7*24

Hot topic

Market

Activities

According to 1M AI News, the AI lab PrismML, co-founded by Caltech mathematician Babak Hassibi, has emerged from stealth mode and released the open-source 1-bit Bonsai series of large language models. The flagship model, 1-bit Bonsai 8B, features 8.2 billion parameters and occupies only 1.15 GB of memory, which is approximately 14 times more compressed than comparable 16-bit models (around 16 GB). The weights are available for download under the Apache 2.0 license on HuggingFace, along with two smaller models: 4B (0.5 GB) and 1.7B (0.24 GB). Bonsai 8B is a true end-to-end 1-bit model: the embedding layer, attention layer, MLP layer, and output head all represent weights using only +1 or -1, without any high-precision patches. PrismML claims its inference and language understanding capabilities on standard benchmarks are comparable to those of 16-bit full-precision models. The core compression mathematics was developed by the team over several years at Caltech, with the intellectual property owned by Caltech, making PrismML the sole exclusive licensee. The model was trained using Google v4 TPU. Measured speeds include 136 tokens/s on an M4 Pro Mac, 440 tokens/s on an RTX 4090, and approximately 44 tokens/s on an iPhone 17 Pro Max, while standard 16-bit 8B models cannot be loaded onto any iPhone. Energy consumption is reduced by about 4-5 times compared to 16-bit models. PrismML notes that existing hardware is not designed for 1-bit inference, and the speed and energy advantages primarily come from reduced memory usage; if hardware specifically designed for 1-bit operations (requiring only addition and subtraction, without multiplication) emerges in the future, efficiency could improve by an order of magnitude. PrismML has completed $16.25 million in SAFE and seed round financing, with investors including Khosla Ventures, Cerberus Capital, and Caltech. Vinod Khosla, founder of Khosla Ventures, stated that this is 'not a minor iteration, but a significant technological breakthrough, a mathematical breakthrough, not just another small model.'

[Caltech Releases Open Source True 1-Bit Model Bonsai: 8B Parameters at Only 1.15GB, Achieving 44 Tokens/s on iPhone]According to 1M AI News, the AI lab PrismML, co-founded by Caltech mathematician Babak Hassibi, has emerged from stealth mode and released the open-source 1-bit Bonsai series of large language models. The flagship model, 1-bit Bonsai 8B, features 8.2 billion parameters and occupies only 1.15 GB of memory, which is approximately 14 times more compressed than comparable 16-bit models (around 16 GB). The weights are available for download under the Apache 2.0 license on HuggingFace, along with two smaller models: 4B (0.5 GB) and 1.7B (0.24 GB). Bonsai 8B is a true end-to-end 1-bit model: the embedding layer, attention layer, MLP layer, and output head all represent weights using only +1 or -1, without any high-precision patches. PrismML claims its inference and language understanding capabilities on standard benchmarks are comparable to those of 16-bit full-precision models. The core compression mathematics was developed by the team over several years at Caltech, with the intellectual property owned by Caltech, making PrismML the sole exclusive licensee. The model was trained using Google v4 TPU. Measured speeds include 136 tokens/s on an M4 Pro Mac, 440 tokens/s on an RTX 4090, and approximately 44 tokens/s on an iPhone 17 Pro Max, while standard 16-bit 8B models cannot be loaded onto any iPhone. Energy consumption is reduced by about 4-5 times compared to 16-bit models. PrismML notes that existing hardware is not designed for 1-bit inference, and the speed and energy advantages primarily come from reduced memory usage; if hardware specifically designed for 1-bit operations (requiring only addition and subtraction, without multiplication) emerges in the future, efficiency could improve by an order of magnitude. PrismML has completed $16.25 million in SAFE and seed round financing, with investors including Khosla Ventures, Cerberus Capital, and Caltech. Vinod Khosla, founder of Khosla Ventures, stated that this is 'not a minor iteration, but a significant technological breakthrough, a mathematical breakthrough, not just another small model.'

Caltech Releases Open Source True 1-Bit Model Bonsai: 8B Parameters at Only 1.15GB, Achieving 44 Tokens/s on iPhone

According to 1M AI News, Google is developing a screenless fitness band under the Fitbit brand, with plans to launch it this year. Insiders reveal that the device will offer basic features, with additional functionalities available through a paid subscription, similar to the subscription models of screenless fitness band Whoop and smart ring Oura. However, Google will charge separately for the hardware, while Whoop relies entirely on subscriptions. NBA Golden State Warriors star Stephen Curry teased the device on social media on Tuesday, stating it will open a 'new relationship with health.' Google confirmed that Curry has been 'working with the team to create something special,' with more details to be announced 'soon.' The device features a gray fabric band with an orange lining, closely resembling the current screenless Whoop band. It will connect to the redesigned Fitbit app's AI health coach, a feature that launched in public preview last October, covering scenarios such as menstrual health, mental health, nutrition, and hydration tracking. With the addition of the Pixel Watch and traditional Fitbit product line, Google's wearable lineup will further expand. Google's entry comes as the screenless wearable market heats up. Whoop recently completed a $575 million Series G funding round at a valuation of $10.1 billion, with sports stars like Cristiano Ronaldo and LeBron James investing personally, and the company plans to go public within two years; smart ring Oura was also valued at $11 billion last year.

[Google's Screenless Fitbit Band Revealed: Similar to Whoop, Integrates AI Health Coach]According to 1M AI News, Google is developing a screenless fitness band under the Fitbit brand, with plans to launch it this year. Insiders reveal that the device will offer basic features, with additional functionalities available through a paid subscription, similar to the subscription models of screenless fitness band Whoop and smart ring Oura. However, Google will charge separately for the hardware, while Whoop relies entirely on subscriptions. NBA Golden State Warriors star Stephen Curry teased the device on social media on Tuesday, stating it will open a 'new relationship with health.' Google confirmed that Curry has been 'working with the team to create something special,' with more details to be announced 'soon.' The device features a gray fabric band with an orange lining, closely resembling the current screenless Whoop band. It will connect to the redesigned Fitbit app's AI health coach, a feature that launched in public preview last October, covering scenarios such as menstrual health, mental health, nutrition, and hydration tracking. With the addition of the Pixel Watch and traditional Fitbit product line, Google's wearable lineup will further expand. Google's entry comes as the screenless wearable market heats up. Whoop recently completed a $575 million Series G funding round at a valuation of $10.1 billion, with sports stars like Cristiano Ronaldo and LeBron James investing personally, and the company plans to go public within two years; smart ring Oura was also valued at $11 billion last year.

Google's Screenless Fitbit Band Revealed: Similar to Whoop, Integrates AI Health Coach

On April 1, according to monitoring by Farside Investors, the US Bitcoin spot ETF recorded a net inflow of $117.5 million yesterday. The breakdown is as follows: BlackRock IBIT: +$98.4 million, Fidelity FBTC: +$16.2 million, Bitwise BITB: +$1.8 million, ARK ARKB: +$1.1 million. Additionally, the US Ethereum spot ETF saw a net inflow of $31.2 million yesterday, with the following contributions: BlackRock ETHA: +$24.7 million, BlackRock ETHB: +$1.1 million, Fidelity FETH: +$1.6 million, Bitwise ETHW: +$1.2 million, 21 Shares TETH: +$2.6 million.

[US Bitcoin Spot ETF Sees $117.5 Million Net Inflow, Ethereum ETF Gains $31.2 Million]On April 1, according to monitoring by Farside Investors, the US Bitcoin spot ETF recorded a net inflow of $117.5 million yesterday. The breakdown is as follows: BlackRock IBIT: +$98.4 million, Fidelity FBTC: +$16.2 million, Bitwise BITB: +$1.8 million, ARK ARKB: +$1.1 million. Additionally, the US Ethereum spot ETF saw a net inflow of $31.2 million yesterday, with the following contributions: BlackRock ETHA: +$24.7 million, BlackRock ETHB: +$1.1 million, Fidelity FETH: +$1.6 million, Bitwise ETHW: +$1.2 million, 21 Shares TETH: +$2.6 million.

US Bitcoin Spot ETF Sees $117.5 Million Net Inflow, Ethereum ETF Gains $31.2 Million

On April 1, Cointelegraph reported that Arizona's cryptocurrency-related bill has passed the House Rules Committee and is set to quickly enter the full House voting process. Senate Bill 1042 (SB1042) proposes to allow the state's financial funds to allocate up to 10% for investment in cryptocurrency assets, while Senate Bill 1649 (SB1649) aims to establish an official digital asset reserve fund.

[Arizona Cryptocurrency Bill Passes House Rules Committee, Set for Full House Vote]On April 1, Cointelegraph reported that Arizona's cryptocurrency-related bill has passed the House Rules Committee and is set to quickly enter the full House voting process. Senate Bill 1042 (SB1042) proposes to allow the state's financial funds to allocate up to 10% for investment in cryptocurrency assets, while Senate Bill 1649 (SB1649) aims to establish an official digital asset reserve fund.

Arizona Cryptocurrency Bill Passes House Rules Committee, Set for Full House Vote

On April 1, according to on-chain analyst Yu Jin, a SIREN 'whale' on-chain address spent 144,000 USDT half an hour ago to buy back 500,000 SIREN at a price of 0.288 USD. This 500,000 SIREN was sold off earlier in the month for 0.947 USD, converting to 473,000 USDT. After the price drop, they spent 144,000 USDT to repurchase the tokens. The controlling address currently holds 645 million SIREN, valued at 155 million USD, down from a value of 1.44 billion USD a week ago.

[On-chain Address Buys Back 500,000 SIREN for 144,000 USDT]On April 1, according to on-chain analyst Yu Jin, a SIREN 'whale' on-chain address spent 144,000 USDT half an hour ago to buy back 500,000 SIREN at a price of 0.288 USD. This 500,000 SIREN was sold off earlier in the month for 0.947 USD, converting to 473,000 USDT. After the price drop, they spent 144,000 USDT to repurchase the tokens. The controlling address currently holds 645 million SIREN, valued at 155 million USD, down from a value of 1.44 billion USD a week ago.

On-chain Address Buys Back 500,000 SIREN for 144,000 USDT

On April 1, Tom Lee stated in an interview with CNBC: "I believe the market has absorbed 90% to 95% of the selling pressure, and the process of selling may have come to an end. We can start to build a bottom now. We must recognize that in a wartime environment, the stock market often bottoms out early. We have studied every war since 1900, and the stock market tends to hit its bottom within the first 10% of the war's progression. So, if this follows the same pattern, we are currently in the early stages of that process. At this stage, any piece of bad news can trigger risk-off behavior. But you know, this is also why positioning is worth paying attention to, because at some point, when people become overly neutral, even if the situation isn't that bad, the market could see a V-shaped rebound." Tom Lee also added on social media that even if the "bottom" has not yet been reached, he believes the U.S. economy can withstand oil prices of $100, or even $120.

[Tom Lee: Market Has Digested Over 90% of Selling Pressure, Stocks Typically Bottom in Early Stages of War]On April 1, Tom Lee stated in an interview with CNBC: "I believe the market has absorbed 90% to 95% of the selling pressure, and the process of selling may have come to an end. We can start to build a bottom now. We must recognize that in a wartime environment, the stock market often bottoms out early. We have studied every war since 1900, and the stock market tends to hit its bottom within the first 10% of the war's progression. So, if this follows the same pattern, we are currently in the early stages of that process. At this stage, any piece of bad news can trigger risk-off behavior. But you know, this is also why positioning is worth paying attention to, because at some point, when people become overly neutral, even if the situation isn't that bad, the market could see a V-shaped rebound." Tom Lee also added on social media that even if the "bottom" has not yet been reached, he believes the U.S. economy can withstand oil prices of $100, or even $120.

Tom Lee: Market Has Digested Over 90% of Selling Pressure, Stocks Typically Bottom in Early Stages of War

According to monitoring by 1M AI News, the Claude Code source code leak incident has triggered a chain of attributions. Developer jakeg questioned under Bun's GitHub issue #28001: "Claude Code is built using Bun, and there is a known bug in Bun (`Bun.serve()` still provides source map files in production mode). Is this the root cause of the leak?" The comment quickly garnered nearly a hundred reactions, with many developers joining the discussion. Anthropic acquired Bun in December 2025, with the acquisition announcement stating, "Bun is key in the infrastructure expansion process for Claude Code," and Jarred Sumner and his team joined Anthropic with the acquisition. Jarred Sumner personally responded, stating, "This has nothing to do with Claude Code. This bug targets Bun's frontend development server. Claude Code is not a frontend application; it is a TUI (terminal user interface program) and does not use Bun.serve() to compile single-file executables." He then locked the issue to prevent non-collaborators from continuing to comment and changed the title to emphasize "Bun's frontend development server" to avoid misleading interpretations. The two issues are technically completely different. Bun #28001 is a configuration bug where `Bun.serve()` still exposes `.map` files to the browser after setting `development: false`, which pertains to the frontend server; the leak of Claude Code involved approximately 60MB of source map files accidentally included in the v2.1.88 npm package, which Anthropic described as a "release packaging issue caused by human error," related to CI/CD build configuration errors. This Bun bug has existed for three weeks since it was submitted on March 11 and has yet to be merged for a fix.

[Is Claude Code Leak Caused by Bun Bug? Bun Founder Denies and Locks Thread]According to monitoring by 1M AI News, the Claude Code source code leak incident has triggered a chain of attributions. Developer jakeg questioned under Bun's GitHub issue #28001: "Claude Code is built using Bun, and there is a known bug in Bun (`Bun.serve()` still provides source map files in production mode). Is this the root cause of the leak?" The comment quickly garnered nearly a hundred reactions, with many developers joining the discussion. Anthropic acquired Bun in December 2025, with the acquisition announcement stating, "Bun is key in the infrastructure expansion process for Claude Code," and Jarred Sumner and his team joined Anthropic with the acquisition. Jarred Sumner personally responded, stating, "This has nothing to do with Claude Code. This bug targets Bun's frontend development server. Claude Code is not a frontend application; it is a TUI (terminal user interface program) and does not use Bun.serve() to compile single-file executables." He then locked the issue to prevent non-collaborators from continuing to comment and changed the title to emphasize "Bun's frontend development server" to avoid misleading interpretations. The two issues are technically completely different. Bun #28001 is a configuration bug where `Bun.serve()` still exposes `.map` files to the browser after setting `development: false`, which pertains to the frontend server; the leak of Claude Code involved approximately 60MB of source map files accidentally included in the v2.1.88 npm package, which Anthropic described as a "release packaging issue caused by human error," related to CI/CD build configuration errors. This Bun bug has existed for three weeks since it was submitted on March 11 and has yet to be merged for a fix.

Is Claude Code Leak Caused by Bun Bug? Bun Founder Denies and Locks Thread

According to CryptoQuant data, Bitcoin's realized price (the average cost basis of all coins on the network weighted by the last transaction) is $54,286, while the spot price is $68,774, a difference of approximately $14,500, or about 21% above the realized price. Looking back at the bear market of 2022, a key signal for the actual bottom was when the spot price fell below the realized price. From June to October 2022, Bitcoin's spot price was below the overall cost basis, and when the spot price was about 15% lower than the realized price, this decline almost perfectly corresponded to the cycle's low point of around $15,500. A similar drop occurred during the early 2020 COVID crash. These periods are true accumulation zones, as the entire network was in a state of loss. Historically, buying when the market is overall in loss has been one of the most reliable entry signals for Bitcoin. Currently, the situation is different. The spot price is 21% higher than the realized price, indicating that the average holder is still in profit, which provides a significant buffer. For Bitcoin to fall back to the realized price from current levels, it would need to drop to around $54,000, representing a further decline of 20%. Notably, the speed at which the gap between the spot price and the realized price is narrowing is very rapid. By the end of 2024, if Bitcoin trades above $119,000, the premium would be around 120%, but it has compressed to 21% within about 15 months, making this one of the fastest periods approaching the realized price line, barring a complete collapse. CryptoQuant analyst Oinonen noted on Monday that Bitcoin has entered a so-called 'accumulation zone' and compared it to the bottom of 2022. However, this judgment may be premature. CryptoQuant's own charts show that the definition of the accumulation zone in 2022 was when the spot price traded at or below the realized price. The box they drew on the current price trend indicates that the spot price is still well above the indicators defining that range.

[Analyst: Bitcoin is Nearing Buy Zone Not Seen in Nearly Three Years]According to CryptoQuant data, Bitcoin's realized price (the average cost basis of all coins on the network weighted by the last transaction) is $54,286, while the spot price is $68,774, a difference of approximately $14,500, or about 21% above the realized price. Looking back at the bear market of 2022, a key signal for the actual bottom was when the spot price fell below the realized price. From June to October 2022, Bitcoin's spot price was below the overall cost basis, and when the spot price was about 15% lower than the realized price, this decline almost perfectly corresponded to the cycle's low point of around $15,500. A similar drop occurred during the early 2020 COVID crash. These periods are true accumulation zones, as the entire network was in a state of loss. Historically, buying when the market is overall in loss has been one of the most reliable entry signals for Bitcoin. Currently, the situation is different. The spot price is 21% higher than the realized price, indicating that the average holder is still in profit, which provides a significant buffer. For Bitcoin to fall back to the realized price from current levels, it would need to drop to around $54,000, representing a further decline of 20%. Notably, the speed at which the gap between the spot price and the realized price is narrowing is very rapid. By the end of 2024, if Bitcoin trades above $119,000, the premium would be around 120%, but it has compressed to 21% within about 15 months, making this one of the fastest periods approaching the realized price line, barring a complete collapse. CryptoQuant analyst Oinonen noted on Monday that Bitcoin has entered a so-called 'accumulation zone' and compared it to the bottom of 2022. However, this judgment may be premature. CryptoQuant's own charts show that the definition of the accumulation zone in 2022 was when the spot price traded at or below the realized price. The box they drew on the current price trend indicates that the spot price is still well above the indicators defining that range.

Analyst: Bitcoin is Nearing Buy Zone Not Seen in Nearly Three Years

On April 1, the official announcement from Cosmos ecosystem lending blockchain UX Chain stated that it will gradually suspend operations. Users must clear their open positions before the full platform liquidation process begins on April 30. The protocol is expected to be completely shut down by May 15. UX reserves and treasury will be used to cover the funding gap of the protocol.

[Cosmos Ecosystem UX Chain to Gradually Suspend Operations, Expected to Fully Shut Down by May 15]On April 1, the official announcement from Cosmos ecosystem lending blockchain UX Chain stated that it will gradually suspend operations. Users must clear their open positions before the full platform liquidation process begins on April 30. The protocol is expected to be completely shut down by May 15. UX reserves and treasury will be used to cover the funding gap of the protocol.

Cosmos Ecosystem UX Chain to Gradually Suspend Operations, Expected to Fully Shut Down by May 15

According to CoinGecko, on August 26th, the total trading volume of the entire cryptocurrency market was $28.883 billion, a decrease of approximately 15.5% compared to August 25th ($34.185 billion).

[The total trading volume of the encryption market yesterday was 28.883 billion US dollars, a single-day drop of 15.5%]According to CoinGecko, on August 26th, the total trading volume of the entire cryptocurrency market was $28.883 billion, a decrease of approximately 15.5% compared to August 25th ($34.185 billion).

The total trading volume of the encryption market yesterday was 28.883 billion US dollars, a single-day drop of 15.5%

Article

Google: Quantum Computer Can Crack Bitcoin Private Keys in Just 9 Minutes

BTC fell below 66000  USDT

"PINDex Global Launch Ceremony" Successfully Concludes, Ushering in a New Era of AI-Native Finance

Claude Code's Latest npm Package Accidentally Includes 60MB Source Map, Source Code Leaked Again After a Year

OpenEden Extends Lock-Up Period for All Team and Advisor Tokens by 9 Months

Alibaba Tongyi Releases CoPaw 1.0 with Custom Models, Security Mechanisms, Multi-Agent Collaboration, and Memory Management

The total trading volume of the encryption market yesterday was 28.883 billion US dollars, a single-day drop of 15.5%

Clause de non-responsabilité：Le contenu ci-dessus ne représente pas les positions de HTX.，HTX ne fournit pas de recommandations de trading.。

Tous les commentaires0RécentPopulaire

Sujets d'actualité