Andrey Sergeenkov @Nikopolos

Disclosure becomes possible in two ways. In both California and Nevada, where she lived, the privilege passes to the personal representative of her estate, who can consent to disclosure. Documents that existed before she gave them to her lawyers, such as messages from Hack VC, do not become privileged by being in Sanford Heisler’s files and can be obtained by subpoena or court order.

Collected what can be verified so far about @hsinju's death. Maybe it saves someone some time. sergeenkov.com/hsin-ju-chuang…

The linked paper is an opinion ("Perspective") article. It explicitly states it is "not proposing that ADHD be reclassified exclusively as a circadian disorder" and that "evidence on remission of ADHD with circadian interventions is lacking." doi.org/10.3389/fpsyt.…

Nicholas Fabiano, MD @NTFabiano

7 days ago

We published a paper reconceptualizing ADHD as a circadian rhythm disorder.

368 4K 36K 3.8M 27K

Arbitrum (@arbitrum) pays its delegates $54,000 a month between them, down from $67,000 under the programme it replaced last year. Over the same change the number of proposals they vote on fell from 5.6 a month to 2.5, so the cost of getting one proposal voted on went from $11,964 to $21,600. The programme got smaller and each vote it buys got more expensive. For scale, the DAO pays one advisory firm $250,000 a month on a two year contract to work on Arbitrum alone. Everyone who votes, together, gets a fifth of that. No delegate can earn more than $700 on a single constitutional vote. And the most common reason one lost their reward was missing the five day deadline to post a comment explaining how they voted. All four figures come from the programme operator's own update, published in April. The two pairs sit four paragraphs apart in it and are never divided.

Copy trading the @HyperliquidX leaderboard is sold as a product by exchanges. In a backtest the moment a leader enters carries no information at all, and the fee alone decides the outcome. I simulated 9,125 copied entries over 150 days, equal weighted, each held one hour. The traders were picked four ways: those who build positions gradually and those who finish most days in profit, each as a fixed list and as a list rebuilt monthly. Before fees the four land between −2.2 and +2.1 basis points a trade. Every 95% interval brackets zero. Break even needs 9 bps, the round trip taker fee at the entry tier, and that is a floor because copy platforms take a cut of profit on top. After the fee the four lose between 6.9 and 11.2 bps a trade. The loss is arithmetic: a fixed tariff applied to an edge of zero. Market drift does not explain it. Over the window the mean coin returned −0.07 bps an hour, so a cohort that sits long every single time picks up about a hundredth of the fee.

A @Polymarket volume figure built from chain data is short by about a third. If the query was written before April, it returns nothing at all. Those figures are how the market gets sized, valued and compared with other venues such as Kalshi. In the last week of August the three contracts took about $558M of taker volume. Reading only the main one loses $171M of it. The main contract handles ordinary yes or no markets and carries 69%. A second handles markets with several outcomes, an election with a field of candidates, and carries 29%. The rest goes through a third contract for multi leg bets. Volume here sits in very few trades. The median fill is $3.60, and the largest 1% of fills carry about 70% of the money. So the multi outcome book is 18% of the trades and 29% of the money, and counting trades will never tell you whether a missing venue matters. Polymarket replaced all three contracts on 28 April. The old addresses still resolve and still hold three years of history, and they have taken no trades since. A query pointed at them returns an empty result with no error, so it reports a market that went quiet.

The only major precedent for a Treasury buyback of this scale is the 2000-2002 program, when the Treasury retired $67.5 billion of old 30-year bonds. Day one back then looked just like yesterday: a sharp spike in long-bond prices and chaos among traders, which then settled into steady repricing. But the effect held for the entire life of the program, so shorting the rally was a losing trade.

Who can apply? You should apply if your project has existing traction and is actively advancing privacy on Ethereum and across the open internet. Our two categories for this epoch are: - Privacy: everything from private transfers, shielded pools, wallet/RPC privacy, encrypted mempools, and zero-knowledge (ZK) tooling - Open internet: includes secure messaging, metadata-protecting communication tools, private compute hardware, and the privacy libraries that everything else is built on. Important dates: Deadline to apply: 10 September, 2026 Epoch Accelerator: 6 - 10 October, 2026 Epoch allocation window: 14 - 21 October, 2026 Apply now: octant.fillout.com/epoch-13

@Aly_Cabral The em dash is syntactically the cheapest option. Commas, semicolons, parentheses, and colons all force you to pick a specific relationship between parts of a sentence. The em dash fits almost anywhere and almost never reads as an error.

You've mixed up a config stub with a distilled model. Kimi-K3-0.18B has the same architecture config as K3 but fresh random weights and zero inherited knowledge, built with llm-compressor's create-tiny-model so people can test their inference stack on a laptop before touching the real 1.4TB weights.

@mustafa_ramax @Cointelegraph Moonshot's own serving measures at 32 tok/s output.

@mustafa_ramax @Cointelegraph That's one rack, roughly two senior engineer salaries, and it serves hundreds of users at once. Any regional cloud or mid-size company already burning $30K a month on frontier APIs covers that in about a year and then owns the thing outright, with the data staying in-house.

Keeping Ethereum's core devs paid costs $30M a year. That's 0.17% of the money parked in ETH ETFs on top of their code. The institutions earning fees on it contribute nothing. morrow.report/18-billion-in-…

Ethereum staking yield is quietly paying for Bitcoin's infrastructure. @OctantApp's latest round sent its biggest grant to the Tor Project, the network behind roughly 1 in 4 $BTC nodes. 148.7 ETH to Tor all-time, all of it from yield on staked $ETH. btcpeers.com/ethereum-staki…

It's not a bug, it's a feature. "As time goes to infinity, all keys will be lost."

Coin Bureau @coinbureau

3 months ago

🔥ZCASH CO-FOUNDER: BITCOIN SHOULD REMOVE 21M SUPPLY CAP AND GROW 4% PER YEAR Eli Ben-Sasson said Bitcoin’s 21M cap “doesn’t make sense because as time goes to infinity, all keys will be lost.” He suggested 4% annual issuance tied to population growth to ensure “there’s enough

keyboard_arrow_left Previous keyboard_arrow_right Next

883 49 581 570K 102

The AI pivot works because narrative capital is cheaper than building. Question is what happens to projects that skip narratives entirely, Tor and Solidity don't pivot, they just get funded through @OctantApp rounds while the pivoters burn through their third rebrand. Boring survives.

wstETH crossing into consumer apps is the yield layer eating fintech. Fun fact, the same staked-ETH yield already funds crypto's unglamorous infrastructure, @OctantApp runs entirely on staking returns from 100k ETH and just distributed 88 ETH to 25 projects. Composability of yield might matter more than composability of tokens.

Growth ceilings cut both ways, teams also leave when ecosystem funding dries up. Ethereum's answer has been recurring public goods money, rotki and L2BEAT keep shipping partly because rounds like @OctantApp's fund them epoch after epoch. Which chain keeps its builders is a funding question as much as a scaling one.

Everyone measures these buys in future price targets, nobody measures them in yield. 42k ETH staked throws off enough annually to run serious grant programs forever. The Golem Foundation does exactly that with 100k ETH and @OctantApp has turned the yield into $9M for public goods without selling a single coin.

Interesting that cashback is the retention hook here, $18M paid out to keep people spending. Meanwhile $GLM lockers at @OctantApp route their staking rewards the opposite direction, into funding Tor and open source devs. Crypto finally has both consumer loops working, extraction and contribution.