BTC $63,523, ETH $1,885.80, SOL $75.81, XRP $1.008, HYPE $56.18, DOGE $0.070466, AAVE $88.65, ZEC $492.31.
Thursday morning is about crypto rails getting institutional habits.
The August 10-12 digests already covered BIP-110, CLARITY calendar risk, ETF flows, miner AI pivots, tokenized equities, wallet and bridge failures, SEC Reg Crypto setup, Fidelity staking, Harmony’s mint, Cisco exploitation, CoreWeave, and the latest round of GitHub agent-interface picks.
This morning moves the board again.
Goldman is buying NEOS to get bitcoin and ether income ETFs instead of building the shelf slowly. Wintermute plans a $1 billion push into AI infrastructure and traditional-market HFT because crypto flow alone is no longer enough. Kalshi is adding a Wall Street-style low-latency data feed from DoubleZero. Miden is preparing a private USDC-backed stablecoin with selective disclosure. Solana came within about 20 million staked SOL of a finality halt after a connectivity failure. Anchorpoint started beta access for Hong Kong’s HKDAP stablecoin. Russia’s central bank wants retail crypto narrowed to BTC, ETH, and USDT. CPI cooled exactly as expected, which kept bitcoin stuck inside the same range. New York City opened a probe into prediction-market marketing. GitHub’s fresh board points to financial-market foundation models, AI-native PowerPoint production, and tiny models for edge devices.
Less “which token won the day?” More “who owns the wrapper, feed, privacy layer, validator path, regulated currency rail, and execution model when crypto markets start looking like finance?”
Price snapshot via CoinGecko simple-price data around 05:18 HKT.
1. Goldman Bought The Bitcoin Income Shelf
CoinDesk reported that Goldman Sachs agreed to acquire NEOS Investments in a cash-and-equity deal worth up to $2.25 billion.
NEOS manages BTCI, a bitcoin synthetic income ETF with about $1.1 billion in assets, plus other options-based income funds. The deal gives Goldman a roughly $30 billion options ETF platform and lifts its combined ETF assets above $130 billion.
The bitcoin detail is useful.
BTCI doesn’t simply hold bitcoin. It uses covered-call exposure on bitcoin ETPs to pay monthly income, with a yield around 27%, while giving up some upside when bitcoin rallies. The buyer is different from a spot ETF buyer.
Spot ETFs solved access. Income ETFs solve behavior.
Financial advisers have clients who want distributions, statements, guardrails, and recognizable portfolio roles. Goldman is buying the machinery for that demand because bitcoin has stopped being a single-product asset class.
The wrapper war is moving from “can you hold BTC?” to “can you turn BTC into the payout profile a portfolio already knows?“
2. Wintermute Wants Half Its Revenue Outside Crypto
CoinDesk reported that Wintermute plans to spend about $1 billion over five years on high-frequency trading and AI data-center infrastructure.
The target is blunt: non-crypto markets should generate more than half of revenue by the end of 2027, up from about 10% today.
That says plenty about the tape.
Wintermute’s average daily trading volume has fallen to about $10 billion this year from $15 billion in 2025. It is still one of crypto’s main market makers, but the firm’s next leg is stocks, commodities, foreign exchange, ETFs, prediction markets, and compute-heavy quantitative models.
Crypto taught firms how to move fast in fragmented, always-open markets. Now the better operators are taking that playbook back into traditional finance.
That is a compliment and a warning.
If a top crypto-native market maker can earn more by selling speed, models, and liquidity outside crypto, then token markets have to compete for attention like any other venue.
3. Kalshi Added Low-Latency Market Data
CoinDesk reported that Kalshi is adding DoubleZero’s low-latency market data feed to its order book.
DoubleZero describes the product as a transport layer that sends live exchange and onchain data over dedicated fiber, then distributes it to connected traders at the same time.
That sounds dry. It isn’t.
Prediction markets are becoming pricing venues for macro releases, policy odds, event risk, and eventually crypto-linked derivatives. If a contract can move on CPI, a rate decision, or a court ruling, the fastest clean data feed becomes part of the market structure.
The interesting bit is where this is happening. Kalshi is a regulated prediction market. DoubleZero is Solana-linked low-latency infrastructure. The customer is institutional trading flow.
That is the direction: event markets stop feeling like novelty betting screens and start asking for the same feed discipline as futures exchanges.
4. Miden Is Betting Privacy Stablecoins Can Be Compliant
CoinDesk reported that Miden plans to launch USDCx, a private stablecoin backed 1:1 by USDC held through Circle’s xReserve infrastructure.
USDCx is expected to debut with Miden’s mainnet near the end of the month. The pitch is private-by-default transactions with selective disclosure for auditors, regulators, and counterparties.
That is the right privacy problem.
Institutions don’t want every balance, payroll run, trading flow, counterparty, and treasury movement visible on a block explorer. Regulators don’t want black-box money movement either. Public-chain transparency made crypto auditable, but it also made serious business activity hard to run without leaking strategy.
Selective disclosure is the compromise worth watching.
If it works, stablecoins can move closer to corporate treasury, B2B payments, payroll, and trading settlement. If it fails, the market gets another privacy tool that compliance teams refuse to touch.
5. Solana Nearly Hit The Freeze Line
CoinDesk reported that a routing failure knocked almost 29% of Solana’s staked tokens offline on Wednesday morning.
Marinade said Solana came within about 20 million tokens of the one-third delinquency threshold where finality stops. Around 90 validators were affected, and the issue traced back to a bad internet route from Teraswitch’s Miami facility that spread to data centers in Europe and Asia.
This is the boring infrastructure risk that matters.
The chain didn’t freeze. That is good. But it got close enough that validator concentration, data-center dependencies, failover paths, and stake distribution became the real story.
Fast chains love throughput numbers. Users care about whether transactions finalize when a network provider has a bad morning.
Solana’s next credibility win won’t come from another speed chart. It will come from making this kind of correlated outage harder to repeat.
6. Hong Kong’s HKD Stablecoin Reached Distribution
CoinDesk reported that Standard Chartered-led Anchorpoint Financial started beta access for HKDAP, its Hong Kong dollar-backed stablecoin.
HashKey Exchange and OSL Group are authorized distributors, giving eligible institutions and professional investors a way to mint and redeem the token. Anchorpoint plans broader access channels and possible retail availability by late 2026.
That makes Hong Kong’s stablecoin push more real.
Licences are paperwork. Distribution is the test.
HKDAP starts with institutional payments and settlement, which is sensible. A local-currency stablecoin only matters if banks, exchanges, brokers, market makers, and corporates can use it without treating every mint and redemption as a bespoke legal project.
The question now is whether HKDAP becomes a real Hong Kong dollar rail or just a regulated demo with better branding.
7. Russia Picked A Three-Asset Retail List
CoinDesk reported that Russia’s central bank wants non-qualified retail investors limited to BTC, ETH, and USDT on regulated exchanges from September 1.
The draft rules also put a 300,000-ruble annual purchase limit, roughly $3,600, on non-qualified investors at each intermediary. Qualified investors would not face the same cap.
That is a whitelist model, not a broad market.
The inclusion of USDT is the tell. Bitcoin and ether are the obvious “major asset” choices. Tether is the practical dollar pipe. Retail crypto access, in this design, becomes three approved instruments inside regulated venues while domestic crypto payments remain prohibited.
This is what controlled adoption looks like.
Governments can decide they want exposure rails, tax visibility, and investor limits without accepting open-ended token choice. The market gets access, but only inside a narrow box.
8. CPI Removed The Selloff, Not The Range
The Block reported that bitcoin slipped below $64,000 after July CPI landed exactly in line with forecasts.
Headline CPI cooled to 3.4% from 3.5%. Core inflation eased to 2.5% from 2.6%. Both matched consensus, which gave the market no reason to reprice hard in either direction.
That is why bitcoin still feels stuck.
ETF access, tokenization deals, stablecoin launches, and policy movement are all real, but macro has not handed crypto a clean breakout signal. The current band near $62,000 to $66,000 is less about disbelief and more about waiting.
The next catalyst probably has to be specific.
Either the Fed changes the rate path, liquidity gets easier, or crypto produces a demand story strong enough to matter without macro help. An in-line CPI print only buys time.
9. Prediction Markets Got A Marketing Probe
The Block reported that the New York City Council opened a probe into Kalshi, Polymarket, Coinbase, and Gemini Titan over alleged false, deceptive, or abusive marketing tactics.
The council’s letters cite concerns that younger people are being targeted. They also reference allegations reported by The Wall Street Journal that Polymarket paid creators to film themselves placing fake bets and fake wins on site replicas.
This is where prediction markets leave the nerd zone.
The policy fight has been about whether event contracts are federally regulated derivatives or state-regulated gambling. The marketing fight is about something messier: how these products are sold to normal users once they become mainstream.
Prediction markets want to be information infrastructure. Fine. Then the ads, affiliate loops, influencer scripts, screenshots, and fake-win content have to survive the same scrutiny as the market design.
10. GitHub Trending - Markets, Decks, And Edge Models
The featured-repo tracker ruled out recent repeats including msitarzewski/agency-agents, paperclipai/paperclip, 3b1b/manim, NanmiCoder/MediaCrawler, ZhuLinsen/daily_stock_analysis, openai/openai-cookbook, and the last few agent-infrastructure picks.
Fresh picks from GitHub Trending, GitHub Trending Python, and repo metadata:
shiyu-coder/Kronos has about 36.9k stars and describes itself as a foundation model for the language of financial markets. The useful signal is that market data is becoming a pretraining domain, not just a spreadsheet input for indicators.
hugohe3/ppt-master has about 45.5k stars and turns documents or topics into native PowerPoint decks with charts, tables, animations, and speaker-note narration. It fits the enterprise-agent pattern: the output has to land in the file format teams already use.
cactus-compute/needle has about 4.1k stars and packages a 14MB foundation model for phones, wearables, smart-home devices, and robots. That is the opposite of the giant data-center story. It points to agent features that run at the edge because latency, privacy, battery, or connectivity force them there.
Today’s GitHub board was not another generic agent framework. It was finance-native modeling, office-native artifact generation, and tiny-device inference.
Morning Read
Read Goldman’s NEOS acquisition, then read Miden’s USDCx launch plan, then read the Solana near-freeze report.
The number to remember is 29%.
That is the share of Solana staked tokens Marinade says briefly went offline during the routing failure. The second number is $2.25 billion, because Goldman is buying a ready-made options-income ETF platform with crypto wrappers inside it.
Thursday’s read is that crypto is being folded into real market plumbing. The winners won’t only have assets. They’ll have distribution wrappers, compliant privacy, fast feeds, resilient validators, local-currency settlement, and outputs that fit the tools institutions already use.
Evening Update - 18:24 HKT
BTC $63,576, ETH $1,883.02, SOL $75.80, XRP $1.007, HYPE $57.58, DOGE $0.070140, AAVE $88.82, ZEC $494.44.
Thursday evening is about crypto finance getting judged like finance.
The morning digest already covered Goldman buying NEOS, Wintermute’s AI and traditional-market expansion, Kalshi’s DoubleZero feed, Miden’s USDCx privacy stablecoin, Solana’s routing scare, HKDAP distribution, Russia’s three-asset retail whitelist, CPI’s range-bound bitcoin tape, New York’s prediction-market marketing probe, and GitHub’s market-model, deck, and edge-model picks.
Tonight moves to balance sheets, banking access, incentive design, and security access.
Brazil’s largest bitcoin treasury firm is preparing an ETF that mostly owns Strategy preferred shares. Metaplanet created yen-denominated BitBonds instead of selling bitcoin. BitGo grew revenue almost 80% and still posted a loss. Mirae Asset is putting another $35 million into Korbit after buying the Korean exchange. The Blockchain Association backed Custodia’s Supreme Court petition over Fed master-account access. Zerohash’s first national trust-bank bid came back from the OCC with deficiencies. Securitize missed earnings even as tokenized AUM and transaction volume hit records. The CFTC warned prediction markets that trading incentives can create wash-trading and manipulation risk. Bitcoin firms asked AI labs for the same model access attackers can already get. DOGE leverage rebuilt to 2025 levels while the token is down about 70% year over year. GitHub’s evening board points to enterprise agent workspaces, team memory hubs, and provenance stripping.
That is a different kind of maturity test.
Crypto is no longer only trying to prove that tokens can trade. It has to prove that credit products, public companies, regulated exchanges, tokenized funds, event markets, security teams, and AI-era tooling can hold up when investors ask boring questions.
Evening price snapshot via CoinGecko simple-price data around 18:24 HKT.
11. Brazil Is Packaging Bitcoin Treasury Credit
CoinDesk reported that OranjeBTC plans to list DIGY11, an ETF on Brazil’s B3 that would initially allocate 95% to Strategy’s STRC preferred shares and 5% to Strive’s SATA.
The target is not spot bitcoin exposure.
DIGY11 is designed as a monthly income product in reais. OranjeBTC expects distributions near Brazil’s CDI interbank rate plus 3 to 5 percentage points, net of costs, though returns are not guaranteed. The fund also plans monthly FX hedges and a 0.90% management fee.
That is the wrapper story getting stranger.
Brazilian investors would be buying a local-currency ETF holding preferred shares issued by bitcoin-treasury companies. The bitcoin sits on corporate balance sheets. The investor gets credit, distributions, issuer risk, currency hedging, and a management fee.
This is what happens after the spot ETF unlocks access. The next product turns bitcoin balance sheets into income plumbing.
12. Metaplanet Added BitBonds To The Treasury Toolkit
CoinDesk reported that Metaplanet completed its first BitBonds sale, issuing four privately placed unsecured series worth about 200 million yen, or $1.3 million.
The bonds mature in roughly three years and carry annual interest of 4% to 4.3%. They sit beside common stock, equity-linked securities, and preferred shares as another funding channel.
The sale is small. The structure matters.
Metaplanet is trying to build a yen credit market around a bitcoin-heavy corporate balance sheet without selling the underlying BTC. CEO Simon Gerovich also denied rumors that the company sold roughly $320 million of bitcoin, saying a 5,014 BTC movement was a custody transfer and holdings remain 43,000 BTC.
Bitcoin treasury companies are becoming capital-structure experiments.
The easy version was “issue equity, buy bitcoin.” The harder version is matching debt cost, maturity, liquidity, custody movement, shareholder dilution, and BTC volatility without losing trust.
13. BitGo Grew Fast And Still Lost Money
The Block reported that BitGo posted $4.33 billion in second-quarter revenue, up 79.6% year over year, while recording a $19 million net loss.
The revenue growth came from higher digital asset sales and stablecoin-as-a-service growth. Normalized assets on the platform rose 31% to $65.2 billion, and client count rose 26%. BitGo also held $159 million of cash, $147.7 million of bitcoin, and authorized a $50 million share repurchase.
That is a useful split-screen.
Custody and stablecoin infrastructure can grow while accounting marks and operating costs still drag the income statement. BitGo’s loss narrowed from the first quarter, but the year-over-year swing from profit came partly from digital-asset valuation.
Infrastructure companies want software multiples. Public markets still see crypto exposure on the balance sheet.
The next test is whether stablecoin and custody revenue can become durable enough that token marks stop dominating the story.
14. Mirae Asset Put More Capital Behind Korbit
The Block reported that Mirae Asset is injecting roughly 50 billion won, or $35.3 million, into DigitalX, the operator of South Korean exchange Korbit.
Mirae completed its Korbit acquisition in July after buying 97.15% of the company for about 130 billion won, or $91.5 million. DigitalX said the new capital will improve its financial structure and address urgent management needs.
That points to the Korean exchange race changing owners.
Korbit remains one of South Korea’s five fully licensed fiat-to-crypto exchanges, but Upbit and Bithumb still dominate. Mirae’s bet is that an exchange can become part of a broader platform for digital and traditional assets rather than a standalone trading venue.
Korea’s banks and brokers are not waiting for crypto exchanges to stay independent.
Hana has exposure to Dunamu. Samsung affiliates have exposure to Dunamu. Korea Investment & Securities has a Coinone stake through OKX Ventures. Mirae now has Korbit.
The exchange layer is becoming financial-group infrastructure.
15. Custodia’s Banking Fight Reached The Supreme Court Lobby
The Block reported that the Blockchain Association filed an amicus brief supporting Custodia Bank’s Supreme Court petition over Fed master-account access.
Custodia first applied in October 2020. The Kansas City Fed denied the application in January 2023, and lower courts have sided against Custodia so far. The Kansas City Fed is due to respond to the petition by Sept. 11.
This is not just a Custodia story.
A Fed master account gives a bank direct access to central-bank payment rails. Without one, digital-asset banks often need correspondent-bank relationships, which means more dependency on institutions that may not want the risk.
The Blockchain Association’s argument is that eligible state-chartered banks should not be shut out by discretionary federal pressure.
Kraken Financial got a limited-purpose master account in March. Custodia still hasn’t.
That split matters because the industry needs to know whether access is rule-based, relationship-based, risk-model-based, or simply discretionary.
16. Zerohash’s Trust-Bank Bid Hit The OCC Wall
CoinDesk reported that the OCC returned Zerohash’s national trust-bank application after finding material deficiencies.
Zerohash handles crypto infrastructure for firms including Morgan Stanley’s E*Trade. The company said the return does not affect current operations and that it plans to refile with a narrower scope by the end of the month.
That is the other side of the access story.
The OCC has granted many provisional digital-asset trust-bank charters, so this is not a blanket anti-crypto posture. A returned application says the regulator was not comfortable with the package in front of it.
Zerohash’s next version will likely be more focused on national trust activities tied to its rollout timeline.
That is how crypto banking gets normalized: not through one grand approval wave, but through narrower charters, staged permissions, and a lot of paperwork that forces infrastructure firms to describe exactly what business they are in.
17. Securitize Proved Tokenization Still Needs Revenue
CoinDesk reported that Securitize fell 20% after hours after reporting a second-quarter earnings miss in its first report since going public.
The tokenization firm reported revenue of $14.4 million, down 5% from a year earlier and below the $20.6 million analyst estimate. It posted a $2.37 per-share loss versus an expected $0.15 loss.
The activity data looked much better.
Average tokenized assets under management reached a record $4.3 billion. Transaction volume rose 147% to $5.3 billion. Its fund-services arm oversaw 663 active funds and $24.3 billion in assets under administration.
That is the tokenization gap in one earnings report.
Wall Street likes the story. BlackRock’s BUIDL gives Securitize a flagship proof point. But public investors still want revenue conversion, margins, and earnings discipline.
Tokenized assets can hit records while the stock sells off. The rail has to monetize.
18. The CFTC Warned Prediction Markets On Incentives
CoinDesk reported that the CFTC warned prediction-market platforms about deficient filings for trading-incentive programs.
The agency said some filings do not give enough notice or compliance analysis. It also warned that volume rewards can push traders toward wash trading, pre-arranged trading, or other manipulative behavior. Some market-maker stipends and rebates can create similar problems.
This is a different issue from New York’s marketing probe in the morning digest.
The morning story was about how prediction markets are sold to users. The evening story is about how liquidity is manufactured inside regulated event-contract venues.
Both matter.
Prediction markets want deep books and tight spreads. Incentives can help create that. But if the wrong structure pays users to hit volume targets or guarantees market-maker economics too loosely, the venue starts importing the same incentive abuse that exchanges have spent years policing.
Event markets are becoming real market structure. That means real surveillance and real incentive design.
19. Bitcoin Security Teams Want Better AI Access
CoinDesk reported that more than three dozen crypto firms asked major AI labs to give open-source security researchers early access to more capable models.
The letter was organized by the Bitcoin Policy Institute and signed by Coinbase, Block, BitGo, Blockstream, Anchorage Digital, ARK Invest, Bitwise, Foundry, Casa, Exodus, Brink, Chaincode, Btrust, and others.
Their argument is simple: attackers can use stolen access, open-weight models, and hacking tools, while defenders hit safety filters when they try to audit real software.
Recent BTCPay Server and Lightning-node bugs found partly by an AI-powered volunteer Bitcoin Red Team made the point sharper.
This is one of the better AI-security arguments.
Guardrails that stop malware generation are useful. Guardrails that block legitimate maintainers from finding bugs in critical open-source financial infrastructure create an asymmetry. The right answer probably needs trusted researcher programs, secure sandboxes, logging, and direct contacts with AI lab security teams.
Bitcoin does not need AI hype. It needs defenders with tools good enough to match the attack surface.
20. DOGE Leverage Rebuilt While Price Stayed Broken
CoinDesk reported that DOGE futures open interest has climbed to about $1.21 billion from roughly $930 million in late June.
Measured in coins, open interest is back near 17.18 billion DOGE, close to the 17.78 billion level from October 2025. The difference is price. DOGE trades near 7 cents now versus about 25 cents then and is down almost 70% year over year.
The positioning is crowded on the bullish side.
CoinDesk said more than three Binance accounts were long for every short, and OKX’s ratio was above five to one. That does not mean the dollar value of longs is five times shorts, but it does show which side retail accounts prefer.
This is why meme leverage is worth watching even when the asset is not leading the market.
If price keeps slipping, forced liquidations can turn a sleepy token into a market-structure event. The interesting part is not DOGE’s narrative. It is the mismatch between weak spot price and rebuilt futures risk.
21. GitHub Trending - Agent Workspaces, Shared Memory, And Provenance Cleanup
The featured-repo tracker ruled out recent repeats including shiyu-coder/Kronos, hugohe3/ppt-master, cactus-compute/needle, harveyai/harvey-labs, 3b1b/manim, denoland/deno, ShawnPana/phone-harness, oil-oil/oil-motion, and Flaminis/Dalaran.
Fresh picks from OSSInsight Trending, GitHub Trending, and repo metadata:
cloudflare/cloudflare-os was listed with about 49 daily stars and describes itself as an agent workspace on Cloudflare Workers for documents, apps, and company-context agents. The useful signal is that agent runtime is moving closer to deployable worker infrastructure, not only local chat sessions.
TencentCloud/TencentDB-Agent-Memory was listed with about 39 daily stars and packages team-level agent memory across chat memory, skills, LLM wiki, and code graphs. That matters because memory is becoming an enterprise governance problem, not just a retrieval trick.
guillaumemeyer/watermarks-remover was listed with about 23 daily stars and focuses on removing multi-vendor provenance marks and metadata from common formats. The uncomfortable signal is demand for provenance cleanup at exactly the moment AI-content detection, watermarking, and C2PA metadata are becoming policy and platform issues.
Today’s evening repo board is sharper than another agent framework list. It points to where agents actually meet production: runtime, shared memory, and the evidence trail around generated artifacts.
Evening Read
Read Securitize’s earnings miss, then read the CFTC prediction-market incentive warning, then read the bitcoin firms’ AI-security letter story.
The number to remember is $4.3 billion.
That is both BitGo’s quarterly revenue and Securitize’s average tokenized AUM. One shows custody and stablecoin infrastructure growing through accounting pain. The other shows tokenization activity growing before revenue catches up.
Thursday evening’s read is that the wrappers are multiplying, but the market is getting less forgiving. A bitcoin-income ETF still has credit risk. BitBonds still depend on issuer quality. Trust-bank access still depends on regulators. Tokenized AUM still has to become revenue. Prediction markets still need clean incentive design. AI defense still needs access to serious tools.
Crypto finance is getting real enough to be boring. That is progress. It is also when the questions get harder.