Matt Page (@mattjohnpage) header
Matt Page (@mattjohnpage) profile picture

Music Man I Macro I Digital Assets with a touch of metals and a smattering of interesting alts. On new adventures...always learning!

Singapore
Joined October 2011
Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)Photo by Matt Page (@mattjohnpage)
I've been waiting months to put the Mac Studio M5 Ultra 512GB head to head against four DGX Sparks. Today I finally did, with GLM-5.3-Flash at max reasoning from 2K to 1M tokens.
28
12
10
210
49,930
Just switched from @WisprFlow to Hex (free, open source, one-click install) and its night and day difference on speed (it's native Rust!), and accuracy is really good. local first diction. Get it. Also a reminder that WisprFlow, at least for meeting notes, 100% trains on your private data—they say so when they force you to disable HIPAA to use it. Big Data is coming after you. Break the cycle. Link in comments.
6
1
15
763
oil goes up, so rates go up. every idiot on the internet knows that one. this week canada did it. first hike in three years, to 2.5%, with brent near $116. the clueless grey brigade at the bank of england's expected to follow next week, and the market's pricing up to 75bp from warsh over the next year. but at $116 oil's already a tax, levied on every commuter and every trucker. a central bank that hikes into it collects that tax twice. it's just that weak men in grey suits feel the need to do something, anything. the refiners have noticed. crude's screaming war while the 3-2-1 crack spread's fallen from over $72 to about $50 as diesel floods in. the people who turn oil into something you can actually buy are telling you demand's already cracking. it's the conceit and arrogance of a well-formed argument. we've seen this film. in july 2008 the ecb hiked with oil above $140. you know how that autumn went. tighten into a supply shock and you manufacture the slowdown that would've killed the inflation anyway. a 10y at 5% is the price of believing them. watch core pce tomorrow. if it beats 0.3% m/m and the 10y breaks back through 5.3%, the hawks have a case. if not, they're hiking into a recession they're building themselves.
23
32
5
273
20,028
Matt Page retweeted
Singapore: TOKEN2049 Week This week in Singapore, I’m on stage five times across the week, catch me at:  - TokenizeThis 2026 by RedStone on October 6: luma.com/sk9weqs5 - Blockworks Digital Asset Summit on October 7: lnkd.in/dA3ceSJ5  - Agentic Finance by More & More on October 8: lnkd.in/d8rNUsVN - HSC Asset Management Conference Singapore on October 8: lnkd.in/dJRUXZSM  - RWA SUMMIT on October 9 as part of RWA WEEK: luma.com/rwasummit Come meet the broader team on October 7 (9:00-12:00): Opening Bell, an institutional breakfast co-hosted with @trex_network and @KaikoData. 10 minutes from Marina Bay Sands and timed to kick off your TOKEN2049 day. Relaxed setting, high-quality conversations, and the right people in the room. Quality breakfast on us. RSVP: luma.com/ht5jyi8k Zth.
6
9
1
79
2,793
Don’t think NVIDIA is your friend. Last quarter, $89 billion of its $96.2 billion in revenue—92.5%—came from its Data Center business. Just 7.5% came from everything else combined—including you, the NVIDIA-loving consumer. NVIDIA started out building consumer gaming cards, but the reality is that it couldn’t care less about that business anymore. Follow the money. Don’t confuse a company’s biggest commercial opportunity with a mission to put powerful AI in your hands. @JensenHuang makes nearly all his money from Big Corp. He could absolutely put powerful GPUs with TEE (Trusted Execution Environment) support into normal people’s hands. He could let us build our own secure, community-run neoclouds and make compute cheap for the masses. He could level the playing field and still get massively rich. But instead, NVIDIA bends the knee to keep Big Corp happy. It has announced a 64GB DGX Spark starting at US$4,999—half the memory of the older 128GB version, with a roughly $5,000 entry fee. That’s more than the 128GB version cost at launch in October 2025. It’s worse than the older version in every way—and more expensive. To me, that’s an insult, not an exciting vision of accessible personal AI. My concern is simple: if capable local hardware stays expensive and memory-constrained, ordinary people will forever be renting their intelligence from Big Corp—and subject to whatever rules it imposes. Does that prove NVIDIA deliberately designed it to keep you dependent? No, the GPU and memory crunch is real. But I’m not going to mistake their incentives for mine. Apple truly has an opening here. Keep high-memory Mac Studios reasonably priced. Make powerful local AI something more people can own—not another service they need permission to use. Apple doesn’t need to become a charity. Make empowering the individual a profitable business—just like they've always done. Earn our money by helping us need the cloud less—just as you chose not to milk your users’ privacy for profit. I repeat: NVIDIA is not your friend. Apple isn’t automatically your friend either. The company I’ll back is the one that puts more intelligence under our control—not the one that makes me more dependent on Big Corp. We’ll see how this unfolds. We’re looking hopefully in your direction, @johnternus.
11
3
4
40
3,483
Matt Page retweeted
Bybit team is standby to help in any ways that we can. Bitget helped us when we had the hack. we are updating lazarusbounty.com/en/ to help Bitget to capture and trace the stolen fund movement.
Here is what we can confirm at this stage: On the attack: Our security team has made initial progress in tracing the source. The attacker compromised a critical backend system within our wallet infrastructure, used it to spoof transaction data, and triggered our authorization process to move funds out. Private key compromise has been ruled out — this excludes the more severe risk scenarios. Loss containment is confirmed. No further unauthorized transfers are possible. The specific method of system intrusion remains under active investigation. A full technical report will follow once confirmed. On withdrawal restoration: Multiple technical teams are working in parallel on system remediation and security hardening. Withdrawal restoration is being prepared in parallel. We will announce a timeline as soon as one is confirmed — we will not commit to a window we cannot guarantee.
348
296
128
3,876
756,399
Is crypto back? Everyone who quietly rebranded to AI last year seems to think so. Some thoughts on the market, and on @RealVision. Tis the season of humble brags on X, so here's mine. In reality, this was a pretty tough bear market. Crypto had to face the reality of a big drawdown outside of a general equities sell-off. But dogged analysis while the poo was flying thick and fast generated targeted calls and positioning insights that took the portfolio I run for the Pro tier up 2x in 2 months. I don't overtrade. I make high-conviction calls. My primary calls: 1. Derive ethereum:0xb1d1eae60eea9525032a6dcb4c1ce336a1de71be, up 5x from the March Deep Dive 2. Zcash $ZEC (2 trades), up 5x from the April Deep Dive 3. NEAR Protocol $NEAR, up 2x from the June Deep Dive 4. The SCP vs BTC, and the DeFi vs SCP outperformance call 5. In July, deployed 75% of the cash built up since the Q3 2025 top 6. Lousy call on Circle $CRCL, down 20% At @RealVision, my Pro members' portfolio compounded at 40% p.a. for 3 years, including the 47% drawdown. Importantly, it has outperformed both NDX and BTC. Now for the reality check. Altcoin open interest is at the 91st percentile versus Bitcoin. Things are frothy. But the first move out of a bear can be like that without derailing the cycle. This is not the time to get super aggressive or reach for leverage. Expect more volatility. I never like to overtrade the start of a cycle, but taking some profits makes sense. I want to hold Quality names for the medium to long term. (ps: I am bringing out a Quality index/basket to help narrow the field for Alpha tier subscribers.) On the macro front, I am still concerned about the general state of global liquidity. DXY is strong, the yield curve is flattening, and real rates are climbing. We are late stage in the 5-year refi cycle. Bessent blinked, but it will take more to get through this. Crypto could easily see a 30%+ pullback which means (50%+ for many high flyers). But for long-term investors who are now waking up and need to take their crypto allocation from nothing to something (depending on risk tolerance), dips are for buying quality names. The structural tailwinds are coalescing. Some of this move is ahead of the data, but that's generally how markets work. Assets are pricing in what the agent economy + tokenisation looks like in 12 months, not today. Just a short comment on @RealVision. We have been around for 13+ years. After being force-fed the bile of BBG and CNBC for 20 years in TradFi, I saw straight away how transformative the platform was back in 2014 and became a foundational subscriber. I was a subscriber for almost a decade before joining @RaoulGMI to build out the crypto product. It's been an incredible 3 years. Whether as a strategist or a subscriber, I am lucky each day to read the phenomenal macro, equity, tech, AI, and geopolitics analysis from the team of independent strategists team: Raoul, @AndreasSteno, @RosenvoldGeo, and @DMattin. That's 5 strategists who often agree but, more importantly, often disagree. There is no 'house' view. If that upsets you, go become a client of Morgan Stanley, JPM or some brokerage with a conflict of interest. My lane is crypto. That's what I love, and I have never been more bullish on the secular trend that inspired me to leave the TradFi world years ago. Brutal, undignified, sleep-destroying. Best job I have ever had.
32
25
356
23,788
Great exposition form @krunkosaurus on the current statements coming out around slowing down AI
There are now a lot of models sitting on the Pareto frontier. Each is better at some things and worse at others, but there is no longer one model that is obviously best at everything. What if we’re reaching a point where frontier intelligence is beginning to converge? Maybe that is the backroom conversation between Dario and Sam. Not AGI. Not a hard wall. Just a slower, more brutal climb upward—while smaller, cheaper MoE models close the gap much faster than the frontier labs can widen it. Imagine Sam and Dario looking at the scaling curves and reaching roughly the same conclusion: “We’re probably not going to make the next model 2–3x smarter than the current one. The frontier is converging around a similar level of intelligence, and from here the gains get increasingly expensive.” Maybe getting another true multiple in capability requires something absurd: a 100-trillion-parameter-class system, vastly more compute, vastly more data, and vastly more inference capacity. Even assuming you could train it, how do you serve it to hundreds of millions of people? They already struggle to economically serve today’s frontier models at global scale. A model requiring 10x the compute per useful answer would be a completely different problem. Viewed through that lens, all the talk about “slowing the frontier” starts to make more sense. Maybe the frontier isn’t being slowed voluntarily. Maybe frontier progress itself is slowing. The biggest labs can keep pushing enormous experimental systems that only they can afford to run, while the public models improve incrementally and DeepSeek, GLM, Qwen, and others keep compressing yesterday’s frontier intelligence into smaller and cheaper architectures. At some point, the gap closes faster than the frontier moves. And even if someone builds a model that is 10x more capable, there may simply not be enough compute in the world to serve that level of intelligence to billions of people. The next phase of AI may not be one lab racing infinitely ahead. It may be intelligence converging—and becoming cheaper.
15
The OpenBuilder event here in Singapore last night was probably one of the best MiniMax H3-focused events in the world. What you can do with an open-source video model is incredible and game changing in both price and performance. My full presentation in the comments.
4
2
1
16
1,361
In just the past few weeks MiniMax H3 has changed what’s possible with open-source AI video. Team Sogni will be at SGInnovate alongside @MiniMax_AI, presenting: “1,000+ Hours of H3: What We Learned Making It Sogni’s #1 Video Model” We will go over how we scaled H3 on our community GPU network, tips and tricks on getting the most impressive renders, and so much more. If you’re in Singapore, come join us. Registration is free, but space is limited. Sign up through the event link:
2
2
11
497
Elizabeth Holmes You Can See Everything (Coraline edition) Made with Minimax H3 + Sogni Creative agent skill
1
1
1
11
4,719
this make-money-v0 trading agent ... makes money! +$2,996.72 in 17 days 70% win rate
42
13
8
333
55,294
Matt Page retweeted
The first atomically composable EVM ever. @ZenithFdn x @CantonNetwork One ledger, two languages. The Canton EVM. Zenith. Zth.
16
18
6
91
6,514
Matt Page retweeted
🚨 If you learn one thing today, let it be this 👇 Zenith built THE FIRST atomically composable EVM. No other network has ever added a new VM and made it truly atomically composable with its existing one. Subnets, appchains, rollups, validiums, sidechains, prividiums… none of them achieve this with the parent chain. They settle asynchronously, rely on bridges or message passing, introduce latency windows, trust assumptions, or partial-failure risk. Assets and logic live in silos. Liquidity fragments. Value leaks. Composability is a farce. @ZenithFdn flips that model completely. When you ship on Zenith, you ship on @CantonNetwork… not on top of it, not extracting from it. This is fundamentally new architecture that alters how L1 scaling works. Traditional scaling pushes execution off the L1 into isolated environments that trade atomicity for throughput. This is what Ethereum exemplifies today: fragmented liquidity, broken multi-leg operations, and economic value leaked to L2s that never returns to Ethereum all glued together with fragile economic bridges. Zenith scales by extending the L1 itself with a new VM that remains fully native. You get Ethereum-compatible reliability and developer experience while preserving Canton’s privacy, compliance, and crucially, true atomic composability across these environments. A single transaction can lock a Canton-native tokenized Treasury, invoke a Solidity lending vault on Zenith, mint an ERC-3643 receipt, and settle everything in one indivisible operation under the same consensus. Liquidity stays unified. Risk is compressed. Capital velocity increases. Daml and EVM finally operate as one system instead of two parallel universes. This is not just better interop. This is the first atomically composable EVM. 
This is the Zenith. Zth.
Our Head of Product @NorbertVadas just joined the @indexed_pod for a wide-ranging technical conversation about what Zenith is building on @CantonNetwork and what it means for institutional DeFi. If you want to hear Norbert walk through the architecture directly -why EVM compatibility matters for Canton, how CIP-56 assets work, what atomic composability between Canton and Zenith actually looks like in practice, and where the roadmap goes from here - this is the episode worth setting aside time for. A few timestamps worth jumping to if you don't have the full hour: (00:51) What is Zenith? (17:14) Why EVM compatibility? (39:13) RWA lending example (42:15) Bringing DeFi to Canton (56:46) Who is using Zenith? Zth.
11
7
3
58
4,320
Matt Page retweeted
We are proud to announce that SupraNova now moves assets in both directions. Ethereum → Supra. Supra → Ethereum. ✅No multisig. ✅No external bridge operator. 🔒Every transfer is verified against Ethereum's own consensus.
🤖 Made with AI
7
44
2
203
4,738
Matt Page retweeted
Looks native. 🙈 Sounds native. 🙉 Must be native. 🙊 Atomic by design, nuclear by impact. Zth.
Every EVM transaction on Zenith routes through Canton's Global Synchronizer as native network activity. Fees stay within Canton's economic model, and apps built on Zenith participate in the same reward pool as other Canton-native apps, based on usage.
8
8
63
2,467
Matt Page retweeted
Back in 2021, we established our options division. Over the years, the product has gone through many changes and upgrades. Today, Bybit Options offers some of the deepest liquidity in the market and is portfolio margined alongside our other product lines, including our upcoming TradFi stocks and stock options offerings. Stay tuned. We keep pushing.
A milestone worth celebrating, and we’re only getting started. Bybit Options now ranks No. 2 globally, capturing 22.4% of options trading volume in H1 2026 and reaching 27.7% in June. This reflects the relentless work of our teams and the trust of our users. Next target: No. 1.
52
19
2
200
45,736