7*24

Hot topic

Market

Activities

According to 1M AI News, OpenAI co-founder Greg Brockman was asked in an interview whether he agrees with Nvidia CEO Jensen Huang's statement that "AGI has already been achieved." He provided an uncertain response: "By my personal definition, I would say I'm about 70-80% there." He believes AGI will be realized in a few years, at which point "almost all intellectual tasks, everything you can do on a computer, AI will be able to do," but the process will still be "jagged." Currently, AI has surpassed humans in many tasks, such as coding, but still struggles with some "very basic tasks that humans can easily accomplish." He illustrated the complexity of his feelings with a personal comparison: "If you showed me today's system five years ago, I would say 'Oh yes, that's what we were talking about.' But it is different from what any of us imagined." The next concrete milestone towards AGI is the planned launch of an "automated AI researcher" this fall. When pressed about what it would specifically do, Brockman said: "It will take all the work that a research scientist does from start to finish and put it onto silicon." He compared this system to managing a junior researcher: it cannot be left unattended for too long, or it will go in unproductive directions, but having senior researchers provide feedback, review charts, and guide the direction is sufficient. He defined the current stage as "the early phase of takeoff": the stronger the AI, the more it can accelerate its own improvement, creating a positive feedback loop; the external supply chain is also accelerating, with chip developers investing more resources, and the developer economy around AI continuously expanding, all energies are converging. He also disclosed an internal cultural detail: after the release of ChatGPT, he felt a sense of "we won" at the company's holiday party, which he described as the most frightening moment for him at OpenAI. "No, we are challengers, we always have been." He believes that the emergence of competitors is "very healthy for OpenAI," as it has helped the company realign its direction.

["If you showed me today's system five years ago, I would say that's AGI": OpenAI co-founder assesses progress at 70-80%, plans to launch AI researcher this fall]According to 1M AI News, OpenAI co-founder Greg Brockman was asked in an interview whether he agrees with Nvidia CEO Jensen Huang's statement that "AGI has already been achieved." He provided an uncertain response: "By my personal definition, I would say I'm about 70-80% there." He believes AGI will be realized in a few years, at which point "almost all intellectual tasks, everything you can do on a computer, AI will be able to do," but the process will still be "jagged." Currently, AI has surpassed humans in many tasks, such as coding, but still struggles with some "very basic tasks that humans can easily accomplish." He illustrated the complexity of his feelings with a personal comparison: "If you showed me today's system five years ago, I would say 'Oh yes, that's what we were talking about.' But it is different from what any of us imagined." The next concrete milestone towards AGI is the planned launch of an "automated AI researcher" this fall. When pressed about what it would specifically do, Brockman said: "It will take all the work that a research scientist does from start to finish and put it onto silicon." He compared this system to managing a junior researcher: it cannot be left unattended for too long, or it will go in unproductive directions, but having senior researchers provide feedback, review charts, and guide the direction is sufficient. He defined the current stage as "the early phase of takeoff": the stronger the AI, the more it can accelerate its own improvement, creating a positive feedback loop; the external supply chain is also accelerating, with chip developers investing more resources, and the developer economy around AI continuously expanding, all energies are converging. He also disclosed an internal cultural detail: after the release of ChatGPT, he felt a sense of "we won" at the company's holiday party, which he described as the most frightening moment for him at OpenAI. "No, we are challengers, we always have been." He believes that the emergence of competitors is "very healthy for OpenAI," as it has helped the company realign its direction.

"If you showed me today's system five years ago, I would say that's AGI": OpenAI co-founder assesses progress at 70-80%, plans to launch AI researcher this fall

On April 2, according to monitoring by HyperInsight, as Ethereum fell to the 'Maji' liquidation price (approximately $2049.15), the latter significantly reduced its holdings once again. Two hours prior, during a rapid market decline triggered by Trump's speech, 'Maji' was still buying on the dips. As of the time of this report, 'Maji' has reduced its 25x Ethereum long positions to 7,200 units.

[Ethereum Drops to 'Maji' Liquidation Price, Significant Reduction in Holdings]On April 2, according to monitoring by HyperInsight, as Ethereum fell to the 'Maji' liquidation price (approximately $2049.15), the latter significantly reduced its holdings once again. Two hours prior, during a rapid market decline triggered by Trump's speech, 'Maji' was still buying on the dips. As of the time of this report, 'Maji' has reduced its 25x Ethereum long positions to 7,200 units.

Ethereum Drops to 'Maji' Liquidation Price, Significant Reduction in Holdings

On April 2, Senate Democratic Leader Chuck Schumer sharply criticized President Trump's nationwide address on Iran's military actions delivered on April 1, describing it as "disorganized and illogical." He stated that Trump failed to clearly outline the objectives of the military actions against Iran, alienated allies, and ignored the everyday issues faced by the American public. Schumer wrote on social media, "Is there a more disorganized, illogical, and sadly delivered presidential war speech than this? Donald Trump's actions in Iran will be seen as one of the most significant policy failures in American history."

[Senate Democratic Leader Criticizes Trump's Speech]On April 2, Senate Democratic Leader Chuck Schumer sharply criticized President Trump's nationwide address on Iran's military actions delivered on April 1, describing it as "disorganized and illogical." He stated that Trump failed to clearly outline the objectives of the military actions against Iran, alienated allies, and ignored the everyday issues faced by the American public. Schumer wrote on social media, "Is there a more disorganized, illogical, and sadly delivered presidential war speech than this? Donald Trump's actions in Iran will be seen as one of the most significant policy failures in American history."

Senate Democratic Leader Criticizes Trump's Speech

On April 2, the Drift protocol vulnerability incident has impacted multiple DeFi protocols. Reflect Money, Ranger Finance, Neutral Trade, Elemental DeFi, Project 0, Lulo Finance, Asgard Finance, DeFi Carrot, Pyra, xPlace, and Fuse Wallet have confirmed they are affected, with some protocols suspending minting, redemption, or deposit functions. Ranger Finance confirmed a risk exposure of approximately $900,000, accounting for about 6% of its total locked value of $14.6 million. Pyra stated that user funds were affected due to being deposited in Drift for yield, and has suspended the Pyra Card function. Asgard Finance indicated that the risk exposure related to Drift is not significant, has disabled that credit source, and contacted affected users. Fuse Wallet has suspended deposits to Drift for its Earn product, but the wallet itself is unaffected. DeFi Carrot has paused minting and redemption functions, while its Boost and Turbo products remain unaffected. xPlace has suspended deposits and withdrawals for its Savings product, and temporarily disabled its credit model and lending functions. Protocols such as Elemental DeFi and Project 0 have stated that related fund allocations have been paused, awaiting Drift's recovery. Lulo Finance warned that Classic deposit users may be affected, while its Protected and Boosted products have no risk exposure.

[11 DeFi Protocols Affected by Drift Security Incident, Some Suspend Core Functions Awaiting Recovery]On April 2, the Drift protocol vulnerability incident has impacted multiple DeFi protocols. Reflect Money, Ranger Finance, Neutral Trade, Elemental DeFi, Project 0, Lulo Finance, Asgard Finance, DeFi Carrot, Pyra, xPlace, and Fuse Wallet have confirmed they are affected, with some protocols suspending minting, redemption, or deposit functions. Ranger Finance confirmed a risk exposure of approximately $900,000, accounting for about 6% of its total locked value of $14.6 million. Pyra stated that user funds were affected due to being deposited in Drift for yield, and has suspended the Pyra Card function. Asgard Finance indicated that the risk exposure related to Drift is not significant, has disabled that credit source, and contacted affected users. Fuse Wallet has suspended deposits to Drift for its Earn product, but the wallet itself is unaffected. DeFi Carrot has paused minting and redemption functions, while its Boost and Turbo products remain unaffected. xPlace has suspended deposits and withdrawals for its Savings product, and temporarily disabled its credit model and lending functions. Protocols such as Elemental DeFi and Project 0 have stated that related fund allocations have been paused, awaiting Drift's recovery. Lulo Finance warned that Classic deposit users may be affected, while its Protected and Boosted products have no risk exposure.

11 DeFi Protocols Affected by Drift Security Incident, Some Suspend Core Functions Awaiting Recovery

According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman reflected on the leap in AI programming capabilities expected by December 2025 during an interview. He used a test prompt he had kept for years to measure progress: asking the AI to build a website that took him months to complete when he was learning programming. Throughout 2025, this task required multiple prompts and about four hours to accomplish; by December, it could be completed with a single prompt and with high quality. He stated that the new model allowed AI to jump from 'being able to complete about 20% of tasks' to 'about 80%', a shift that forces everyone to 'reorganize workflows around AI.' Regarding the allocation of the $110 billion in funding, Brockman likened computing power to 'hiring salespeople': as long as the product has a scalable sales channel, hiring more salespeople can generate more revenue. Computing power is not a cost center but a revenue center. He recalled a conversation with his team on the eve of ChatGPT's release: 'They asked, 'How much computing power should we buy?' I said, 'All of it.' They replied, 'No, no, no, seriously, how much should we buy?' I said, 'No matter how we build, we won't keep up with demand.' This judgment still holds true today, and computing power procurement needs to be locked in 18 to 24 months in advance. On how to utilize this computing power, Brockman revealed that OpenAI is no longer solely pursuing the largest scale of pre-training but is instead jointly optimizing pre-training capabilities and inference costs: 'You don't necessarily have to make it as large as possible, because you also need to consider the numerous downstream inference use cases; what you really want is the optimal solution of intelligence multiplied by cost.' However, he firmly opposed the notion that 'pre-training is no longer important,' believing that the smarter the foundational model, the higher the efficiency of subsequent reinforcement learning and inference stages, and that there is still an 'absolute need' for Nvidia GPUs to support large-scale centralized training.

[OpenAI Co-Founder Claims $110 Billion Still Can't Meet Demand, Pre-Training Shifts to Cost Joint Optimization]According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman reflected on the leap in AI programming capabilities expected by December 2025 during an interview. He used a test prompt he had kept for years to measure progress: asking the AI to build a website that took him months to complete when he was learning programming. Throughout 2025, this task required multiple prompts and about four hours to accomplish; by December, it could be completed with a single prompt and with high quality. He stated that the new model allowed AI to jump from 'being able to complete about 20% of tasks' to 'about 80%', a shift that forces everyone to 'reorganize workflows around AI.' Regarding the allocation of the $110 billion in funding, Brockman likened computing power to 'hiring salespeople': as long as the product has a scalable sales channel, hiring more salespeople can generate more revenue. Computing power is not a cost center but a revenue center. He recalled a conversation with his team on the eve of ChatGPT's release: 'They asked, 'How much computing power should we buy?' I said, 'All of it.' They replied, 'No, no, no, seriously, how much should we buy?' I said, 'No matter how we build, we won't keep up with demand.' This judgment still holds true today, and computing power procurement needs to be locked in 18 to 24 months in advance. On how to utilize this computing power, Brockman revealed that OpenAI is no longer solely pursuing the largest scale of pre-training but is instead jointly optimizing pre-training capabilities and inference costs: 'You don't necessarily have to make it as large as possible, because you also need to consider the numerous downstream inference use cases; what you really want is the optimal solution of intelligence multiplied by cost.' However, he firmly opposed the notion that 'pre-training is no longer important,' believing that the smarter the foundational model, the higher the efficiency of subsequent reinforcement learning and inference stages, and that there is still an 'absolute need' for Nvidia GPUs to support large-scale centralized training.

OpenAI Co-Founder Claims $110 Billion Still Can't Meet Demand, Pre-Training Shifts to Cost Joint Optimization

According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman discussed the new generation pre-trained foundational model codenamed Spud from a technical perspective for the first time in an interview. He described Spud as 'the culmination of about two years of research,' serving as the foundation for all subsequent capability enhancements at OpenAI. He explained the model production process: first, a foundational model is produced through pre-training, followed by reinforcement learning to practice problem-solving in various scenarios, and finally, post-training fine-tuning in terms of behavior and usability. His primary focus over the past 18 months has been on expanding GPU infrastructure and training frameworks. Regarding Spud's capabilities, Brockman provided qualitative rather than quantitative expectations: the model will be able to solve more difficult problems, understand instructions more accurately, grasp context more deeply, and reduce the 'big model smell.' He cited an engineer he closely collaborated with as an example: between GPT-5.2 and GPT-5.3, this engineer transitioned from 'completely unable to use AI for foundational systems engineering' to 'giving it a design document, and it can implement functionality, add monitoring metrics and observability, run performance analyzers and optimize, producing exactly what he wanted.' The interviewer referenced a previous statement by Anthropic CEO Dario Amodei: 'Some players are gambling (yoloing), pulling the risks too high, and I am very concerned,' mentioning that Amodei stated that if predictions are slightly off, the company could go bankrupt. Brockman directly responded: 'I disagree. We have always been very thoughtful and have a clear view of the trends that are coming.' He believes that OpenAI was one of the first companies to recognize the computational power bottleneck, while others 'probably only realized this at the end of last year and started scrambling for computing power, but it was already too late to buy.' Regarding bankruptcy risks, he stated, 'In reality, there are more exit ramps,' and attributed this gamble to a judgment about the entire industry rather than a single company: 'Do you believe this technology can produce and deliver the kind of immense value we are seeing?'

[OpenAI Co-Founder Discusses Spud Model, Responds to Dario's 'High-Stakes' Criticism]According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman discussed the new generation pre-trained foundational model codenamed Spud from a technical perspective for the first time in an interview. He described Spud as 'the culmination of about two years of research,' serving as the foundation for all subsequent capability enhancements at OpenAI. He explained the model production process: first, a foundational model is produced through pre-training, followed by reinforcement learning to practice problem-solving in various scenarios, and finally, post-training fine-tuning in terms of behavior and usability. His primary focus over the past 18 months has been on expanding GPU infrastructure and training frameworks. Regarding Spud's capabilities, Brockman provided qualitative rather than quantitative expectations: the model will be able to solve more difficult problems, understand instructions more accurately, grasp context more deeply, and reduce the 'big model smell.' He cited an engineer he closely collaborated with as an example: between GPT-5.2 and GPT-5.3, this engineer transitioned from 'completely unable to use AI for foundational systems engineering' to 'giving it a design document, and it can implement functionality, add monitoring metrics and observability, run performance analyzers and optimize, producing exactly what he wanted.' The interviewer referenced a previous statement by Anthropic CEO Dario Amodei: 'Some players are gambling (yoloing), pulling the risks too high, and I am very concerned,' mentioning that Amodei stated that if predictions are slightly off, the company could go bankrupt. Brockman directly responded: 'I disagree. We have always been very thoughtful and have a clear view of the trends that are coming.' He believes that OpenAI was one of the first companies to recognize the computational power bottleneck, while others 'probably only realized this at the end of last year and started scrambling for computing power, but it was already too late to buy.' Regarding bankruptcy risks, he stated, 'In reality, there are more exit ramps,' and attributed this gamble to a judgment about the entire industry rather than a single company: 'Do you believe this technology can produce and deliver the kind of immense value we are seeing?'

OpenAI Co-Founder Discusses Spud Model, Responds to Dario's 'High-Stakes' Criticism

On April 2, Cryptopolitan reported that the Russian government submitted a bill to the State Duma proposing criminal penalties for illegal cryptocurrency mining activities, with violators facing fines of up to 2 million rubles (approximately $25,000) and up to 5 years of imprisonment. Currently, about 50,000 individuals and entities are engaged in mining activities, but fewer than 1,500 are registered. Meanwhile, Russia has expanded the mining ban to 13 regions, including the Buryat Republic in Siberia and the Transbaikal Territory, with the ban set to last until 2031. The Energy Minister of Moscow Oblast has also proposed a ban on mining in the capital and Moscow Oblast, stating that local mining consumes about 1 GW of electricity and provides no substantial contribution to the regional economy. Russia had previously legalized mining by the end of 2024, but required practitioners to register with the Federal Tax Service and pay taxes in accordance with the law.

[Russia Plans Legislation to Combat Illegal Mining, 13 Regions Have Fully Banned Mining Activities]On April 2, Cryptopolitan reported that the Russian government submitted a bill to the State Duma proposing criminal penalties for illegal cryptocurrency mining activities, with violators facing fines of up to 2 million rubles (approximately $25,000) and up to 5 years of imprisonment. Currently, about 50,000 individuals and entities are engaged in mining activities, but fewer than 1,500 are registered. Meanwhile, Russia has expanded the mining ban to 13 regions, including the Buryat Republic in Siberia and the Transbaikal Territory, with the ban set to last until 2031. The Energy Minister of Moscow Oblast has also proposed a ban on mining in the capital and Moscow Oblast, stating that local mining consumes about 1 GW of electricity and provides no substantial contribution to the regional economy. Russia had previously legalized mining by the end of 2024, but required practitioners to register with the Federal Tax Service and pay taxes in accordance with the law.

Russia Plans Legislation to Combat Illegal Mining, 13 Regions Have Fully Banned Mining Activities

According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman revealed in the same interview that the research plan behind Sora has not ended with the product shutdown, but has been integrated into the robotics direction for continued advancement. He stated that Sora's world model and the core GPT reasoning series belong to 'two different branches of the technology tree.' Advancing both branches simultaneously under limited computational power 'is very difficult for us,' which is why the consumer application of Sora was cut, but the research team's work has shifted to 'a very clear application that will bring about transformation: robotics.' Brockman explained why the robotics direction remains in the research phase rather than a product priority: 'Robotics has not yet matured to the point of being deployable at scale; we will see a real takeoff in the knowledge work domain next year, not in robotics.' This aligns with OpenAI's logic of concentrating all productization resources on the GPT reasoning series. He also emphasized that the GPT series is not limited to text; capabilities such as voice interaction and image generation are based on different tuning of the same model architecture and do not constitute independent branches. He responded to Google DeepMind head Demis Hassabis's earlier claim that image/video generation is closest to AGI, admitting 'it's absolutely possible, yes.' OpenAI's choice to bet on the GPT reasoning route means 'you have to make choices and take action,' citing an early OpenAI principle: 'The sum of random vectors is zero, but if you align the vectors, you can move in one direction.'

[Shutting Down Products Does Not Mean Abandoning Research: OpenAI Co-Founder Says Sora's World Model Technology Has Shifted to Robotics]According to monitoring by 1M AI News, OpenAI co-founder Greg Brockman revealed in the same interview that the research plan behind Sora has not ended with the product shutdown, but has been integrated into the robotics direction for continued advancement. He stated that Sora's world model and the core GPT reasoning series belong to 'two different branches of the technology tree.' Advancing both branches simultaneously under limited computational power 'is very difficult for us,' which is why the consumer application of Sora was cut, but the research team's work has shifted to 'a very clear application that will bring about transformation: robotics.' Brockman explained why the robotics direction remains in the research phase rather than a product priority: 'Robotics has not yet matured to the point of being deployable at scale; we will see a real takeoff in the knowledge work domain next year, not in robotics.' This aligns with OpenAI's logic of concentrating all productization resources on the GPT reasoning series. He also emphasized that the GPT series is not limited to text; capabilities such as voice interaction and image generation are based on different tuning of the same model architecture and do not constitute independent branches. He responded to Google DeepMind head Demis Hassabis's earlier claim that image/video generation is closest to AGI, admitting 'it's absolutely possible, yes.' OpenAI's choice to bet on the GPT reasoning route means 'you have to make choices and take action,' citing an early OpenAI principle: 'The sum of random vectors is zero, but if you align the vectors, you can move in one direction.'

Shutting Down Products Does Not Mean Abandoning Research: OpenAI Co-Founder Says Sora's World Model Technology Has Shifted to Robotics

According to 1M AI News, OpenAI President and Co-Founder Greg Brockman provided the first detailed description of the super app's specific form and implementation path during an interview on the Big Technology Podcast. This app will integrate ChatGPT, the AI programming platform Codex, and the AI browser Atlas into a unified entry point, aiming to allow users to "directly request anything they want the computer to do." Brockman positions it as a "personal AGI," a personal assistant that understands user preferences, aligns with user goals, and can represent users in handling digital world affairs. He revealed that the first step is to expand the existing Codex application to support general knowledge work, not just programmers. He described this effort as building on Codex's existing general agent foundation, integrating tools like spreadsheets and documents: "The Codex application is essentially a combination of two things: a general agent foundation plus a coding agent. That general foundation can be applied in many areas." There have already been numerous spontaneous cases within OpenAI where Codex has been used for non-programming tasks, such as the communications team using it to integrate feedback via Slack and email. The entire super app plan will be rolled out in phases over the next few months, rather than being released all at once, with the standalone ChatGPT mobile app remaining unchanged; the merger will only apply to the desktop version. This strategy also implies a resource competition. Brockman admitted that OpenAI's current computing power is "not enough to support both the personal assistant and Codex lines simultaneously," which is the fundamental reason for shutting down Sora and scaling back the product line. He described this as a "recognition of technological maturity and the impending significant impact," rather than a simple shift from consumer to enterprise.

[Codex Leads the Way: OpenAI Co-Founder Details Super App Roadmap]According to 1M AI News, OpenAI President and Co-Founder Greg Brockman provided the first detailed description of the super app's specific form and implementation path during an interview on the Big Technology Podcast. This app will integrate ChatGPT, the AI programming platform Codex, and the AI browser Atlas into a unified entry point, aiming to allow users to "directly request anything they want the computer to do." Brockman positions it as a "personal AGI," a personal assistant that understands user preferences, aligns with user goals, and can represent users in handling digital world affairs. He revealed that the first step is to expand the existing Codex application to support general knowledge work, not just programmers. He described this effort as building on Codex's existing general agent foundation, integrating tools like spreadsheets and documents: "The Codex application is essentially a combination of two things: a general agent foundation plus a coding agent. That general foundation can be applied in many areas." There have already been numerous spontaneous cases within OpenAI where Codex has been used for non-programming tasks, such as the communications team using it to integrate feedback via Slack and email. The entire super app plan will be rolled out in phases over the next few months, rather than being released all at once, with the standalone ChatGPT mobile app remaining unchanged; the merger will only apply to the desktop version. This strategy also implies a resource competition. Brockman admitted that OpenAI's current computing power is "not enough to support both the personal assistant and Codex lines simultaneously," which is the fundamental reason for shutting down Sora and scaling back the product line. He described this as a "recognition of technological maturity and the impending significant impact," rather than a simple shift from consumer to enterprise.

Codex Leads the Way: OpenAI Co-Founder Details Super App Roadmap

火币HTX数据监测显示，BTC突破85000  USDT，现报85015.73 USDT，24H涨幅0.69%。

【BTC突破85000  USDT】火币HTX数据监测显示，BTC突破85000  USDT，现报85015.73 USDT，24H涨幅0.69%。

BTC突破85000  USDT

Article

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

BTC broke through 69000  USDT

U.S. Officials: Trump to Reiterate Timeline for Withdrawal from Iran in 2 to 3 Weeks

Riot Platforms Suspected of Selling 500 Bitcoins Worth $34.13 Million

SEC Approves NYSE American Listing of Multi-Crypto Asset Trust Options

BTC fell below 68000  USDT

BTC突破85000 USDT

Disclaimer：Il contenuto di cui sopra non rappresenta le posizioni di HTX.，HTX non fornisce alcuna raccomandazione di trading.。

Articoli correlati

Tutti i commenti0RecentePopolare

Temi di tendenza

Articoli correlati