Mark

90 posts

Mark banner
Mark

Mark

@MachSpeedx0

AI enthusiast and blockchain fanatic | NFA, DYOR

Katılım Haziran 2022
870 Takip Edilen89 Takipçiler
Mark
Mark@MachSpeedx0·
Everyone read the headline as a hold. Three governors actually voted to put rates up, and the 30 year yield rose about 9bp. A no change decision with votes leaning higher is what people mean by a hawkish hold. Bonds took it seriously, bitcoin and ether rose anyway
English
0
0
0
3
Mark
Mark@MachSpeedx0·
Everyone I spoke to today read the hold as dovish. The tape read it the other way, 2s bid and 30s offered, which is a market deciding inflation gets tolerated for longer. That long end is what crypto discounts against, so a no change meeting can still tighten
English
0
0
0
11
LinkedIn Lunatics
LinkedIn Lunatics@LinkedInLunat1c·
I checked his post history…this guy is being serious
LinkedIn Lunatics tweet media
English
126
68
2.5K
209.6K
Mark retweetledi
Kay Capital
Kay Capital@keyahayek·
历史反复验证:韩国永远是全球金融危机的第一预警器。 1997 年亚洲金融风暴:韩元率先断崖贬值,KOSPI 暴跌,外汇储备几乎耗尽;随后风险迅速传导至全球市场。 2000 年互联网泡沫:韩国半导体板块提前见顶回落,几个月后纳斯达克崩盘 78%,科技泡沫彻底破裂。 2008 年次贷危机:韩股率先进入加速下跌,一个多月后雷曼破产,全球金融体系当场宕机。 后来韩国人发现,只靠韩国市场发送预警,效率还是太低,于是开始主动出海: 2021 年,Bill Hwang 通过 Archegos 把杠杆塞进总收益互换,一次爆仓让全球投行损失上百亿美元,其中瑞信单家亏掉55亿美元,开启了中概一路趴到现在的大熊市。 2022 年,Do Kwon 发明的 LUNA 和 UST 梯云纵 400 多亿美元市值几天内灰飞烟灭,随后三箭资本、Celsius 等机构连环爆雷,半年后 FTX 倒闭。 2025 年,韩国危机预警系统进一步完成了产业化升级。据说南方东英最初只给三星电子在香港市场推出了两倍杠杆产品 7747。SK 海力士看完很不高兴,主动找上门问: 「为什么三星有,我没有?」 南方东英从善如流,随后推出 7709,满足了韩国企业在金融创新领域不患杠杆不足,患杠杆不均的合理诉求。 结果以 7709 和 7747 为首的两支杠杆兵团迅速做大:SK海力士相关产品最高管理规模约 230 亿美元,三星相关产品约 80 亿美元,合计超过 300 亿美元。其中 7709 单只产品规模一度超过 200 亿美元,成为全球最大的单一个股 ETN 杠杆产品。 这批产品每天自动再平衡: 股票上涨,它们必须继续买入; 股票下跌,它们必须被迫卖出。 别人投资讲究低买高卖,韩国金融创新实现了高买低卖全自动化。 偏偏三星和 SK 海力士又占据 KOSPI 超过一半的权重。于是两员大将上涨时负责把韩国股市顶上天,下跌时负责按照两倍杠杆纪律集体砸盘。一次暴跌,仅 SK 海力士相关杠杆基金就能制造约 50 亿美元的机械性卖盘。 最终,KOSPI 从 9385 点一路被干到 6000 点附近,一个多月回撤约 36%,期间反复熔断,单日盘中再跌超过 11%。
中文
190
308
1.8K
527.7K
Mark
Mark@MachSpeedx0·
@tengyanAI Advanced packaging capacity is sold out well ahead, so the margin line partly reflects contracts signed a year or two back. Mix can improve while price is still stuck at yesterday's negotiation. I would watch capex commentary more than tonight's gross margin
English
0
0
0
6
Teng Yan
Teng Yan@tengyanAI·
Amkor ($AMKR) reports earnings tonight. the interesting line is probably the beat. it’s whether AI is changing the economics of the business. Gross margin expanding with advanced packaging mix → HBM / CoWoS adjacency is flowing into the P&L. Gross margin staying flat → AI exposure is driving volume, not value.
English
7
0
8
3.3K
Mark retweetledi
Yarchi
Yarchi@undefinedKi·
Loop engineering and graph engineering, explained like you are five. Everyone throws the words around, almost nobody defines them A loop is one worker with a to-do list. It picks a task, does it, looks at what happened, picks the next one. Everything it knows lives in one conversation. When that conversation fills up, things start falling out of it. A graph is an assembly line. Each station does one job. Between stations the work gets saved to disk. If station four fails, you rerun station four, not the whole line. Three questions decide which one you need: 1. Will you do this a hundred more times? If not, use the loop. Building the line costs more than doing the work by hand 2. Can a script tell you the output is correct? If only your taste can judge it, the checkpoints are decoration 3. Do you know the steps yet? If not, the loop will find a path you would never have drawn The loop is not dead, it got demoted. Every station on the line still has a worker running a loop inside it. Bookmark this
Yarchi tweet media
Yarchi@undefinedKi

x.com/i/article/2080…

English
21
140
792
70K
Mark
Mark@MachSpeedx0·
@Whdysseus 野村的17.13美元建立在20倍市盈率的基础上,是美光估值的两倍,但DRAM属于周期股,在周期顶部倍数本应压缩而非扩张。空头看2美元,不是看错了公司基本面,而是拒绝给一只周期股支付成长股的估值倍数。
中文
0
0
0
3
大晖哥Whdysseus
大晖哥Whdysseus@Whdysseus·
长鑫科技(CXMT)7月27日(周一)上市,多空分歧炸了 投行给出的目标股价达到17.13美元,市值约合1.15万亿美元(7.76万亿),有真金白银的玩家股价最低看到2美元。 Hyperliquid链上最大空头:看跌到2-4.5美元 上市首日,长鑫科技在Hyperliquid上的合约最大持仓地址仍在持续加仓空单,当前仓位已达1726万美元,开空均价6.4美元(约43元)。 就在刚才,他在2美元-4.5美元(约13-30元)挂上了平仓单。 野村证券:目标股价17.13美元(约116元),市值看到约1.15万亿美元(7.76万亿元)。 同一天,野村证券启动对长鑫科技的覆盖,给予买入评级,目标价17.13美元(约116元)。 对应20倍市盈率,是美光估值的两倍。 从IPO价格1.28美元(8.66元)算,上涨空间1239%,对应市值约1.15万亿美元(7.76万亿元)。 一方看到2美元,一方看到17.13美元。 同一个标的,完全不同的剧本。多空对决,才刚刚开始。 #CXMT
中文
2
0
3
860
Mark
Mark@MachSpeedx0·
@karlmehta The interesting thing is the ~$7bn need, showing that in 2014 vs now, the capital gap did not close and are only going up
English
0
0
0
3
Karl Mehta
Karl Mehta@karlmehta·
Demis Hassabis says he sold DeepMind in 2014 because staying independent needed $7 billion, and he could barely raise $10 million: "He saw Google as an AI company. The reason I sold was I knew it would require 7 billion dollars of investment in order to take on, try and become one of those companies." "And I was barely able to scrape together 10 million dollar rounds at that point." "And the problem was is that we'd done enough, the Atari thing, we had alerted enough of the Silicon Valley titans. They were coming after my people." "We had no business, we didn't even have a product, obviously not, because I was trying to make the core technology." "I was paying myself zero and people were being paid like 100k a year and then they were getting offers of 10 million a year." "And then if that didn't work, they could afford to come back and say, here's 50 million a year." "So at some point, that was going to go wrong." P.S: I'm hosting the AI Assurance & Governance Summit 2026 on October 1 at the Stanford Faculty Club, one day and one track, to work on the part of this that money did not solve. Frontier labs, regulated industries, insurers and investors in the room. Link in the reply. - Demis Hassabis (@demishassabis), co-founder and CEO of Google DeepMind, at the Worshipful Company of Information Technologists Annual Lecture.
English
43
98
1K
301.2K
Mark
Mark@MachSpeedx0·
@cao_lab Can't agree more... as all of that is now concentrated in a few hands, especially with AI outsmarting any individual humans in the forseeable future (if not already)
English
0
0
0
6
DrCAO | AIWeb3 | ComputeFlux
One Thought #018 We spent years decentralizing money. The next decade is about decentralizing intelligence.
English
1
0
2
113
Mark
Mark@MachSpeedx0·
@Whdysseus 沃什闭麦这一点被市场低估了。前瞻指引本身就是一种政策工具,取消它等于把定价工作丢还给市场,波动率因此成为传导渠道之一。7月加息概率从约10%跳升至约33%,部分反映基本面变化,部分则只是信息真空中的重新揣测。
中文
0
0
0
2
大晖哥Whdysseus
大晖哥Whdysseus@Whdysseus·
“美联储传声筒”放话:这次真的很难猜 Nick Timiraos最新判断——7月28-29日这次议息会议,可能是近年来最难预测的一次。 五个关键变量: 油价重新抬头——通胀压力又被点燃 关税风险升温——政策不确定性在添堵 部分官员转向加息——按兵不动的共识开始松动 市场预期跳跃——7月加息概率从约10%跳到约33% 18位官员严重分裂——一半人预期加息,一半人认为没必要 美联储主席沃什继续闭麦 上任后就是惜字如金,拒绝任何前瞻指引。投资者只能从其他官员发言里“连蒙带猜”。 多空叠加+官员分歧+主席沉默,确实是这几年少见的“迷雾局”。
中文
1
0
0
468
Mark
Mark@MachSpeedx0·
Competition is a great thing - thanks Moonshot AI and OpenAI for Opus 5
English
0
0
1
28
Mark
Mark@MachSpeedx0·
Winding down financial engineering is good for the industry imo
matthew sigel, recovering CFA@matthew_sigel

List of DATs who have abandoned Bitcoin and crypto accumulation strategies since: 🔴 Full Liquidations / Complete Exits Satsuma Technology (SATS LN) — Shareholders voted on July 21, 2026 (>>90% approval) to liquid all 668 BTC (~$43.5M), return capital to investors, and delist from the London Stock Exchange. Pantera Capital Management was reportedly among investors pushing for the wind-down. Bitdeer (BTDR US) — Fully emptied its corporate Bitcoin treasury as of February 20, 2026, liquidating its remaining 943 BTC to fund its pivot to AI data centers. The company had also sold 798 BTC in December 2025. CEO Jihan Wu stated the zero balance "will not always be zero in the future." Sequans Communications (SQNS US) — Sold 1,025 BTC in Q1 2026 to fund convertible debt redemptions and an ADS buyback, reducing holdings from 2,139 BTC to 1,114 BTC. Subsequently sold nearly 80% of remaining holdings to fully redeem all convertible debt. Sequans has explicitly stated it will not acquire more Bitcoin and is refocusing on its core IoT semiconductor business, with plans to monetize its remaining 658 BTC in a disciplined manner over coming quarters. Genius Group (GNS US) — Liquidated its entire Bitcoin treasury in early April 2026 to repay approximately $8.5M in debt. The company had previously reduced holdings from 200 to 138 BTC in November 2025 to cover short-term cash needs, with plans to recommence Bitcoin purchases at end of 2026. Prenetics (PRE US) — Fully liquidated its entire ~510 BTC treasury (~$41.3M in proceeds) in May 2026 and adopted a policy prohibiting any future digital asset purchases. Vaultz Capital (V3TC AQ) — Shareholders voted on July 21, 2026 to withdraw its Bitcoin treasury policy and dispose of all Bitcoin holdings, repositioning as a cash-backed acquisition platform. Exact BTC quantity not disclosed. Alpha Compute (ALP US) — Completed a full wind-down of its digital asset treasury on July 10, 2026, returning the final tranche of ~$6M in TON/GRAM holdings to Animoca Brands affiliates. No longer holds any digital assets. AEG (AEG LN) — Fully liquidated its entire digital asset holdings post year-end, realizing approximately £97,945, converting proceeds back to working capital. MAIA Biotechnology (MAIA US) — Digital asset treasury strategy placed on hold due to cryptocurrency volatility; holds approximately $0 in digital assets as of March 23, 2026. 🟠 Partial / Ongoing / Forced Sales MARA Holdings (MARA US) — Sold approximately 15,133 BTC (~$1.1B) in March 2026 to repurchase over $1B of face value of its Senior Convertible Notes and reduce its line of credit by $200M. MARA has formally expanded its digital asset management strategy to allow for the sale of Bitcoin held on its balance sheet — not just current production — to enhance financial flexibility and fund capital projects. Empery Digital (EMPD US) — Has been systematically selling Bitcoin since at least March 2026, reportedly selling nearly half of its BTC holdings for approximately $87M as of July 10, 2026, to fund a $200M share repurchase program and repay a term loan. Disclosed weekly tranches include: Strategy / MicroStrategy (MSTR US) — Sold 32 BTC (~$2.5M) between May 26–31, 2026 to fund preferred stock distributions. The board authorized a BTC Monetization Program of up to $1.25B to generate USD reserves, with a minimum USD reserve policy of at least 12 months coverage. Between June 29–July 5, 2026, sold a further 3,588 BTC (~$135.2M), bringing total sales to approximately 3,620 BTC (~$137.7M combined). Michael Saylor stated at BTC Prague: "I never said the company wouldn't sell its Bitcoin." Nakamoto Inc. (NAKA US) — Sold approximately 284 BTC for $20M in March 2026, with proceeds used for working capital and to fund operations following its acquisitions of BTC Inc. and UTXO Management. Additionally sold ~40 BTC received as premium income from its derivatives program. As of December 31, 2025, approximately 3,717 of its 5,342 BTC were pledged as collateral for a Kraken loan maturing December 4, 2026 — a potential binary event. Smarter Web Company (SWC LN) — Sold 178 BTC on July 23, 2026 at an average price of $65,762 (~$11.7M) to repay its "Smarter Convert" convertible instrument to the TOBAM Group ahead of maturity. Cango Inc. (CANG US) — Completed a Bitcoin sale in February 2026 to strengthen its financial position and advance an AI transformation strategy. Exact amount not disclosed. DIGI (DIGI CN) — Sold its entire 118,000 XRP position and pivoted its digital asset strategy to Bitcoin accumulation, holding 10.88 BTC as of February 28, 2026. 🟢 Strategy Shifts (Reduced Accumulation / Active Management) Exodus Movement (EXOD US) — Has been materially reducing its ETH holdings: from 2,742 ETH (July 2025) to 1,840 ETH (February 2026) to 1,433 ETH (May 2026). MPU Capital (MPU US) — Shifted primary treasury emphasis away from BTC/ETH to stablecoin governance tokens in August 2025; may continue to buy, hold, or sell BTC and ETH. ZeroStar (ZSTK US) — Treasury policy explicitly contemplates periodically selling cryptocurrency for general corporate purposes. DigitalX (DCC AU) — Moved away from pure accumulation to actively managing Bitcoin holdings within a disciplined capital allocation framework.

English
0
0
0
22
Mark
Mark@MachSpeedx0·
@zengjiajun_eth Hmm interesting, gonna do it rn. Are there specific verticals I should train these small models for to get maximum quality output from them?
English
1
0
0
78
Zeng Jiajun 曾嘉俊
Zeng Jiajun 曾嘉俊@zengjiajun_eth·
I would strongly suggest everyone to go through the process from collecting data, processing data, pre-training, SFT, to RL and RLHF. You can do it in one day together with Claude code, depending on how hand on you wanna be. It will blow your mind. For small model you don’t even need to rent GPU. Renting GPUs will certainly help u understand scaling laws better with bigger model.
Kevin Simback 🍷@KSimback

If you haven’t post-trained a model, you definitely should Pick a use case -> run evals -> build training data set -> rent a GPU -> train -> run evals

English
2
0
7
1.4K
Mark retweetledi
Yield Maxi
Yield Maxi@0xYield_Maxi·
BlackRock, Strategy and Coinbase just formed a BTC Security Consortium: $15 million to prepare Bitcoin for quantum computing. Bitcoin's biggest challenge is getting what it needs to be solved. The institutions aren't selling. They're building.
Yield Maxi tweet media
English
1
1
8
120
Mark
Mark@MachSpeedx0·
@Sheng5Wu May I add “being intimidated by CLI interfaces like Claude Code, Codex or things like Hermes” I think those environments are the ones that give people wings
English
1
0
1
39
Sheng
Sheng@Sheng5Wu·
3 common mistakes people make when using AI 1/ Vague prompts: This is the most common mistake I see when people use AI. Writing vague prompts and then blaming AI is like giving poor instructions to your employees and saying they are not competent enough. Identify your task, choose the right AI tool and learn prompt engineering to get the best out of every tool.
English
2
0
6
167
Mark
Mark@MachSpeedx0·
End of an era
BitMEX@BitMEX

Dear BitMEX Users, Today, we share with a very heavy heart that BitMEX exchange will shut down its operations, effective 23 September 2026 at 04:00:00 UTC. The owner and operator of BitMEX, HDR Global Trading Limited, has made the difficult decision to close operations following a strategic review of the business. It may not look the same today, but we are proud of our 11+ year legacy and the role we played in shaping the crypto industry. We invented the 100x leverage perpetual swap, which for most of you, was the first step to your crypto trading journey. It is now the most traded financial product in the crypto industry, adopted by thousands of users and exchanges. And we remain proud of our robust security infrastructure, which has allowed us to maintain a flawless track record of 0 customer funds lost to hacks in our entire operating history. We want to reassure you that your assets remain fully safe and under your control during this transition period. This announcement is just to give enough time to ensure a smooth withdrawal process for everyone. From today we strongly encourage all users to close their positions and withdraw their funds as soon as convenient. For more details on the full process, please read our blog: bitmex.com/blog/bitmex-cl… BitMEX was once home to some of the greatest traders today. Our team has dedicated tremendous effort and passion into building the platform into what it is, and we are glad to have reached some of you during your time with us. To everyone who has traded, supported, and grown alongside us - thank you for your trust over the last 11 years. The BitMEX Team

English
0
0
0
31
Mark retweetledi
Yield Maxi
Yield Maxi@0xYield_Maxi·
Morpho just launched fixed-rate lending on Base Fixed-rate lending on cbBTC: ~4% APY. Fixed-rate Prodigy.fi vaults on cbBTC: 12%+ APY The difference? You're not just lending — you're earning option premiums from traders who want exposure to price moves. That's what Prodigy.fi does. Same chain. Higher yield. Different mechanism.
Morpho 🦋@Morpho

Predictability, now in your hands Introducing Morpho Fixed Rate Markets on the new Markets App Lend and borrow directly from fixed rate, fixed term markets

English
0
2
7
145
Mark
Mark@MachSpeedx0·
@cryptopunk7213 haha meta probably really wants the training-data. The router can logs which model wins for which prompt type, which is exactly the dataset they need to distil a cheaper in-house model
English
0
0
1
23
Ejaaz
Ejaaz@cryptopunk7213·
meta is building a competitor to OpenRouter (coming right after news openrouter might get acquired), but theres another reason meta's doing this: > no more $100Ms wasted on inference. the platform would efficiently route ai prompts to multiple different models, saving on costs and produce better answers. > brand new revenue stream: Meta's intention is to release this publicly for anyone to save money > powerful new way to train their models: used internally, meta would create a very valuable dataset of how/when to use models for various purposes. zuck's still in this race
Ejaaz tweet mediaEjaaz tweet media
English
8
0
26
3.9K
Mark
Mark@MachSpeedx0·
@Route2FI Tbh, I think the edge in poker and forex (or even crypto) isn't loving risk, it's being the one sizing it correctly whilst the table loves it blindly
English
0
0
0
91
Route 2 FI
Route 2 FI@Route2FI·
I've always been attracted to money games like poker, sports betting, stocks, forex, and, eventually, crypto. I think what all of us in crypto have in common is that we love probability games & taking risks. Being self-sustainable & 100% responsible for your own future is priceless.
English
30
11
193
16.9K