I ASKED MY CHINESE QUANT FRIEND HOW HE MAKES $75,000/MONTH He sent me this MIT lecture. 77 minutes on random walks, Markov chains, and why a trade can look attractive while having zero edge. The example at 23:00 is worth your attention: In a fair random-walk model, you aim to win $100 and stop if you lose $50. That's a 2:1 reward-to-risk ratio. But the probability of reaching +$100 first is only 1/3. (1/3 × $100) − (2/3 × $50) = $0 before costs. A bigger profit target doesn't create an edge. The probabilities still have to work. Three places to start: • 10:28 — Simple random walks • 23:00 — The $100 / $50 stopping problem • 32:05 — Markov chains: what the current state tells you MIT 18.S096 · Lecture 5 · Dr. Choongbum Lee. Bookmark this. Work through the $100 / $50 example before your next backtest.

1995. MARC ANDREESSEN MAKES $58M OVERNIGHT AT AGE 24... AND IMMEDIATELY PICKS A FIGHT WITH THE RICHEST MAN ON EARTH august 9, 1995. the opening bell rings. Netscape is 16 months old, has zero profit, and goes public anyway. by sunset it's worth $2.9B. a 24-year-old kid from rural Wisconsin wakes up worth $58M. Time magazine puts him on the cover sitting barefoot on a literal throne. just two years earlier, he was making $6.85 an hour at a university lab in Illinois. eating pizza, drinking soda at 3am, writing 9,000 lines of code to put images on a screen for the first time. then he drops the line that wakes up a sleeping giant: "Windows will become just a set of poorly debugged device drivers" Bill Gates hears it. Microsoft declares total war. Gates makes Internet Explorer free and bundles it into every copy of Windows on earth. Netscape gets crushed from 90% market share to dust. AOL buys the scraps. most 27-year-olds with tens of millions in the bank would retire to a beach and tell war stories for the rest of their lives. Andreessen didn't leave the table. he teamed up with Ben Horowitz and pushed his chips right back into the fire: a cloud startup called Loudcloud in 1999. timing? catastrophic. the dot-com bubble explodes. customers go bankrupt. the stock tanks to $2. the press writes their obituary every single week. to survive, they sell off the core business, pivot into enterprise software as Opsware, stare down bankruptcy for seven straight years... and sell it to HP for $1.6 BILLION in cash. that near-death torture taught him one thing: Wall Street suits in boardrooms have no idea how to build a company. they panic and fire the engineers when things get hard. so in 2009, Marc and Ben put up $300M to start Andreessen Horowitz (a16z) with one radical rule: never fire the technical founder. two years later, when the world still thought tech was just another industry, he wrote six words that predicted the next two decades: "software is eating the world" napkin math: the college job: $6.85/hour in 1993 the first IPO: $58M at age 24 the empire built from the ashes: $45B+ under management. early checks into Facebook, GitHub, Coinbase, Stripe, Solana, OpenAI Gates won the browser war. the kid making $6.85 an hour ended up funding the entire modern internet.

i'm sorry, ARGON-4 can WHAT? it can replace an entire 8-person quant team and run an autonomous hedge fund loop locally: mining, killing and deploying strategies while you sleep? since when?? i've been staring at charts and manually backtesting for YEARS. read guide below👇

Entropy will change our world > volume: $1.5B > launched: on Aug 24 > funding: $14M raised link: entropy.io/?r=zerqfer

Years of overbuilt crypto rails are finally monetizing through the application layer, turning products like stablecoins and tokenized assets into institutional revenue drivers he real venture upside now concentrates in adjacent financial primitives: credit, FX, and custody layers underwritten before mainstream traction shows up in the data.

sparsity is finally available for everyone sparsitydata.com/

early crypto projects - @spicenetio - @D3inc - @QFEX keep your eye on them

so true If you had bought $SOL in June for $60, you would already have 2x by now

The most interesting metric here: day-60 leaders still produced a median 15.9x gain after already surviving two separate 50% drawdowns. Deep pullbacks reset the funding rates and shift supply from fast rotators to long-term holders who build the next floor. So when you track genuine cultural leaders on Solana or Robinhood Chain, drawdowns are normal market structure. The real expansion phase often starts right after crash number two.

someone pls ddos me

aave just turned tesla and nvidia into defi collateral. the oracle takes weekends off. live since friday on base. 7 coinbase tokenized stocks, borrow usdc against them. here's how i'd actually use it 1. discovery what went live: - aapl, amzn, googl, meta, msft, nvda, tsla (the coinbase b20 tokens) - collateral only. you deposit stock, you borrow usdc, that is the whole product - collateral factors 65-79% depending on the ticker - caps are small on purpose: ~$29m collateral, $32m usdc supply, $21m usdc borrow - non-us only, issued offshore, no us securities registration you cannot borrow one stock against another. cross collateral is off. 2. research before you size anything, go read the feed schedule. chainlink's tokenized equity feeds publish sunday evening to friday evening et. outside that window the feed holds friday's close. no heartbeat off hours. the contract still answers you, it just answers with a stale number. so the token keeps trading on aerodrome all weekend while aave values it at friday close. someone measured this properly back in august across 4 aerodrome usdc pools: every weekend gap came in under 0.6%, with ~$6m displayed liquidity and ~$7m of 24h volume. tight spreads, but that is $6m of depth, not $600m. what i look for before depositing: - whether that ticker has earnings inside my loan window - pool depth for the specific token i'm holding, not the basket - my liquidation price measured against friday close, not against the live dex price 3. execution how i'd open it: > deposit on base, borrow usdc, keep the first one small - start at half the max ltv. 79% available does not mean 79% used - on a 70% collateral factor, borrowing at 35% leaves room for a 50% drawdown - borrow the usdc and park it. do not lever it into a memecoin on friday afternoon - write your liquidation price down before you close the tab skip rules: - skip any ticker with earnings before your next check in - skip max ltv on anything that isn't msft or aapl - skip opening fresh friday after the close. you are flying blind until sunday evening - skip the whole thing if you can't look at the position monday morning 4. management monday open is the event. the rest is waiting. - friday: deleverage before the close, not after - weekend: the number on your dashboard is not a live price - sunday evening: feed wakes up, your real ltv shows up with it - monday: if the stock gapped down 8% overnight, your ltv moves 8% in a single print there's no slow bleed into liquidation here. it arrives as a jump. risks worth saying out loud - this market is one day old, small caps, fresh risk params, no track record - stale oracle plus thin weekend liquidity is the core problem, everything else is detail - dividends auto reinvest after fees and withholding, splits go through an onchain multiplier. corporate actions are a brand new surface onchain - offshore issuer, non-us only. know what you actually hold - aave v4 hub and spoke is new code carrying this the main takeaway: your ltv on friday is not your ltv on monday, and the difference lands as one print instead of a slide you can react to. size for the gap, not for the spread. want a part two after monday's open, with where the ltvs actually landed and whether anything got liquidated?

kalshi vs beni

just woke up and found out that bitget was hacked for $350M

I just crossed 300 Variational points. At the current OTC rate ($24/pt), that’s over $7,500+ locked in Here is how you milk this before you get diluted to zero: 1. Hedge everything: Run delta-neutral positions across Arbitrum perps so you bleed $0 while generating $1M+ in volume. 2. Farm off-hours: The 150k weekly pool is zero-sum. Farm on Sunday nights when volume is dead to get 2x points per $1 volume. 3. Sybil the base tier: 1 point qualifies a wallet. Run 15 sub-wallets with small volume to capture the Genesis tier bonuses. Farming costs me $6.50. Whales are begging to buy my points off-market at $25. Do the fucking math. Stop scrolling and start printing before the crowd wakes up.

btc is 34% off the october 2025 high. eth is 46% off its own eth is in a real bear. btc is in a drawdown that already bounced 20% off the 69k june low. and capital kept rotating into ai names the entire way down.

传统钓鱼偷的是你某一个代币的额度。7702 钓鱼偷的是整个地址的执行权。 USENIX 上的一份研究把样本里的 EIP-7702 授权交易拉出来看，63% 指向恶意合约，已确认被盗超过 230 万美元。 这东西的签名弹窗，长得跟普通签名几乎一样。我的检查流程如下。 1. 先搞清楚它和 approve 差在哪 - approve：你给出的是某一个代币的额度。想转别的，还得再弹一次 - 7702 委托：一笔 0x04 类型的交易，把 0xef0100 加上一个合约地址，写进你 EOA 的代码位 - 写进去之后，你的地址就能跑那个合约的任意逻辑。批量把代币和 NFT 一次性扫走，全程不再弹窗 额度被盗，你丢的是一个币。委托被盗，你丢的是这个地址的控制流。 2. 钓鱼话术长什么样 - 常见包装：钱包升级、安全增强、开启智能账户、一键领取 - 弹出来的不是 approve，是一个 authorization 授权签名 - 最阴的一点：授权里的 chain_id 可以填 0。按规范，填 0 的授权在所有链上都有效 - 你以为只在一条链上签了一下，实际是全链通行 3. 签之前我看三样东西 - 交易类型：出现 0x04、SET_CODE、delegation、智能账户 这些字样，先停 - 被委托的合约地址：是不是你钱包官方的实现合约。浏览器上没开源的，直接 pass - chain_id：不是当前链的具体 ID 而是 0，直接 pass 跳过规则： - 要求你先升级钱包才能领空投 = 跳过。没有任何正经空投需要你改账户类型 - 陌生站点弹委托签名 = 跳过 - 弹窗文案看不懂在授权什么 = 跳过 4. 怎么查自己有没有被委托 - 打开 revoke.cash，账户页里有个 delegations 标签，直接看 - 或者查你地址的 code 字段。普通 EOA 应该是空的 - 如果返回 0xef0100 开头的 23 字节，就是有委托。后面那 20 字节就是被委托的合约地址 主力地址建议现在就查一遍，委托是持久的，不会自己过期。 5. 怎么解除 - revoke(.)cash 只能看，解不了。要回钱包里解 - MetaMask：账户详情里把 Enable Smart Contract Account 关掉，每条链要分别关 - 协议层面的做法是再签一次授权，地址填全零地址，把 code 清空 6. 风险和边界，说实话 - 7702 本身不是漏洞，是 Pectra 升级里的正常功能。很多钱包靠它做批量交易和 gas 代付 - 真正的问题是签名界面，把交出执行权这件事显示得像一次普通签名 - Inferno Drainer 那套改过的工具，半年内从三万多个地址卷走九百万美元以上 - 我自己的做法：7702 只在日常小号上开，冷钱包保持最朴素的 EOA，一行代码都不写进去 核心要点：approve 是借出一把钥匙，7702 委托是把锁换成对方的。弹窗里只要出现升级、智能账户这类词，先停下来，把被委托的地址看清楚再说。 要不要写个第二篇，带你一步一步查主力地址的历史委托记录？

a memecoin paired against tokenized nvidia is two trades in one. most people only notice the second one when it goes against them. robinhood chain had 400+ stock-paired pools by early september, around $8.8m of stock tokens doing $95m in daily volume. your meme doesn't quote against usdc there. it quotes against nvda, tsla, aapl, hims. and since toro launched on the 18th, some of them claim a real asset backing too. so "what is this actually backed by" is finally a question you can answer onchain. here's how i check it. 1. discovery - pons is the launchpad doing the volume. ~$1 to deploy, 1b fixed supply, ~25k tokens minted in a single day in early sept - i ignore the live launch feed completely. 646k tokens from 167k addresses since july means that feed is pure noise - i only open names that already survived 3-7 days and still have a funded pool and people posting 2. read the quote asset before the chart - usdc pair = normal meme math, one leg - stock token pair = two legs, and one of them is an equity you didn't research - if nvda gaps down 6% overnight and your meme is flat in nvda terms, you are down 6% in dollars every stock token has a live chainlink feed. read that, not the usd number a dashboard prints for you. dividends and splits come through a uiMultiplier() call, so a "price drop" is sometimes just a split. 3. the mint window nobody checks robinhood's tokenization window for minting and burning stock tokens runs monday 02:00 cet to saturday 02:00 cet. saturday and sunday, inventory is frozen. no new stock tokens get created. - weekday: meme goes viral, more stock tokens get minted, the pool absorbs it - weekend: same viral move, fixed inventory, everyone fights over what is already sitting in the pool - dashboard shows one price, your average fill lands somewhere else i size way down on weekend entries into stock-paired pools. usually i just wait for monday. 4. if it claims a backing, open the chest toro takes a 3% tax on every trade and buys tokenized gold and equities with it. what i want to see before touching anything calling itself backed: - a chest address i can open myself in an explorer - backing per token against current mc. that ratio is your floor and it is almost always smaller than the marketing suggests - a written redemption path i haven't found toro's redemption mechanics spelled out anywhere public. that alone isn't a rug flag, but until someone can actually claim against the chest, the backing is a narrative number. skip rules: - no public chest address = skip - backing worth 2% of mc while the pitch says "asset backed" = skip - fat launchpad fees, anon team, no redemption doc = skip. noxa collected $12m in fees and went dark in july 5. risks, straight - stock tokens are tokenised debt securities issued by robinhood assets (jersey) limited. you hold an issuer obligation, not shares - they're restricted for us, canada, uk and swiss persons. toro bars us persons as well - thin pools distort. a meme going viral can move the price of its own quote asset - the asset-backed category is about a week old. treat every number as provisional the main takeaway: on robinhood chain the quote asset is half your position. work out what you are really priced against, and when its inventory is allowed to move, before you open a chart. want a part two where i price the premium over backing on a live chest token, step by step?

gm guys grokification went kinda wrong

alt open interest just passed bitcoin for the first time since dec 2024 that is not a bull signal by itself. it is a crowding signal. here's how i'm trading this rotation instead of chasing it the setup in numbers: - alt OI around $40b vs btc around $24b - altseason index at 49, it was 32 a week ago - alt mc outside the top 10 above $200b, up ~10% since sept 1 - total mc back over $3t, btc back near 85k last time alts flipped btc on OI was dec 2024. a lot of mid caps bled out in the weeks after while btc just sat there. worth keeping in the back of your head. 1. separate leverage moves from spot moves open coinglass or coinalyze, pull up the name you actually want to buy. - OI up + price up + funding hot = rented bid, someone is paying rent on that candle every 8h - price up while OI is flat or falling = spot is doing the work. that's the one i want - OI that's big relative to float is a liquidation map, not a trend zec is the loud version of this right now: about $2.4b OI against a ~$19b mc. works great until it doesn't, and it unwinds fast. 2. pick the lane before you pick the ticker i'm not trying to catch every name that moves. two lanes only: - majors with real spot flow (eth, sol), sized bigger, held longer - one or two mid caps under $1b where the story is still fresh and OI hasn't gone vertical yet skip rules i use: - funding above ~0.05% per 8h and still climbing, i don't open a fresh long - top 10 holders over 25-30% on a low cap, skip - already +200% in 90 days with OI at ath, skip. that's someone's exit - if my only reason is "it's altseason", skip 3. execution - scale in, two or three clips, never one market buy into a green candle - on perps i'm staying under 3x here. crowded longs don't need mine stacked on top - spot for anything i'd be annoyed to get wicked out of - invalidation level before entry. if i can't name the level i'm not in the trade 4. management - first trim at +40-50%, that pays for the position and the rest rides free - funding rips positive and OI makes a new high while price goes sideways, i'm trimming not adding - index above 75 is where i'm taking risk off, not adding it - btc dominance turning back up is my cue to cut the long tail first. mid caps bleed before majors do risks, honestly: - this can keep running for weeks and make every rule above look dumb - OI crossovers are a tiny sample, dec 2024 is one data point - funding and OI lag. they describe crowding, they don't call the top - i trim early more often than i call tops correctly, and i've eaten a lot of upside doing it the main takeaway: this crossover is a positioning read. i let it change my size and my stops, i don't let it change what i think a name is worth. want a part two on reading funding vs OI divergence on one chart, step by step?