7*24

Hot topic

Market

Activities

According to monitoring by 1M AI News, Meta is preparing to release a new generation of AI models developed under the leadership of Alexandr Wang, with plans to offer open-source versions of some of the models. Wang joined Meta last year through a $15 billion deal from AI data infrastructure company Scale AI, where he previously served as CEO. There had been speculation that Meta might completely abandon the open-source route, but according to Axios, citing informed sources, Wang believes that Meta can play a unique role in popularizing the latest AI technologies while ensuring developers have access to an American-made open-source option. However, Meta's largest new models will remain closed-source, marking a shift from the company's previous full open-source approach to a hybrid strategy that includes both open and closed-source elements. Wang's strategy focuses on differentiated competition: while Anthropic and OpenAI increasingly target government and enterprise clients, Meta aims for the consumer market, with the goal of distributing its models as widely as possible to global users. Meta's advantage lies in its global reach through its three free platforms: WhatsApp, Facebook, and Instagram, which provide a distribution channel that is difficult for competitors to match. Meta acknowledges that its new models may not fully compete with the next-generation models soon to be released by OpenAI and Anthropic across all dimensions, but it believes it will have its own competitive edge in consumer scenarios. Previously, the Llama 4 series significantly lagged behind competitors, and the primary goal of the new model series is to close that gap. The trend of reducing open-source initiatives is not limited to Meta alone; Alibaba has also recently transitioned its strongest new Qianwen model to closed-source, reversing its earlier strategy of full openness. The industry is forming a consensus: even the most open-minded companies are retaining control over their most powerful systems.

[Meta's New AI Model to Adopt Hybrid Strategy with Partial Open Source]According to monitoring by 1M AI News, Meta is preparing to release a new generation of AI models developed under the leadership of Alexandr Wang, with plans to offer open-source versions of some of the models. Wang joined Meta last year through a $15 billion deal from AI data infrastructure company Scale AI, where he previously served as CEO. There had been speculation that Meta might completely abandon the open-source route, but according to Axios, citing informed sources, Wang believes that Meta can play a unique role in popularizing the latest AI technologies while ensuring developers have access to an American-made open-source option. However, Meta's largest new models will remain closed-source, marking a shift from the company's previous full open-source approach to a hybrid strategy that includes both open and closed-source elements. Wang's strategy focuses on differentiated competition: while Anthropic and OpenAI increasingly target government and enterprise clients, Meta aims for the consumer market, with the goal of distributing its models as widely as possible to global users. Meta's advantage lies in its global reach through its three free platforms: WhatsApp, Facebook, and Instagram, which provide a distribution channel that is difficult for competitors to match. Meta acknowledges that its new models may not fully compete with the next-generation models soon to be released by OpenAI and Anthropic across all dimensions, but it believes it will have its own competitive edge in consumer scenarios. Previously, the Llama 4 series significantly lagged behind competitors, and the primary goal of the new model series is to close that gap. The trend of reducing open-source initiatives is not limited to Meta alone; Alibaba has also recently transitioned its strongest new Qianwen model to closed-source, reversing its earlier strategy of full openness. The industry is forming a consensus: even the most open-minded companies are retaining control over their most powerful systems.

Meta's New AI Model to Adopt Hybrid Strategy with Partial Open Source

According to monitoring by 1M AI News, an AI usage ranking named 'Claudeonomics', after Anthropic's flagship product Claude, has surfaced within Meta's internal network. This ranking was created by employees using company data, compiling token consumption from over 85,000 individuals and listing the top 250 users. A copy of the ranking seen by The Information shows that the total consumption in the past 30 days has surpassed 60 trillion tokens, which roughly translates to about $900 million based on the public average price of Claude Opus 4.6 (approximately $15 per million tokens), though the actual model combinations and pricing protocols at Meta remain unclear. The top individual user has an average consumption of 281 billion tokens, potentially costing millions of dollars. The ranking features gamified incentives, with tiers ranging from bronze to emerald, and the highest titles including 'Token Legend' and 'Session Immortal', along with 'Model Connoisseur' and 'Cache Wizard'. Some employees have been running AI agents for hours on end purely to inflate their usage for ranking purposes. Meta CEO Mark Zuckerberg and CTO Andrew Bosworth did not make it into the top 250. Recently, there has been a trend in Silicon Valley known as 'tokenmaxxing', where token consumption is becoming a new metric for measuring engineer productivity. Bosworth mentioned at a tech conference in February that the token expenditure of a top engineer is equivalent to their salary, with productivity improvements reaching up to 10 times, stating, 'This is a no-lose deal; keep burning, there’s no limit.' NVIDIA CEO Jensen Huang remarked last month that if an engineer with a $500,000 annual salary consumes less than $250,000 in tokens each year, he would be 'deeply concerned'. Meta engineers are currently using external models from Anthropic, OpenAI, and Google, as well as internal tools like MyClaw (Meta's version of OpenClaw) and the recently acquired Manus. In an internal memo this year, Zuckerberg made a 'bold request' to the engineering team to rewrite Meta's codebase to enable AI agents to directly understand and modify the code.

[Meta's Internal 'Claudeonomics' Ranking Revealed: 85,000 Employees Burn 60 Trillion Tokens in 30 Days]According to monitoring by 1M AI News, an AI usage ranking named 'Claudeonomics', after Anthropic's flagship product Claude, has surfaced within Meta's internal network. This ranking was created by employees using company data, compiling token consumption from over 85,000 individuals and listing the top 250 users. A copy of the ranking seen by The Information shows that the total consumption in the past 30 days has surpassed 60 trillion tokens, which roughly translates to about $900 million based on the public average price of Claude Opus 4.6 (approximately $15 per million tokens), though the actual model combinations and pricing protocols at Meta remain unclear. The top individual user has an average consumption of 281 billion tokens, potentially costing millions of dollars. The ranking features gamified incentives, with tiers ranging from bronze to emerald, and the highest titles including 'Token Legend' and 'Session Immortal', along with 'Model Connoisseur' and 'Cache Wizard'. Some employees have been running AI agents for hours on end purely to inflate their usage for ranking purposes. Meta CEO Mark Zuckerberg and CTO Andrew Bosworth did not make it into the top 250. Recently, there has been a trend in Silicon Valley known as 'tokenmaxxing', where token consumption is becoming a new metric for measuring engineer productivity. Bosworth mentioned at a tech conference in February that the token expenditure of a top engineer is equivalent to their salary, with productivity improvements reaching up to 10 times, stating, 'This is a no-lose deal; keep burning, there’s no limit.' NVIDIA CEO Jensen Huang remarked last month that if an engineer with a $500,000 annual salary consumes less than $250,000 in tokens each year, he would be 'deeply concerned'. Meta engineers are currently using external models from Anthropic, OpenAI, and Google, as well as internal tools like MyClaw (Meta's version of OpenClaw) and the recently acquired Manus. In an internal memo this year, Zuckerberg made a 'bold request' to the engineering team to rewrite Meta's codebase to enable AI agents to directly understand and modify the code.

Meta's Internal 'Claudeonomics' Ranking Revealed: 85,000 Employees Burn 60 Trillion Tokens in 30 Days

On April 7, several former OpenAI employees jointly established a venture capital fund named Zero Shot, with a target size of $100 million. The first round has raised approximately $20 million and completed several investments. The fund was initiated by Evan Morikawa, Andrew Mayne, Shawn Jain, along with participation from Kelly Kovacs and Brett Rounsaville. The invested projects include the enterprise automation platform Worktrace AI and the robotics company Foundry Robotics. The team stated that they will focus on AI sectors with long-term potential while adopting a cautious approach towards certain areas, including 'vibecoding' platforms and robotic video data training.

[Former OpenAI Employees Launch $100 Million VC Fund Zero Shot, Focusing on AI Investments]On April 7, several former OpenAI employees jointly established a venture capital fund named Zero Shot, with a target size of $100 million. The first round has raised approximately $20 million and completed several investments. The fund was initiated by Evan Morikawa, Andrew Mayne, Shawn Jain, along with participation from Kelly Kovacs and Brett Rounsaville. The invested projects include the enterprise automation platform Worktrace AI and the robotics company Foundry Robotics. The team stated that they will focus on AI sectors with long-term potential while adopting a cautious approach towards certain areas, including 'vibecoding' platforms and robotic video data training.

Former OpenAI Employees Launch $100 Million VC Fund Zero Shot, Focusing on AI Investments

On April 7, the cryptocurrency wallet application Phantom experienced a temporary service interruption, impacting token prices and balance displays. The team is actively working to resolve the related issues.

[Phantom Experiences Temporary Service Interruption Affecting Token Prices and Balance Display]On April 7, the cryptocurrency wallet application Phantom experienced a temporary service interruption, impacting token prices and balance displays. The team is actively working to resolve the related issues.

Phantom Experiences Temporary Service Interruption Affecting Token Prices and Balance Display

According to monitoring by 1M AI News, Joanne Jang, General Manager of OpenAI Labs, announced her departure on X, concluding her 4.5-year tenure at OpenAI. She joined when the company was still a research lab with fewer than 200 employees. Jang was involved in the launch of GPT-4, DALL·E 2, text-to-speech, Chat API, and memory features, and established 'Model Behavior' as an independent discipline, overseeing the personality design and post-training reinforcement learning for GPT-4o, GPT-4.5, and o3. She also led the development of OpenAI's AI behavior guidelines, Model Spec. In September 2025, she transitioned from the model behavior team to establish OpenAI Labs, exploring new interaction paradigms for human-AI collaboration. Jang was named a 'Thinker' in TIME's 100 Most Influential People in AI in 2025. In her farewell letter, she expressed that if she could be remembered for one thing, 'I hope it is that I brought frogs and pink text into OpenAI'.

[OpenAI's Former Model Behavior Lead Joanne Jang Announces Departure, Led ChatGPT Personality Design]According to monitoring by 1M AI News, Joanne Jang, General Manager of OpenAI Labs, announced her departure on X, concluding her 4.5-year tenure at OpenAI. She joined when the company was still a research lab with fewer than 200 employees. Jang was involved in the launch of GPT-4, DALL·E 2, text-to-speech, Chat API, and memory features, and established 'Model Behavior' as an independent discipline, overseeing the personality design and post-training reinforcement learning for GPT-4o, GPT-4.5, and o3. She also led the development of OpenAI's AI behavior guidelines, Model Spec. In September 2025, she transitioned from the model behavior team to establish OpenAI Labs, exploring new interaction paradigms for human-AI collaboration. Jang was named a 'Thinker' in TIME's 100 Most Influential People in AI in 2025. In her farewell letter, she expressed that if she could be remembered for one thing, 'I hope it is that I brought frogs and pink text into OpenAI'.

OpenAI's Former Model Behavior Lead Joanne Jang Announces Departure, Led ChatGPT Personality Design

On April 7, Aave founder Stani Kulechov tweeted that Aave respects Chaos Labs' decision to exit its risk management role and thanked them for their contributions over the past years. The Aave protocol's smart contracts, asset listings, and network deployments will remain unaffected, and both parties will maintain close cooperation during the exit process. Aave currently employs a dual-layer economic risk management model, jointly managed by Chaos Labs and LlamaRisk. Given the scale of the Aave protocol, it will continue to uphold this dual model in the future, and additionally introduce a technical risk layer managed by Aave Labs. Aave Labs will work closely with LlamaRisk to ensure uninterrupted risk coverage during the transition and support LlamaRisk in increasing its budget and team size.

[Aave Founder Responds to Chaos Labs Exit: Will Maintain Dual Risk Management Model and Support LlamaRisk Takeover]On April 7, Aave founder Stani Kulechov tweeted that Aave respects Chaos Labs' decision to exit its risk management role and thanked them for their contributions over the past years. The Aave protocol's smart contracts, asset listings, and network deployments will remain unaffected, and both parties will maintain close cooperation during the exit process. Aave currently employs a dual-layer economic risk management model, jointly managed by Chaos Labs and LlamaRisk. Given the scale of the Aave protocol, it will continue to uphold this dual model in the future, and additionally introduce a technical risk layer managed by Aave Labs. Aave Labs will work closely with LlamaRisk to ensure uninterrupted risk coverage during the transition and support LlamaRisk in increasing its budget and team size.

Aave Founder Responds to Chaos Labs Exit: Will Maintain Dual Risk Management Model and Support LlamaRisk Takeover

On April 7, the Polygon Foundation announced that the Giugliano hard fork will be released at block height 85,268,500 on the Polygon mainnet, expected to occur at 2 PM UTC on April 8. This upgrade includes: achieving faster finality by allowing block producers to broadcast blocks earlier; incorporating fee parameters directly into the block header; and adding RPC support for fee data. All node operators must upgrade Bor to v2.7.0 (or upgrade Erigon to v3.5.0).

[Polygon Foundation: Giugliano Hard Fork Scheduled for April 8 on Polygon Mainnet]On April 7, the Polygon Foundation announced that the Giugliano hard fork will be released at block height 85,268,500 on the Polygon mainnet, expected to occur at 2 PM UTC on April 8. This upgrade includes: achieving faster finality by allowing block producers to broadcast blocks earlier; incorporating fee parameters directly into the block header; and adding RPC support for fee data. All node operators must upgrade Bor to v2.7.0 (or upgrade Erigon to v3.5.0).

Polygon Foundation: Giugliano Hard Fork Scheduled for April 8 on Polygon Mainnet

On April 7, Aave founder Stani Kulechov tweeted that Aave respects Chaos Labs' decision to exit its risk management role and thanked them for their contributions over the past years. The Aave protocol's smart contracts, asset listings, and network deployments will remain unaffected, and both parties will maintain close cooperation during the exit process. Aave currently employs a dual-layer economic risk management model, jointly managed by Chaos Labs and LlamaRisk. Given the scale of the Aave protocol, it will continue to maintain this dual model in the future, with an additional technical risk layer managed by Aave Labs. Aave Labs will work closely with LlamaRisk to ensure uninterrupted risk coverage during the transition and support LlamaRisk in increasing its budget and team size.

[Aave Founder Responds to Chaos Labs Exit: Will Maintain Dual Risk Management Model and Support LlamaRisk Takeover]On April 7, Aave founder Stani Kulechov tweeted that Aave respects Chaos Labs' decision to exit its risk management role and thanked them for their contributions over the past years. The Aave protocol's smart contracts, asset listings, and network deployments will remain unaffected, and both parties will maintain close cooperation during the exit process. Aave currently employs a dual-layer economic risk management model, jointly managed by Chaos Labs and LlamaRisk. Given the scale of the Aave protocol, it will continue to maintain this dual model in the future, with an additional technical risk layer managed by Aave Labs. Aave Labs will work closely with LlamaRisk to ensure uninterrupted risk coverage during the transition and support LlamaRisk in increasing its budget and team size.

On April 7, sources reported that SpaceX plans to publicly release its IPO prospectus by the end of May. The company is set to kick off its IPO roadshow during the week of June 8. (Jin Shi)

[Sources: SpaceX Plans to Release IPO Prospectus by End of May]On April 7, sources reported that SpaceX plans to publicly release its IPO prospectus by the end of May. The company is set to kick off its IPO roadshow during the week of June 8. (Jin Shi)

Sources: SpaceX Plans to Release IPO Prospectus by End of May

&nbsp;according to on-chain analyst monitoring, the address 0x2d4…f2B9A accumulated 5,089.56 ETH (9.943 million USD) in the past 23 hours at an average price of 1,953.65 USD, just buying at the low point before the ETH rebound; then sold 4,000 ETH (8.235 million USD) at the short-term high of 2,058.91 USD nine hours ago, gaining a profit of 421,000 USD; now the remaining 1,089.56 ETH still has an unrealized profit of 104,000 USD.

[A major whale engaged in swing trading of ETH, buying 5089 coins at the low point and selling 4000 coins at the high point.]&nbsp;according to on-chain analyst monitoring, the address 0x2d4…f2B9A accumulated 5,089.56 ETH (9.943 million USD) in the past 23 hours at an average price of 1,953.65 USD, just buying at the low point before the ETH rebound; then sold 4,000 ETH (8.235 million USD) at the short-term high of 2,058.91 USD nine hours ago, gaining a profit of 421,000 USD; now the remaining 1,089.56 ETH still has an unrealized profit of 104,000 USD.

A major whale engaged in swing trading of ETH, buying 5089 coins at the low point and selling 4000 coins at the high point.

Article

Arc Blockchain Releases Post-Quantum Cryptography Roadmap

China's AI Model Usage Surges Over 31%, Outpacing the U.S. for Five Consecutive Weeks

Software Engineering Job Openings Reach Three-Year High; a16z Co-Founder Andreessen Calls AI Unemployment Claims 'Completely False'

Santiment: Bitcoin Shows Short-Term Price Top Signal at Weekend Close

Iranian Senior Official: Iran Will Not Accept 'Diktat'

BTC broke through 70000  USDT

A major whale engaged in swing trading of ETH, buying 5089 coins at the low point and selling 4000 coins at the high point.

Disclaimer：The content above does not represent HTX's positions.，HTX does not provide any trading recommendations.。

Related Articles

All Comments0LatestHot

Trending Topics

Related Articles