BREAKING: Brazil ETF $⁠EWZ crashes -12% as Lula takes the lead in the presidential runoff.

paradigm走了，MC也走了，现在A16Z也走了。 对BTC的信仰感觉圈子里的人真没有17、20年那会强了。那是愚蠢但火热的年代，那时候的人们： 相信三浪的EOS 相信反杀BTC的BCH 相信存储天王FIL 相信公链天王ICP 奇怪的是，现在ETF通过了，美国支持了，币圈信仰的空气却越来越稀薄……

I dumped it all. The Maelstrom directional book, took profits on $ETH, $ENA, $UNI and other shitcoins. I think we gotta go down to go up. It looks a bit dicey out there in macro and geopolitical land. Going to sit on the sidelines and wait for the all clear signal … Brrrrr

Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications. On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments. While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software. After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely. openai.com/index/securing-ou…

我也疑惑这个问题，为啥 Grokbot 能做起来，而且惊为天人，但是 Grok 一直不温不火； 不过所以感觉 Grokbot 和 Grok 团队应该不是一个吧！

OpenAI nears up to $30 billion funding round at $1.25 trillion valuation led by UAE investors cnbc.com/2026/10/22/openai-f…

CHINA’S CSI 300 FALLS 1.5%, ENTERS BEAR MARKET

We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent. It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.

*DEEPSEEK RAISES $15 BLN IN FUNDING DEAL JOINED BY TENCENT, CATL

🚨NEW: OCC Finalizes GENIUS Act Rules, Backs Off Stablecoin Rewards Restrictions The @USOCC has finalized its GENIUS Act rules, dropping proposed restrictions on stablecoin rewards and leaving a major fight between banks and crypto firms unresolved. cryptoinamerica.com/p/occ-fi…

JUST IN: Anthropic recorded a staggering $80,000,000,000.00 net loss in the first half of 2026.

JUST IN: Anthropic recorded a staggering $80,000,000,000.00 net loss in the first half of 2026.

Two important changes to timeline recommendations: - New AI/SI ranking model that predicts probable time spent much better - Retrieval model trained to surface posts people read all the way through User active seconds is growing exponentially and hitting all-time highs almost every day:

Business Insider confirms it: Amazon has warehouses of GPUs that can’t be installed, and is buying them so far in advance that a supply chain team said they “didn’t have a clear forecasted path” to installing them. businessinsider.com/amazon-2…

A note on transferability. Papertrade is a complicated primitive, and managing it from a risk standpoint is a complicated problem. The PAPER token is far more integrated into the economics of the primitive than most tokens are in their respective protocols today, and, broadly, was designed to act as a counter insolvency incentive coupled with a rebate mechanism to losing traders. Market makers and exchanges are by far the two most profitable businesses in crypto, and our goal was to try and externalise as much of their margins back to users as possible. We are now just over a week into the life of papertrade and we are becoming increasingly confident that this is both possible, but also a likely outcome as agentic led software becomes the norm. As it stands, the paper token traded as a perp becomes a leveraged bet on bankroll solvency - and being able to short bankroll solvency materially changes the payoff curve of a manipulation attack. The target number goes from the bankroll, to the bankroll + short side exposure on paper. We can, and do limit bankroll size specifically to control this payoff curve and ensure it never becomes existentially problematic, but external perp markets are not something we would have any control over if left to float of their own accord. For example, if our bankroll is $10m, an attack costing $15m manipulating the price of bitcoin or ether is not viable. You simply lose money. If however you were also able to put on $10m of PAPER short side exposure, suddenly an attack costing $15m is very viable indeed. If we were to make the token transferable at this point, we are confident we would quickly enter a state too dangerous to allow any material exposure to be taken on papertrade, an end result that benefits no party beyond the wolves who have done the math on the short and are waiting in the wings with sharpened knives. We would also like to thank the various perps teams who have reached out about pre-market listings for their understanding and patience regarding this issue, and especially to the paragon team who, despite jumping the gun on us, have behaved with the utmost integrity and transparency once a dialogue was opened. Perps on PAPER will always act as a self-limiting exposure ceiling for a token with this type of utility, and we think it would be a shame to limit the growth of papertrade to the extent that allowing perps would necessitate. As such, at this juncture we do not have plans to allow off-contract token transferability. Structurally, papertrade is limited by the quality of the data it receives from hyperliquid, but we strongly believe that hyperliquid has a very large part to play in the future of global exchanges, and expect their books to continue to thicken and strengthen over time. A note on fees. Papertrade launched with a deadband fee from entry of 0.2bps. This means you don't earn anything until the price has moved by 0.002% in your favour. Over the months before launch, we played around with many different curves and fee structures affecting exclusively positions closed very close to entry price in order to try and devise a structure that removed the edge against the bankroll low time frame scalp bots have, whilst also minimally impacting the traders around which and for whom the product was devised. We were never able to find anything we were absolutely happy with and so decided to stay with the 0.2bp deadband fee through launch. This has however meant we have had to be very restrictive when it came to certain parameters, such as OI caps. Now, however, we have enough live data to fully understand exactly which strategies work, and to make the necessary changes to stop them. The changes we are making will have an on hours and off hours fee mode to account for the relative differences in liquidity the hyperliquid books see between these two time periods, and the ease with which low time frame moves are signalled. The changes we are making will only affect positions closed within 6bps of entry (0.06%) in on hours, and 10bps (0.1%) in off hours. Any trades closed above these threshold price shifts will see no change to their fees at all. On a look back to launch basis, these changes affect the profitability of bot strategies roughly 15x more than they do normal traders, for whom a very small proportion of overall winnings come from within this band. These changes should serve to flatten much of the edge our microfee structure necessarily creates on low time frames, and will help with long term exchange sustainability. This should hopefully allow us to open up parameters such as OI caps a little more so those of you who wish to can put on the size you want to, when you want to. It is possible these changes are still too lax, and that further revisions to this structure may be necessary going forwards. If so, we will notify users ahead of time. A note on paper tokenomics. In conjunction with the fee changes, we will also be implementing a change to the token issuance curve. Going forward, every $1 lost mints a fixed share of PAPER's supply at that moment, roughly 0.1% of supply per $1M lost (the exact number depends on the mint rate at the time the contracts are upgraded). PAPER minting in and near insolvency will also change, switching from a flat 100 per $1 of loss, to a time weighted multiplier on the curve growing over time, effectively acting as the house running a discounted auction on its own future cash flows. This change also removes the risk of the paper-mint-straddle-attack, temporarily pushing the protocol insolvent by closing only the winning side of a large straddle position and then attempting to mint as much paper as possible. In the immediate term these changes will only have a very subtle effect on PAPER minting, but in the long run will structurally change the outlook of the system beyond the first few weeks, serving to slowly dilute out older holders and leaving more recent users with an advantage. Reconstituting the PAPER system into something much closer to the exchange rakeback it was intended to be. We have been very reticent to touch this particular aspect of the system for obvious reasons, but ultimately we believe that these changes provide all users with the optionality to take a smaller piece of a potentially much larger pie than the current dynamic might necessarily allow. Other, miscellaneous notes. These contract upgrades will include a few other contract changes, including the addition of builder codes so that builders can utilise papertrade in whatever capacity they choose to. This will be a whitelist only to begin with and we will have more for you later in the week on where/how this will be rolled out. A small change to contract permissions on the reward distributor and a change on how below bust closes are processed by the exchange are also included, largely to block gas griefing attacks. These changes will be pushed to docs in full over the coming day, and we will further announce once the changes are queued into the timelock pending final security passes, scheduled to go live 24 hours after that point. Final note. We'd like to thank everyone that has tried out papertrade over the last 10 days. It has been a labour of love building this product and both of us have been taken aback at all of the positive feedback we have received. We'd also like to thank all those of you who have spent considerable time musing on the system we have built, and much of your advice has been taken onboard and will likely affect the colour of what papertrade becomes going forward. We know some of these changes will be unpopular with some, but we would not be making them if we did not feel them to be absolutely necessary. - blurr, jez

🚨 JUST IN: OpenAI is reportedly rushing to close a roughly $30 billion funding round that could value it above $1.4 trillion before Anthropic’s expected mid-November IPO, per Reuters.

A note on transferability. Papertrade is a complicated primitive, and managing it from a risk standpoint is a complicated problem. The PAPER token is far more integrated into the economics of the primitive than most tokens are in their respective protocols today, and, broadly, was designed to act as a counter insolvency incentive coupled with a rebate mechanism to losing traders. Market makers and exchanges are by far the two most profitable businesses in crypto, and our goal was to try and externalise as much of their margins back to users as possible. We are now just over a week into the life of papertrade and we are becoming increasingly confident that this is both possible, but also a likely outcome as agentic led software becomes the norm. As it stands, the paper token traded as a perp becomes a leveraged bet on bankroll solvency - and being able to short bankroll solvency materially changes the payoff curve of a manipulation attack. The target number goes from the bankroll, to the bankroll + short side exposure on paper. We can, and do limit bankroll size specifically to control this payoff curve and ensure it never becomes existentially problematic, but external perp markets are not something we would have any control over if left to float of their own accord. For example, if our bankroll is $10m, an attack costing $15m manipulating the price of bitcoin or ether is not viable. You simply lose money. If however you were also able to put on $10m of PAPER short side exposure, suddenly an attack costing $15m is very viable indeed. If we were to make the token transferable at this point, we are confident we would quickly enter a state too dangerous to allow any material exposure to be taken on papertrade, an end result that benefits no party beyond the wolves who have done the math on the short and are waiting in the wings with sharpened knives. We would also like to thank the various perps teams who have reached out about pre-market listings for their understanding and patience regarding this issue, and especially to the paragon team who, despite jumping the gun on us, have behaved with the utmost integrity and transparency once a dialogue was opened. Perps on PAPER will always act as a self-limiting exposure ceiling for a token with this type of utility, and we think it would be a shame to limit the growth of papertrade to the extent that allowing perps would necessitate. As such, at this juncture we do not have plans to allow off-contract token transferability. Structurally, papertrade is limited by the quality of the data it receives from hyperliquid, but we strongly believe that hyperliquid has a very large part to play in the future of global exchanges, and expect their books to continue to thicken and strengthen over time. A note on fees. Papertrade launched with a deadband fee from entry of 0.2bps. This means you don't earn anything until the price has moved by 0.002% in your favour. Over the months before launch, we played around with many different curves and fee structures affecting exclusively positions closed very close to entry price in order to try and devise a structure that removed the edge against the bankroll low time frame scalp bots have, whilst also minimally impacting the traders around which and for whom the product was devised. We were never able to find anything we were absolutely happy with and so decided to stay with the 0.2bp deadband fee through launch. This has however meant we have had to be very restrictive when it came to certain parameters, such as OI caps. Now, however, we have enough live data to fully understand exactly which strategies work, and to make the necessary changes to stop them. The changes we are making will have an on hours and off hours fee mode to account for the relative differences in liquidity the hyperliquid books see between these two time periods, and the ease with which low time frame moves are signalled. The changes we are making will only affect positions closed within 6bps of entry (0.06%) in on hours, and 10bps (0.1%) in off hours. Any trades closed above these threshold price shifts will see no change to their fees at all. On a look back to launch basis, these changes affect the profitability of bot strategies roughly 15x more than they do normal traders, for whom a very small proportion of overall winnings come from within this band. These changes should serve to flatten much of the edge our microfee structure necessarily creates on low time frames, and will help with long term exchange sustainability. This should hopefully allow us to open up parameters such as OI caps a little more so those of you who wish to can put on the size you want to, when you want to. It is possible these changes are still too lax, and that further revisions to this structure may be necessary going forwards. If so, we will notify users ahead of time. A note on paper tokenomics. In conjunction with the fee changes, we will also be implementing a change to the token issuance curve. Going forward, every $1 lost mints a fixed share of PAPER's supply at that moment, roughly 0.1% of supply per $1M lost (the exact number depends on the mint rate at the time the contracts are upgraded). PAPER minting in and near insolvency will also change, switching from a flat 100 per $1 of loss, to a time weighted multiplier on the curve growing over time, effectively acting as the house running a discounted auction on its own future cash flows. This change also removes the risk of the paper-mint-straddle-attack, temporarily pushing the protocol insolvent by closing only the winning side of a large straddle position and then attempting to mint as much paper as possible. In the immediate term these changes will only have a very subtle effect on PAPER minting, but in the long run will structurally change the outlook of the system beyond the first few weeks, serving to slowly dilute out older holders and leaving more recent users with an advantage. Reconstituting the PAPER system into something much closer to the exchange rakeback it was intended to be. We have been very reticent to touch this particular aspect of the system for obvious reasons, but ultimately we believe that these changes provide all users with the optionality to take a smaller piece of a potentially much larger pie than the current dynamic might necessarily allow. Other, miscellaneous notes. These contract upgrades will include a few other contract changes, including the addition of builder codes so that builders can utilise papertrade in whatever capacity they choose to. This will be a whitelist only to begin with and we will have more for you later in the week on where/how this will be rolled out. A small change to contract permissions on the reward distributor and a change on how below bust closes are processed by the exchange are also included, largely to block gas griefing attacks. These changes will be pushed to docs in full over the coming day, and we will further announce once the changes are queued into the timelock pending final security passes, scheduled to go live 24 hours after that point. Final note. We'd like to thank everyone that has tried out papertrade over the last 10 days. It has been a labour of love building this product and both of us have been taken aback at all of the positive feedback we have received. We'd also like to thank all those of you who have spent considerable time musing on the system we have built, and much of your advice has been taken onboard and will likely affect the colour of what papertrade becomes going forward. We know some of these changes will be unpopular with some, but we would not be making them if we did not feel them to be absolutely necessary. - blurr, jez

Next year, I will give one of my SpaceX shares to every kid in America. This will be the first of many.