large portion of the gains from the prior day.
Now the market’s most sensitive point is very clear: big companies are aggressively pouring money into compute power and infrastructure—can they recoup costs, and can they keep generating profits?
Meta rents out idle compute capacity to cover downside from early investment. Infrastructure ramp-up always outpaces software monetization. If Meta can’t fully consume the capacity through its own AI business, it rents it out for cash flow—turning hardware that would otherwise be pure cost into a rotating asset and lifting overall returns.
But when the AI agents fail to deliver on the rollout, market perception shifts completely. If Meta’s own AI can run, then renting out compute looks like monetizing an asset. If Meta’s own business can’t make it work, then renting out becomes merely passive damage control—and investors are more confident that Meta is putting its compute into its own high-margin ad and social business, not earning meager rental income.
Separately, at Anthropic’s talks with Samsung about developing chips in-house, the core goal is to cut costs and reduce reliance on a single GPU.
The industry value chain is now being squeezed from both ends. Upstream players like Nvidia aren’t satisfied with just selling hardware; they push for revenue-sharing and subscriptions to lock cloud providers in. Downstream, Meta is moving into compute-rental—directly grabbing business from intermediary resellers. That kind of middleman has an increasingly shrinking space to survive.
The recent volatility isn’t because AI demand has collapsed. It’s a necessary adjustment as the industry shifts from mindless capacity expansion to making money more realistically.
Going forward, you shouldn’t focus on how aggressive the capital expenditures are. The key is the real investment return data.
The upstream hardware logic has also changed. Previously, expanding capacity alone could create a market tailwind. Now, companies that only stack more capacity without terminal-level moats will face ongoing pressure.
The industry is entering an era of “precision accounting.” If demand growth slows, pure capacity players will be the first to be eliminated.
$NVDAB
The core of the entire AI industry chain is essentially just two companies: OpenAI and Anthropic. Every major U.S.-listed tech giant is laying out AI, but in essence they’re all revolving around them—yet investments are all tied to hard constraints.
Some bloggers have broken it down: OpenAI’s $100B-plus fundraising comes with long-term, trillion-level procurement orders. Investors must continuously supply supporting services like GPUs, compute power, and more.
North American capital is wildly pouring money into compute and storage, which also squeezes domestic electronic components and scarce materials. In the end, everyone is counting on OpenAI and Anthropic to recoup their costs. Both companies’ revenues have climbed to around the $10B range. OpenAI loses hundreds of billions every year, while Anthropic barely manages to scrape out a small profit.
At the moment, the only product that has truly gone mainstream is Anthropic’s Claude Code (CC for short). Most programmers have used it. The open-source tool “Lobster” is essentially a low-spec desktop version of CC. Tencent’s WorkBuddy, Alibaba’s Qoder/QorderWork—all follow the same underlying idea. AI video-editing on social media is popular, but the hallucination problem can’t be cured. Beyond creative work, other scenarios only drive costs higher.
Even Anthropic, which holds the breakout hit, is only barely profitable. Application revenue alone can’t possibly support massive capital expenditures. The market looks for real cash flow and profits, not just ARR on paper.
Selling tokens and building downstream applications have long been a red ocean with very low entry barriers—nothing more than tweaking prompts and integrating third-party tools. The ToC model gets more expensive the more you use it, and so far it hasn’t been made to work. Traditional internet B2B projects already tend to have thin margins, and when you stack AI’s high costs on top, profitability becomes even harder.
With monetization blocked on the application side, capital naturally grows anxious. At the same time, large-model performance is approaching a bottleneck. Today’s LLMs feel more like a “metaverse” with better user experience: the chat experience is impressive, but it still doesn’t reach the industry’s touted level of disruptive transformation.
AI booms tend to come in waves. ChatGPT mainly surged by throwing compute at the problem, but the underlying architecture is still the Transformer from Google in 2017. Technical breakthroughs don’t follow a fixed cycle—you might see a qualitative change tomorrow, or you might have to wait ten more years or even longer. Only engineering deployment brings clear progress; just climbing leaderboards doesn’t count as substantive advancement.
The core of the entire AI industry chain is essentially just two companies: OpenAI and Anthropic. Every major U.S.-listed tech giant is laying out AI, but in essence they’re all revolving around them—yet investments are all tied to hard constraints.
Some bloggers have broken it down: OpenAI’s $100B-plus fundraising comes with long-term, trillion-level procurement orders. Investors must continuously supply supporting services like GPUs, compute power, and more.
North American capital is wildly pouring money into compute and storage, which also squeezes domestic electronic components and scarce materials. In the end, everyone is counting on OpenAI and Anthropic to recoup their costs. Both companies’ revenues have climbed to around the $10B range. OpenAI loses hundreds of billions every year, while Anthropic barely manages to scrape out a small profit.
At the moment, the only product that has truly gone mainstream is Anthropic’s Claude Code (CC for short). Most programmers have used it. The open-source tool “Lobster” is essentially a low-spec desktop version of CC. Tencent’s WorkBuddy, Alibaba’s Qoder/QorderWork—all follow the same underlying idea. AI video-editing on social media is popular, but the hallucination problem can’t be cured. Beyond creative work, other scenarios only drive costs higher.
Even Anthropic, which holds the breakout hit, is only barely profitable. Application revenue alone can’t possibly support massive capital expenditures. The market looks for real cash flow and profits, not just ARR on paper.
Selling tokens and building downstream applications have long been a red ocean with very low entry barriers—nothing more than tweaking prompts and integrating third-party tools. The ToC model gets more expensive the more you use it, and so far it hasn’t been made to work. Traditional internet B2B projects already tend to have thin margins, and when you stack AI’s high costs on top, profitability becomes even harder.
With monetization blocked on the application side, capital naturally grows anxious. At the same time, large-model performance is approaching a bottleneck. Today’s LLMs feel more like a “metaverse” with better user experience: the chat experience is impressive, but it still doesn’t reach the industry’s touted level of disruptive transformation.
AI booms tend to come in waves. ChatGPT mainly surged by throwing compute at the problem, but the underlying architecture is still the Transformer from Google in 2017. Technical breakthroughs don’t follow a fixed cycle—you might see a qualitative change tomorrow, or you might have to wait ten more years or even longer. Only engineering deployment brings clear progress; just climbing leaderboards doesn’t count as substantive advancement.
NVDA4200 (NVIDIA) 🟢
CA: EJu78KrhDyaLn3p8Y89dgVGDgLKxU4rzSE7fER5yNVDA
Recently, the AI Agent narrative has started to heat up again. This one happens to stack several current hot trends at the same time: Nvidia, AI, memes, and US stocks.
They’ve already released an Autonomous AI Agent that can make trading decisions on its own and also manage social media by itself—the gameplay fits the direction the market currently likes.
They put a lot of effort into the logo, visuals, and the official website. It has that “old meme” vibe—something you find more and more agreeable the more you look at it.
There are also things like AI Meme Trader and mini-games that can be rolled out further.
Lately, Nvidia has kept hitting new highs, and AI hype hasn’t cooled down either. As long as the market keeps pumping this narrative, attention shouldn’t be low.
I’ll put it on my watchlist first (I just bought a tiny amount earlier and I’m currently down/underwater).
X: @NVDA4200
CA: EJu78KrhDyaLn3p8Y89dgVGDgLKxU4rzSE7fER5yNVDA
Recently, the AI Agent narrative has started to heat up again. This one happens to stack several current hot trends at the same time: Nvidia, AI, memes, and US stocks.
They’ve already released an Autonomous AI Agent that can make trading decisions on its own and also manage social media by itself—the gameplay fits the direction the market currently likes.
They put a lot of effort into the logo, visuals, and the official website. It has that “old meme” vibe—something you find more and more agreeable the more you look at it.
There are also things like AI Meme Trader and mini-games that can be rolled out further.
Lately, Nvidia has kept hitting new highs, and AI hype hasn’t cooled down either. As long as the market keeps pumping this narrative, attention shouldn’t be low.
I’ll put it on my watchlist first (I just bought a tiny amount earlier and I’m currently down/underwater).
X: @NVDA4200
NVDAonAlpha
NVDAUS-1.47%
Verified
A lot of folks see @tradebetterapp as a prediction market platform, but I prefer to view it as the infrastructure layer for PredictionFi. Right now, prediction markets are overflowing with trading venues, but what’s really missing is an efficient, stable execution layer. Take Polymarket, for instance; many can find smart money addresses, but keeping up with the flow is tough. One of BETTER's standout features is its execution speed, with official data showing execution at 0.6ms. You can copy smart wallets with one click, way faster than the average user can manually trade, getting closer to a pro trading experience. Plus, it’s not just about mindlessly copying trades. BETTER has a dynamic wallet scoring system that continuously tracks and evaluates wallet performance, filtering out wallets that lose their edge, keeping only those with real, sustained Alpha, making following trades way more valuable. Another key point is transparency. Since it relies on Polymarket's public blockchain data, all wallet movements, position changes, and entry/exit records can be validated and audited. The source of signals is crystal clear, which is a crucial foundation for building trust. The upcoming Vault only requires you to deposit funds, while the strategy takes care of execution. The Vault operates on a 20% performance fee after profits, no charge if there are no gains, and it has built-in risk control mechanisms. Compared to many projects that depend on high emissions and are driven by emotions, this model is obviously more sustainable. I've always believed that prediction markets are one of the few sectors not heavily affected by bull or bear cycles. Regardless of market conditions, people will always be predicting and trading on events like elections, economic data, sports events, and tech trends, so there’s a natural demand underpinning the whole market. The truly valuable opportunities often lie not in creating a new venue but in packaging complex venues into a product layer that everyday people can easily use. - Solving execution issues with 0.6ms ultra-low latency - Addressing copy trading quality issues with intelligent screening systems - Resolving trust issues with transparent on-chain data - Lowering participation barriers for everyday users with Vault - Future potential to aggregate more quality signals from prediction markets and trading venues Over the next 3-5 years, PredictionFi is set to grow into a long-lasting super sector, definitely worth keeping an eye on. CA: 0x396ffad9469e3d3e3fc4061b79acce2ad0ce4b9e
Last night I put up a few entertainment shorts, closed two small trades, and this morning I woke up to see I had flattened them.
Now I'm even a bit tempted to go long again, am I crazy? 😂 $LAB
Now I'm even a bit tempted to go long again, am I crazy? 😂 $LAB
I've been using a bunch of prediction market tools lately, but honestly, the experience after unlocking $BETER has been pretty solid @tradebetterapp.
The most noticeable change is the information gap has shrunk significantly.
Back when I was checking Polymarket, I often found that by the time I spotted an opportunity, the price had already moved, or I saw big money entering but couldn't react in time. Now, when I open Terminal, a lot of signals pop up ahead of time—big money flows, market sentiment shifts are almost updated in real-time, and one-click execution saves a lot of back-and-forth page switching.
One thing I really like is the completeness of the information provided. Besides trade direction, I can also see Z-Score, Insider Score, liquidity changes, correlation analysis, and other data. Many trades that look smart are actually hedging or risk management plays—analyzing these details on your own can take a lot of time.
The recently launched private AI feature is pretty neat too. After uploading my research materials, querying and organizing information has become way easier, and all content runs in a Nitro Enclave environment, so the prompts aren’t saved, plus every call gets a verifiable receipt on-chain. For those who regularly conduct market research, this design definitely brings peace of mind.
🆕Better is still in its early project phase and continues to optimize and update, with recent updates coming in hot:
- Private AI + Mesh Router officially live, supporting 46+ models and multiple AI providers
- Nitro Enclave hardware-level privacy protection and on-chain verifiable receipts now in effect
- Private Agent supports uploading PDF, DOCX, CSV files to build a knowledge base
- HFT infrastructure ongoing optimizations to further reduce latency
- Terminal has entered a more stable public testing phase, with noticeable improvements in UI and overall experience
- Vault (the first Polymarket ETF) is approaching closed Beta
- After Vault launches, the $BETER burn mechanism will kick in, with related wallets being transparent
The Better team is focusing on advancing trading execution, AI tools, and capital management simultaneously, and the update frequency is super high.
At least for those who frequently play on Polymarket, getting used to this will make going back to traditional manual operations feel like a major efficiency loss.
$BETTER is still holding, haven’t moved it, and I personally think June will be an explosive period #DYOR
The most noticeable change is the information gap has shrunk significantly.
Back when I was checking Polymarket, I often found that by the time I spotted an opportunity, the price had already moved, or I saw big money entering but couldn't react in time. Now, when I open Terminal, a lot of signals pop up ahead of time—big money flows, market sentiment shifts are almost updated in real-time, and one-click execution saves a lot of back-and-forth page switching.
One thing I really like is the completeness of the information provided. Besides trade direction, I can also see Z-Score, Insider Score, liquidity changes, correlation analysis, and other data. Many trades that look smart are actually hedging or risk management plays—analyzing these details on your own can take a lot of time.
The recently launched private AI feature is pretty neat too. After uploading my research materials, querying and organizing information has become way easier, and all content runs in a Nitro Enclave environment, so the prompts aren’t saved, plus every call gets a verifiable receipt on-chain. For those who regularly conduct market research, this design definitely brings peace of mind.
🆕Better is still in its early project phase and continues to optimize and update, with recent updates coming in hot:
- Private AI + Mesh Router officially live, supporting 46+ models and multiple AI providers
- Nitro Enclave hardware-level privacy protection and on-chain verifiable receipts now in effect
- Private Agent supports uploading PDF, DOCX, CSV files to build a knowledge base
- HFT infrastructure ongoing optimizations to further reduce latency
- Terminal has entered a more stable public testing phase, with noticeable improvements in UI and overall experience
- Vault (the first Polymarket ETF) is approaching closed Beta
- After Vault launches, the $BETER burn mechanism will kick in, with related wallets being transparent
The Better team is focusing on advancing trading execution, AI tools, and capital management simultaneously, and the update frequency is super high.
At least for those who frequently play on Polymarket, getting used to this will make going back to traditional manual operations feel like a major efficiency loss.
$BETTER is still holding, haven’t moved it, and I personally think June will be an explosive period #DYOR
This June has seen a ton of major sports events and esports tournaments, including the NBA Finals, World Cup, French Open, League of Legends MSI, CS2 Major, Valorant, PUBG, and more. The prediction markets are already on fire, and I've secretly infiltrated several prediction betting groups, diving into some projects and tools. Recently, I noticed a lot of influencers talking about a platform made specifically for Polymarket, built with RUST, featuring a unique Polygon/mempool pre-execution stack and millisecond signal ingestion - @tradebetterapp claims to surpass 90% of on-chain bots in trading speed. I just checked it out, and while it's still pretty bare-bones since it just launched, the number of signals is quite impressive, almost a bit overwhelming. I randomly clicked on a few signals, and they seemed decent, looking pretty accurate, but I'm still unclear about the risk-reward ratio and the slippage between signal generation and execution; I haven't funded it yet to give it a shot, so I'll check it out again tomorrow. Here’s a quick rundown of some project info: 1⃣ Backed by @openservai (which raised $100M in Pre-seed) and @getdomeapi. The three co-founders have solid backgrounds, including top-tier Quant traders, a former senior AWS architect/TikTok security engineer, and extensive project delivery experience in GTM. 2⃣ In the testing phase, they’ve already processed over 10 million Alpha signals and are now at a usable stage, filtering out 155,000 +EV (positive expected value) signals daily. You can dive into the terminal to see real wallet flows and leaderboards, with real-time info delivering a decisive edge. Additionally, the project's token $BETTER has launched, so it’s kind of a cold start. The utility of $BETTER: Buy BETTER ➡️ Unlock the terminal ➡️ Copy strategies/learn methods ➡️ Upgrade permissions. There’s a tiered threshold (minimum holding of 3.5k to experience Guppy, 100k-200k to unlock Fish permissions and trial fund library, etc.). Plus, the token comes with a powerful buyback and burn mechanism, with 2% trading tax and light mode fees all going directly to the market for buybacks and burns; they’ll also add a 20% performance fee from the funding pool for burns later on. $BETTER has had a nice increase this month, already doubling in value, and I feel its breakout point might be in June, so I also bought $400 worth of tokens as a lottery bet.
Got wrecked right off the bat, it's rough out here
Better to sit back and observe
Hibit’s event contracts are pretty fun, every time I want to open a position I first hop on to test the market vibes, haha
$BTC
Better to sit back and observe
Hibit’s event contracts are pretty fun, every time I want to open a position I first hop on to test the market vibes, haha
$BTC
AI Sector Flip: Anthropic Turns Profit, OpenAI Continues to Burn Cash
Once seen as just a follower of OpenAI, Anthropic has hit its first profit milestone since its inception, and it’s two years ahead of schedule.
In Q2, Anthropic expects revenue to reach $10.9 billion, doubling its growth quarter over quarter, while snagging $559 million in operating profit. Currently, its annualized revenue is nearing $45 billion, compared to OpenAI's $25 billion during the same period, creating a nearly 2x gap in earnings.
Anthropic's growth rate is pretty rare; in just a year and a half, its annualized revenue skyrocketed from an initial $1 billion to $45 billion, with a quarterly growth rate hitting 80x, far surpassing the market's previous expectations of 10x annual growth.
In the core enterprise services lane, Anthropic has clearly overtaken the competition. The latest industry stats show that the percentage of U.S. enterprises paying for Anthropic's services has reached 34.4%, surpassing OpenAI's 32.3% for the first time. Thanks to the core advantages of the Claude Code product, over 1,000 companies are now paying a million a year, doubling their enterprise client count in just three months, along with ongoing optimization of computing costs, leading to steady improvements in overall profit efficiency.
On the flip side, OpenAI's situation is starkly different. Its Q1 revenue was $5.7 billion, which sounds impressive, but it’s stuck in a deep loss pit, with an adjusted operating profit margin of -122%. In simple terms, for every dollar earned, they’re losing $1.22.
User growth for OpenAI has also completely stalled; ChatGPT's weekly active users are stuck at 905 million, failing to break the billion mark, with some periods even showing a drop in user numbers. Internally, there’s already a warning about a market winter. Relying solely on subscriptions, enterprise business, and dipping toes into advertising barely supports revenue, making it hard to fill the massive loss gap.
The core of the gap between the two lies in computing power strategy. Anthropic isn’t sticking to a single-binding approach; they’ve integrated computing power from Nvidia, Amazon, Google, Microsoft, and SpaceX, locking in $330 billion in computing resources, with a multi-source supply and extreme cost control. Meanwhile, OpenAI is just now starting to catch up with this strategy and has already fallen behind significantly.
Currently, both leading AI firms are racing towards an IPO, with OpenAI aiming for a trillion-dollar valuation, while Anthropic's pre-valuation has already hit $900 billion.
Who would have thought that five years after spinning off from OpenAI to found Anthropic, it would be the first to nail down a profitable model, while OpenAI, sitting on top-tier traffic, is still burning cash at full throttle.
Once seen as just a follower of OpenAI, Anthropic has hit its first profit milestone since its inception, and it’s two years ahead of schedule.
In Q2, Anthropic expects revenue to reach $10.9 billion, doubling its growth quarter over quarter, while snagging $559 million in operating profit. Currently, its annualized revenue is nearing $45 billion, compared to OpenAI's $25 billion during the same period, creating a nearly 2x gap in earnings.
Anthropic's growth rate is pretty rare; in just a year and a half, its annualized revenue skyrocketed from an initial $1 billion to $45 billion, with a quarterly growth rate hitting 80x, far surpassing the market's previous expectations of 10x annual growth.
In the core enterprise services lane, Anthropic has clearly overtaken the competition. The latest industry stats show that the percentage of U.S. enterprises paying for Anthropic's services has reached 34.4%, surpassing OpenAI's 32.3% for the first time. Thanks to the core advantages of the Claude Code product, over 1,000 companies are now paying a million a year, doubling their enterprise client count in just three months, along with ongoing optimization of computing costs, leading to steady improvements in overall profit efficiency.
On the flip side, OpenAI's situation is starkly different. Its Q1 revenue was $5.7 billion, which sounds impressive, but it’s stuck in a deep loss pit, with an adjusted operating profit margin of -122%. In simple terms, for every dollar earned, they’re losing $1.22.
User growth for OpenAI has also completely stalled; ChatGPT's weekly active users are stuck at 905 million, failing to break the billion mark, with some periods even showing a drop in user numbers. Internally, there’s already a warning about a market winter. Relying solely on subscriptions, enterprise business, and dipping toes into advertising barely supports revenue, making it hard to fill the massive loss gap.
The core of the gap between the two lies in computing power strategy. Anthropic isn’t sticking to a single-binding approach; they’ve integrated computing power from Nvidia, Amazon, Google, Microsoft, and SpaceX, locking in $330 billion in computing resources, with a multi-source supply and extreme cost control. Meanwhile, OpenAI is just now starting to catch up with this strategy and has already fallen behind significantly.
Currently, both leading AI firms are racing towards an IPO, with OpenAI aiming for a trillion-dollar valuation, while Anthropic's pre-valuation has already hit $900 billion.
Who would have thought that five years after spinning off from OpenAI to found Anthropic, it would be the first to nail down a profitable model, while OpenAI, sitting on top-tier traffic, is still burning cash at full throttle.
Wow! KFC is actually up for sale, and it's the Jardine Group making the move, with a whopping price tag of 2.8 billion RMB!
Speaking of which, Jardine holds some serious fast food assets, having operated for over a decade in five markets: Hong Kong, Macau, Taiwan, Myanmar, and Vietnam, including KFC and Pizza Hut, plus the local pizza delivery brand PHD in Hong Kong, totaling around 1,000 outlets and 25,000 employees. This deal roughly translates to about 400 million USD, and the non-binding bids are due this week.
Currently, there are quite a few players eyeing this juicy opportunity—Yum China, Carlyle Group, Uni-President Enterprises from Taiwan, and several private equity firms are in the mix. The strategies of the three core players differ significantly: Yum China aims to leverage its mainland operations to boost profit margins, Carlyle is looking to consolidate the East Asia KFC market for capital gains, and Uni-President is focused on its Taiwan outlets to fill the gaps in its Western fast food offerings.
Some might not know, but Jardine has been in China since the 19th century as a British trading company, primarily dealing in real estate, retail, hotels, and transportation. For them, food and beverage is just a sideline, and this segment's profit margin of 4%-5% is considered mid to low tier in the fast food industry. Plus, with Hong Kong's sluggish consumption and the fast food market already dominated by local brands, selling off this asset is essentially a way for Jardine to slim down and reinvest in their core business.
This isn't really a new trend; in the past decade, foreign food brands handing over control to local capital has become the norm. From Citic and Carlyle acquiring McDonald's China in 2017 to Boyu Capital taking Starbucks China and CPE Yuen Fung getting Burger King China by the end of 2025, multinational brands are stepping back to focus on brand licensing, while local capital takes over operations to unlock potential. This 2.8 billion deal by Jardine is just another example of that trend.
Speaking of which, Jardine holds some serious fast food assets, having operated for over a decade in five markets: Hong Kong, Macau, Taiwan, Myanmar, and Vietnam, including KFC and Pizza Hut, plus the local pizza delivery brand PHD in Hong Kong, totaling around 1,000 outlets and 25,000 employees. This deal roughly translates to about 400 million USD, and the non-binding bids are due this week.
Currently, there are quite a few players eyeing this juicy opportunity—Yum China, Carlyle Group, Uni-President Enterprises from Taiwan, and several private equity firms are in the mix. The strategies of the three core players differ significantly: Yum China aims to leverage its mainland operations to boost profit margins, Carlyle is looking to consolidate the East Asia KFC market for capital gains, and Uni-President is focused on its Taiwan outlets to fill the gaps in its Western fast food offerings.
Some might not know, but Jardine has been in China since the 19th century as a British trading company, primarily dealing in real estate, retail, hotels, and transportation. For them, food and beverage is just a sideline, and this segment's profit margin of 4%-5% is considered mid to low tier in the fast food industry. Plus, with Hong Kong's sluggish consumption and the fast food market already dominated by local brands, selling off this asset is essentially a way for Jardine to slim down and reinvest in their core business.
This isn't really a new trend; in the past decade, foreign food brands handing over control to local capital has become the norm. From Citic and Carlyle acquiring McDonald's China in 2017 to Boyu Capital taking Starbucks China and CPE Yuen Fung getting Burger King China by the end of 2025, multinational brands are stepping back to focus on brand licensing, while local capital takes over operations to unlock potential. This 2.8 billion deal by Jardine is just another example of that trend.
Went back to HiBit for some predictions. The good thing about this CEX is that you can place bets anytime, with settlements after 5 minutes. The downside is that the odds are fixed at 1:0.8. But it's pretty engaging, and for now, no KYC is required, so it feels pretty good.
UKong
·
--
Now you can also play predictions on cex. I just deposited 50 dollars in HIBT and played a bit, and luck was quite good, I made a little profit.
To be honest, there are not many things to play now. A few days ago, I played on a copycat contract and lost 4 trades in a row, so I don't want to play contracts for the time being, just exploring new platforms and playing predictions as a transition.
According to rumors, An An should also be launching a prediction module soon. Let's try cex's predictions first; it feels quite smooth.
To be honest, there are not many things to play now. A few days ago, I played on a copycat contract and lost 4 trades in a row, so I don't want to play contracts for the time being, just exploring new platforms and playing predictions as a transition.
According to rumors, An An should also be launching a prediction module soon. Let's try cex's predictions first; it feels quite smooth.
Now you can also play predictions on cex. I just deposited 50 dollars in HIBT and played a bit, and luck was quite good, I made a little profit.
To be honest, there are not many things to play now. A few days ago, I played on a copycat contract and lost 4 trades in a row, so I don't want to play contracts for the time being, just exploring new platforms and playing predictions as a transition.
According to rumors, An An should also be launching a prediction module soon. Let's try cex's predictions first; it feels quite smooth.
To be honest, there are not many things to play now. A few days ago, I played on a copycat contract and lost 4 trades in a row, so I don't want to play contracts for the time being, just exploring new platforms and playing predictions as a transition.
According to rumors, An An should also be launching a prediction module soon. Let's try cex's predictions first; it feels quite smooth.
1. Current AI Background: Evolving from 'Tool' to 'Subject'
In 2026, AI is at a critical point for value realization, bidding farewell to parameter stacking and shifting towards practical applications and the scaling of intelligent agents.
• Industry Landscape: The global AI market size surpasses $900 billion, with China leading the growth at a compound annual growth rate of over 30%.
• Core Leap: AI Agents become mainstream, with Gartner predicting that 40% of enterprises will embed task-oriented intelligent agents, upgrading from 'Q&A tools' to autonomous execution subjects.
• Technical Foundation: Multi-modal integration and the maturity of mixed expert models (MoE) reduce deployment costs by 60% and improve inference efficiency by 3-5 times.
• Implementation Features: Human-machine collaboration becomes the norm, with new positions (AI Trainer, Prompt Engineer) emerging, and industries focusing on cost reduction, efficiency enhancement, and compliance safety.
2. Trends and Applications of AI in the Cryptocurrency Sector
1. Native Penetration of AI Agents: 86.8% of participants recognize that AI can autonomously place orders, clear transactions, and manage risks, becoming the core of trading.
2. RWA + AI Accelerated Implementation: AI empowers the valuation and rights confirmation of real-world assets, with on-chain circulation scaling and institutional focus.
3. Dual Track of Compliance and Privacy: The popularization of zero-knowledge proof (ZK) technology balances privacy and regulation, promoting compliant privacy.
4. Core Applications:
◦ Quantitative Trading: Algorithms like LSTM identify trends, fund rate arbitrage, and track whale addresses, with AI continuously evolving strategies.
◦ On-chain Analysis: AI monitors abnormal transactions in real-time, reducing fraud risk by 60%; generating price probability distributions to optimize entry and exit.
◦ DeFi/DAO: AI automatically makes markets and clears; assisting in governance to improve efficiency and fairness.
3. Trading Suggestions
1. Tool Strategy: Prioritize AI agents with MEV protection and circuit breaker mechanisms to avoid flash crashes and spike risks.
2. Portfolio Configuration: AI trend tracking plus strict stop-loss (no more than 3% per single asset) to diversify assets and reduce drawdown.
3. Risk Control: Beware of overfitting and high-frequency errors; retain manual intervention interfaces to cope with extreme market conditions.
4. Focus Areas: The three main directions are AI Agent infrastructure, RWA tokenization, and privacy computing.
In 2026, AI is at a critical point for value realization, bidding farewell to parameter stacking and shifting towards practical applications and the scaling of intelligent agents.
• Industry Landscape: The global AI market size surpasses $900 billion, with China leading the growth at a compound annual growth rate of over 30%.
• Core Leap: AI Agents become mainstream, with Gartner predicting that 40% of enterprises will embed task-oriented intelligent agents, upgrading from 'Q&A tools' to autonomous execution subjects.
• Technical Foundation: Multi-modal integration and the maturity of mixed expert models (MoE) reduce deployment costs by 60% and improve inference efficiency by 3-5 times.
• Implementation Features: Human-machine collaboration becomes the norm, with new positions (AI Trainer, Prompt Engineer) emerging, and industries focusing on cost reduction, efficiency enhancement, and compliance safety.
2. Trends and Applications of AI in the Cryptocurrency Sector
1. Native Penetration of AI Agents: 86.8% of participants recognize that AI can autonomously place orders, clear transactions, and manage risks, becoming the core of trading.
2. RWA + AI Accelerated Implementation: AI empowers the valuation and rights confirmation of real-world assets, with on-chain circulation scaling and institutional focus.
3. Dual Track of Compliance and Privacy: The popularization of zero-knowledge proof (ZK) technology balances privacy and regulation, promoting compliant privacy.
4. Core Applications:
◦ Quantitative Trading: Algorithms like LSTM identify trends, fund rate arbitrage, and track whale addresses, with AI continuously evolving strategies.
◦ On-chain Analysis: AI monitors abnormal transactions in real-time, reducing fraud risk by 60%; generating price probability distributions to optimize entry and exit.
◦ DeFi/DAO: AI automatically makes markets and clears; assisting in governance to improve efficiency and fairness.
3. Trading Suggestions
1. Tool Strategy: Prioritize AI agents with MEV protection and circuit breaker mechanisms to avoid flash crashes and spike risks.
2. Portfolio Configuration: AI trend tracking plus strict stop-loss (no more than 3% per single asset) to diversify assets and reduce drawdown.
3. Risk Control: Beware of overfitting and high-frequency errors; retain manual intervention interfaces to cope with extreme market conditions.
4. Focus Areas: The three main directions are AI Agent infrastructure, RWA tokenization, and privacy computing.
Recent observations in the market and investment opportunities
1. Milestone Events (March 10)
• The 20 millionth BTC officially mined: 95% of the total supply of 21 million has been produced, with 1 million remaining to be mined over the next 114 years, further highlighting its scarcity.
2. Major Global Regulatory Actions (Compliance Acceleration)
• First batch of stablecoin licenses issued in Hong Kong: 100% reserves, segregated custody, T+1 redemption, HSBC/Standard Chartered/Bank of China Hong Kong, etc., entering the market, forming an Asian compliance hub.
• Key negotiations on the U.S. Clarity Act: Clarifying SEC/CFTC responsibilities, resolving stablecoin yield disputes, benefiting Circle (USDC).
• The EU MiCA regulation fully effective on March 25: The world's first unified crypto regulatory framework, favorable for compliant exchanges and stablecoins.
• Document No. 42 from eight departments in mainland China: All virtual currency businesses are illegal domestically, strictly prohibiting RWA tokenization and cross-border services.
3. Institutions and Market Dynamics (Institutional Bottom Fishing, Retail Panic)
• BTC spot ETF fund inflow: Net inflow of $458 million in a single day on March 3, with BlackRock's IBIT accounting for $263 million, ending five weeks of net outflow.
• Whales increasing holdings against the trend: Net purchase of about 200,000 BTC over two weeks, addresses holding over 100,000 increased by 14,000.
• Critical approval period for ETH staking ETF: Applications from BlackRock and others, approval expected to bring massive incremental funds.
• Massive unlock in March: Approximately $5.8–6 billion in tokens unlocked, with concentrated selling pressure from WBT, RAIN, and SUI.
4. Market and Hot Projects
• Market Fluctuations: BTC oscillating between $67,000 and $71,000, ETH falling below $2,000, frequent liquidations across the network.
• Aptos Deflation: Community unanimously approved a hard cap of 2.1 billion APT, transitioning from inflation to deflation.
• RWA Tokenization Surge: Global scale increased nearly fourfold in a year, with Hong Kong becoming a key area for institutional investments.
• Compliance of Mixers: U.S. Treasury Department acknowledges their legal privacy uses for the first time.
5. Short-Term Key Dates
• March 18: Federal Reserve interest rate decision (interest rate cut expectations affecting liquidity).
• March 25: Full implementation of EU MiCA.
$BTC
1. Milestone Events (March 10)
• The 20 millionth BTC officially mined: 95% of the total supply of 21 million has been produced, with 1 million remaining to be mined over the next 114 years, further highlighting its scarcity.
2. Major Global Regulatory Actions (Compliance Acceleration)
• First batch of stablecoin licenses issued in Hong Kong: 100% reserves, segregated custody, T+1 redemption, HSBC/Standard Chartered/Bank of China Hong Kong, etc., entering the market, forming an Asian compliance hub.
• Key negotiations on the U.S. Clarity Act: Clarifying SEC/CFTC responsibilities, resolving stablecoin yield disputes, benefiting Circle (USDC).
• The EU MiCA regulation fully effective on March 25: The world's first unified crypto regulatory framework, favorable for compliant exchanges and stablecoins.
• Document No. 42 from eight departments in mainland China: All virtual currency businesses are illegal domestically, strictly prohibiting RWA tokenization and cross-border services.
3. Institutions and Market Dynamics (Institutional Bottom Fishing, Retail Panic)
• BTC spot ETF fund inflow: Net inflow of $458 million in a single day on March 3, with BlackRock's IBIT accounting for $263 million, ending five weeks of net outflow.
• Whales increasing holdings against the trend: Net purchase of about 200,000 BTC over two weeks, addresses holding over 100,000 increased by 14,000.
• Critical approval period for ETH staking ETF: Applications from BlackRock and others, approval expected to bring massive incremental funds.
• Massive unlock in March: Approximately $5.8–6 billion in tokens unlocked, with concentrated selling pressure from WBT, RAIN, and SUI.
4. Market and Hot Projects
• Market Fluctuations: BTC oscillating between $67,000 and $71,000, ETH falling below $2,000, frequent liquidations across the network.
• Aptos Deflation: Community unanimously approved a hard cap of 2.1 billion APT, transitioning from inflation to deflation.
• RWA Tokenization Surge: Global scale increased nearly fourfold in a year, with Hong Kong becoming a key area for institutional investments.
• Compliance of Mixers: U.S. Treasury Department acknowledges their legal privacy uses for the first time.
5. Short-Term Key Dates
• March 18: Federal Reserve interest rate decision (interest rate cut expectations affecting liquidity).
• March 25: Full implementation of EU MiCA.
$BTC
Recently, OpenClaw lobster has gained significant popularity in the cryptocurrency space, with discussions increasing from tools to ecosystem. Many are attracted by its concepts of automation, AI agents, and one-click operation, even hoping it can become a tool for "easy profit." However, amidst the excitement, we might reconsider this "lobster" from a more dialectical and rational perspective.
It is undeniable that OpenClaw, as an open-source project, has certain exploratory significance in the direction of intelligent agents and automation tools, and indeed lowers the threshold for some on-chain operations and task executions. Whenever a new tool emerges, it brings expectations and imaginative possibilities; we need not entirely dismiss its technological value and potential future from the outset.
However, at the same time, the actual effectiveness at this stage indeed still needs to be validated.
cz has also mentioned a similar viewpoint: many people claim that after installing the lobster, they don’t have to do anything, but in fact, most of the time is spent debugging this tool that "seems to be able to do everything but actually does nothing."
This is also the genuine feeling of many users:
• Complex configuration, the threshold is higher than expected
• Actual operation is unstable and requires repeated debugging
• The advertised "automated profits" have inconsistent real-world results
The so-called lobster being able to make money on its own currently remains more on the level of expectations and concepts rather than a universal, stable, and reproducible fact. Some may find suitable scenarios for themselves, but for most ordinary users, stable profitability remains in doubt, let alone easy earnings.
The tool itself is neither right nor wrong; the issue lies in the exaggerated expectations and the blind following driven by emotions.
Whether OpenClaw lobster can truly create value and practical applications in the future still requires time and the market to test. Until then, maintaining a wait-and-see attitude, rational experimentation, avoiding mythologizing or belittling may be a more mature stance.
It is undeniable that OpenClaw, as an open-source project, has certain exploratory significance in the direction of intelligent agents and automation tools, and indeed lowers the threshold for some on-chain operations and task executions. Whenever a new tool emerges, it brings expectations and imaginative possibilities; we need not entirely dismiss its technological value and potential future from the outset.
However, at the same time, the actual effectiveness at this stage indeed still needs to be validated.
cz has also mentioned a similar viewpoint: many people claim that after installing the lobster, they don’t have to do anything, but in fact, most of the time is spent debugging this tool that "seems to be able to do everything but actually does nothing."
This is also the genuine feeling of many users:
• Complex configuration, the threshold is higher than expected
• Actual operation is unstable and requires repeated debugging
• The advertised "automated profits" have inconsistent real-world results
The so-called lobster being able to make money on its own currently remains more on the level of expectations and concepts rather than a universal, stable, and reproducible fact. Some may find suitable scenarios for themselves, but for most ordinary users, stable profitability remains in doubt, let alone easy earnings.
The tool itself is neither right nor wrong; the issue lies in the exaggerated expectations and the blind following driven by emotions.
Whether OpenClaw lobster can truly create value and practical applications in the future still requires time and the market to test. Until then, maintaining a wait-and-see attitude, rational experimentation, avoiding mythologizing or belittling may be a more mature stance.
Log in to explore more content