×
Coreweave CEO defends AI circular deals as ‘working together’ | TechCrunch

Coreweave CEO defends AI circular deals as ‘working together’ | TechCrunch

It’s been quite the year for Coreweave. In March, the AI cloud infrastructure provider went public in one of the biggest and most anticipated IPOs of the year that didn’t live up to its hype.

Another setback took place in October, when a planned acquisition of the cloud provider’s business partner, Core Scientific, faltered due to skepticism from the acquisition target’s shareholders. 

In the meantime, the firm has acquired a number of different companies, its stock has gone up and down, and it’s been both criticized and lauded for its role in the booming AI data center market. 

In an interview at Fortune’s AI Brainstorm summit in San Francisco on Tuesday, Coreweave’s co-founder and CEO, Michael Intrator, defended his company’s performance from critics, noting that it was in the midst of creating a “new business model” for how cloud computing can be built and run. Their collection of Nvidia GPUs is so valuable, they borrow against it to help finance their business. The executive seemed to imply: If you’re charting a new path, you’re destined to encounter some road bumps along the way.  

“I think people are myopic a lot of times,” Intrator said, when questioned about his company’s occasionally unstable stock price. “Yes, it is see-sawing,” he admitted, while noting that the Coreweave IPO took place not long before President Trump’s tariffs went into effect—a notably uncertain moment for the overall economy. 

“We came out into one of the most challenging environments, right around Liberation Day and, in spite of the incredible headwinds, were able to launch a successful IPO,” the CEO told Brainstorm editorial director Andrew Nusca. “I couldn’t be prouder of what the company has accomplished,” he added. 

Coreweave’s stock may have debuted amidst the economic doldrums of March but its price has gone on quite the journey since then. It debuted at $40 and, over the past eight months, has climbed to well over $150, but currently rests at around $90. Its more wary critics have compared it to a meme stock due to its penchant for going up and down. 

Techcrunch event

San Francisco
|
October 13-15, 2026

Some of the uncertainty around Coreweave’s stock has been credited to the company’s hefty level of debt. Not long after Coreweave announced a deal on Monday to issue even more debt to finance its data center buildout, its stock dropped some 8 percent.

Intrator seems to see his company as a disruptor, one whose unconventional tactics may take some getting used to. “When you introduce a new model, when you introduce a new way of doing business, when you disrupt what has been a static environment, it’s going to take some people some time,” he said, during his appearance Tuesday. 

Coreweave actually started its corporate life as a crypto-miner but, in short order, built itself into a pivotal provider of “AI infrastructure” to some of the tech industry’s most major players. In that role, it provides GPUs to AI developers, and has made major partnerships with Microsoft, OpenAI, Nvidia, Meta, and other tech titans.  

Another topic broached Tuesday was the notion of “circularity” within the AI industry. “Circular” business deals, in which a small number of powerful AI companies invest in one another, have frequently been criticized, and have raised questions about the industry’s long term economic stability. Perhaps not surprisingly, since Nvidia is one of its investors as well as its supplier of GPUs, Intrator swatted away such concerns. “Companies are trying to address a violent change in supply and demand,” he said. “You do that by working together.”
 
Since the IPO, Coreweave has continued to make efforts to expand its business. After it acquired Weights and Balances, an AI developer platform, in March, it went on to acquire OpenPipe, a startup that helps companies create and deploy AI agents through reinforcement learning. In October, it also made deals to acquire Marimo (the creator of an open source notebook) and Monolith, another AI company. It also recently announced an expansion of its cloud partnership with OpenAI and said it has plans to move into the federal market, where it wants to provide cloud infrastructure to U.S. government agencies and the defense industrial base. 

Source link
#Coreweave #CEO #defends #circular #deals #working #TechCrunch


Yesterday, OpenAI released its latest voice model, GPT-Live-1. The company called it “a new generation of voice models for natural human-AI interaction.” The full-duplex model allows ChatGPT to talk and listen at the same time, giving the back-and-forth a more conversational flow. It’s designed to be smarter, faster, and more natural.

Sounds great! Let’s see how it handles the internet’s one-man voice model red team, TikToker @huskistaken (aka Husk), when he tests out its capabilities by giving the model something simple:

Ooh, so close. Husk gave ChatGPT the straightforward request to tell him how many times the letter E appears in the number “seventeen,” and the voice model managed to get halfway there by answering “two.” It then offered a very awkward sign-off when Husk ended the conversation. Maybe that’s a query that GPT-Live-2 will be able to handle.

To be fair, Husk isn’t really testing any of the new features that OpenAI highlighted with the release of its new model, which, by most accounts, seems like it’s more capable when it comes to things like live translation than previous iterations. But he is giving the model a very simple benchmarking test that, despite being OpenAI’s flagship voice model, it still failed spectacularly.

It seems OpenAI knows just how bad this looks, too. Jason Liu, a Developer Experience Engineer on OpenAI Codex, reposted the video with a very succinct, “FUCK.”

Husk’s made a habit of terrorizing OpenAI over its voice model for some time now, making viral videos out of how poorly it handles certain prompts. One of his videos—in which he tasks ChatGPT with setting a timer and watching it fail—made it all the way to CEO Sam Altman, who tried to laugh it off in a very “I’m not mad, please don’t put in the newspaper that I got mad” kinda way.

Husk is not alone in giving GPT-Live-1 the old stress test. Another user on X gave the model the “Strawberry” test, asking it to count the number of times the letter R appears in the word “strawberry.” It’s a classic, and one that most AI models can answer correctly by now (whether that’s because they really know there are 3 Rs or because they’ve been trained on how to respond to prevent embarrassment is another question), but GPT-Live-1 gets tripped up by it.

Another user noted a new annoyance that has arisen from the model’s full-duplex functionality: While the model now allows ChatGPT to say things like “mhm” and “yeah” while you talk to confirm it is listening, it apparently is incapable of simply shutting up and not doing that. It seems ChatGPT has built an interrupting machine. Just what we’ve all been waiting for.

READ MORE:

The Future Is Always Listening: OpenAI Says Its New Voice Assistant Is ‘One Step Closer to a Truly Accessible AGI’

OpenAI Is Tired of Seeing All Those Videos of People Clowning on Its Voice Mode

#OpenAI #Beat #TikTokerartifical intelligence,ChatGPT,OpenAI,voice model">OpenAI Just Can’t Beat This TikToker
                Yesterday, OpenAI released its latest voice model, GPT-Live-1. The company called it “a new generation of voice models for natural human-AI interaction.” The full-duplex model allows ChatGPT to talk and listen at the same time, giving the back-and-forth a more conversational flow. It’s designed to be smarter, faster, and more natural. Sounds great! Let’s see how it handles the internet’s one-man voice model red team, TikToker @huskistaken (aka Husk), when he tests out its capabilities by giving the model something simple:  Had to give the new voice model the classic test pic.twitter.com/fQYHbBRNuL — Husk (@huskirl) July 8, 2026  Ooh, so close. Husk gave ChatGPT the straightforward request to tell him how many times the letter E appears in the number “seventeen,” and the voice model managed to get halfway there by answering “two.” It then offered a very awkward sign-off when Husk ended the conversation. Maybe that’s a query that GPT-Live-2 will be able to handle.

 To be fair, Husk isn’t really testing any of the new features that OpenAI highlighted with the release of its new model, which, by most accounts, seems like it’s more capable when it comes to things like live translation than previous iterations. But he is giving the model a very simple benchmarking test that, despite being OpenAI’s flagship voice model, it still failed spectacularly.

 It seems OpenAI knows just how bad this looks, too. Jason Liu, a Developer Experience Engineer on OpenAI Codex, reposted the video with a very succinct, “FUCK.” Husk’s made a habit of terrorizing OpenAI over its voice model for some time now, making viral videos out of how poorly it handles certain prompts. One of his videos—in which he tasks ChatGPT with setting a timer and watching it fail—made it all the way to CEO Sam Altman, who tried to laugh it off in a very “I’m not mad, please don’t put in the newspaper that I got mad” kinda way.

 Husk is not alone in giving GPT-Live-1 the old stress test. Another user on X gave the model the “Strawberry” test, asking it to count the number of times the letter R appears in the word “strawberry.” It’s a classic, and one that most AI models can answer correctly by now (whether that’s because they really know there are 3 Rs or because they’ve been trained on how to respond to prevent embarrassment is another question), but GPT-Live-1 gets tripped up by it.  OpenAI’s brand new voice model vs counting the r’s in strawberry. I really tried to help it along. pic.twitter.com/Pm0RfLyxIA — Himelstech (@himelstech) July 9, 2026  Another user noted a new annoyance that has arisen from the model’s full-duplex functionality: While the model now allows ChatGPT to say things like “mhm” and “yeah” while you talk to confirm it is listening, it apparently is incapable of simply shutting up and not doing that. It seems ChatGPT has built an interrupting machine. Just what we’ve all been waiting for. READ MORE: The Future Is Always Listening: OpenAI Says Its New Voice Assistant Is ‘One Step Closer to a Truly Accessible AGI’ OpenAI Is Tired of Seeing All Those Videos of People Clowning on Its Voice Mode      #OpenAI #Beat #TikTokerartifical intelligence,ChatGPT,OpenAI,voice model

called it “a new generation of voice models for natural human-AI interaction.” The full-duplex model allows ChatGPT to talk and listen at the same time, giving the back-and-forth a more conversational flow. It’s designed to be smarter, faster, and more natural.

Sounds great! Let’s see how it handles the internet’s one-man voice model red team, TikToker @huskistaken (aka Husk), when he tests out its capabilities by giving the model something simple:

Ooh, so close. Husk gave ChatGPT the straightforward request to tell him how many times the letter E appears in the number “seventeen,” and the voice model managed to get halfway there by answering “two.” It then offered a very awkward sign-off when Husk ended the conversation. Maybe that’s a query that GPT-Live-2 will be able to handle.

To be fair, Husk isn’t really testing any of the new features that OpenAI highlighted with the release of its new model, which, by most accounts, seems like it’s more capable when it comes to things like live translation than previous iterations. But he is giving the model a very simple benchmarking test that, despite being OpenAI’s flagship voice model, it still failed spectacularly.

It seems OpenAI knows just how bad this looks, too. Jason Liu, a Developer Experience Engineer on OpenAI Codex, reposted the video with a very succinct, “FUCK.”

Husk’s made a habit of terrorizing OpenAI over its voice model for some time now, making viral videos out of how poorly it handles certain prompts. One of his videos—in which he tasks ChatGPT with setting a timer and watching it fail—made it all the way to CEO Sam Altman, who tried to laugh it off in a very “I’m not mad, please don’t put in the newspaper that I got mad” kinda way.

Husk is not alone in giving GPT-Live-1 the old stress test. Another user on X gave the model the “Strawberry” test, asking it to count the number of times the letter R appears in the word “strawberry.” It’s a classic, and one that most AI models can answer correctly by now (whether that’s because they really know there are 3 Rs or because they’ve been trained on how to respond to prevent embarrassment is another question), but GPT-Live-1 gets tripped up by it.

Another user noted a new annoyance that has arisen from the model’s full-duplex functionality: While the model now allows ChatGPT to say things like “mhm” and “yeah” while you talk to confirm it is listening, it apparently is incapable of simply shutting up and not doing that. It seems ChatGPT has built an interrupting machine. Just what we’ve all been waiting for.

READ MORE:

The Future Is Always Listening: OpenAI Says Its New Voice Assistant Is ‘One Step Closer to a Truly Accessible AGI’

OpenAI Is Tired of Seeing All Those Videos of People Clowning on Its Voice Mode

#OpenAI #Beat #TikTokerartifical intelligence,ChatGPT,OpenAI,voice model">OpenAI Just Can’t Beat This TikTokerOpenAI Just Can’t Beat This TikToker
                Yesterday, OpenAI released its latest voice model, GPT-Live-1. The company called it “a new generation of voice models for natural human-AI interaction.” The full-duplex model allows ChatGPT to talk and listen at the same time, giving the back-and-forth a more conversational flow. It’s designed to be smarter, faster, and more natural. Sounds great! Let’s see how it handles the internet’s one-man voice model red team, TikToker @huskistaken (aka Husk), when he tests out its capabilities by giving the model something simple:  Had to give the new voice model the classic test pic.twitter.com/fQYHbBRNuL — Husk (@huskirl) July 8, 2026  Ooh, so close. Husk gave ChatGPT the straightforward request to tell him how many times the letter E appears in the number “seventeen,” and the voice model managed to get halfway there by answering “two.” It then offered a very awkward sign-off when Husk ended the conversation. Maybe that’s a query that GPT-Live-2 will be able to handle.

 To be fair, Husk isn’t really testing any of the new features that OpenAI highlighted with the release of its new model, which, by most accounts, seems like it’s more capable when it comes to things like live translation than previous iterations. But he is giving the model a very simple benchmarking test that, despite being OpenAI’s flagship voice model, it still failed spectacularly.

 It seems OpenAI knows just how bad this looks, too. Jason Liu, a Developer Experience Engineer on OpenAI Codex, reposted the video with a very succinct, “FUCK.” Husk’s made a habit of terrorizing OpenAI over its voice model for some time now, making viral videos out of how poorly it handles certain prompts. One of his videos—in which he tasks ChatGPT with setting a timer and watching it fail—made it all the way to CEO Sam Altman, who tried to laugh it off in a very “I’m not mad, please don’t put in the newspaper that I got mad” kinda way.

 Husk is not alone in giving GPT-Live-1 the old stress test. Another user on X gave the model the “Strawberry” test, asking it to count the number of times the letter R appears in the word “strawberry.” It’s a classic, and one that most AI models can answer correctly by now (whether that’s because they really know there are 3 Rs or because they’ve been trained on how to respond to prevent embarrassment is another question), but GPT-Live-1 gets tripped up by it.  OpenAI’s brand new voice model vs counting the r’s in strawberry. I really tried to help it along. pic.twitter.com/Pm0RfLyxIA — Himelstech (@himelstech) July 9, 2026  Another user noted a new annoyance that has arisen from the model’s full-duplex functionality: While the model now allows ChatGPT to say things like “mhm” and “yeah” while you talk to confirm it is listening, it apparently is incapable of simply shutting up and not doing that. It seems ChatGPT has built an interrupting machine. Just what we’ve all been waiting for. READ MORE: The Future Is Always Listening: OpenAI Says Its New Voice Assistant Is ‘One Step Closer to a Truly Accessible AGI’ OpenAI Is Tired of Seeing All Those Videos of People Clowning on Its Voice Mode      #OpenAI #Beat #TikTokerartifical intelligence,ChatGPT,OpenAI,voice model

Yesterday, OpenAI released its latest voice model, GPT-Live-1. The company called it “a new generation of voice models for natural human-AI interaction.” The full-duplex model allows ChatGPT to talk and listen at the same time, giving the back-and-forth a more conversational flow. It’s designed to be smarter, faster, and more natural.

Sounds great! Let’s see how it handles the internet’s one-man voice model red team, TikToker @huskistaken (aka Husk), when he tests out its capabilities by giving the model something simple:

Ooh, so close. Husk gave ChatGPT the straightforward request to tell him how many times the letter E appears in the number “seventeen,” and the voice model managed to get halfway there by answering “two.” It then offered a very awkward sign-off when Husk ended the conversation. Maybe that’s a query that GPT-Live-2 will be able to handle.

To be fair, Husk isn’t really testing any of the new features that OpenAI highlighted with the release of its new model, which, by most accounts, seems like it’s more capable when it comes to things like live translation than previous iterations. But he is giving the model a very simple benchmarking test that, despite being OpenAI’s flagship voice model, it still failed spectacularly.

It seems OpenAI knows just how bad this looks, too. Jason Liu, a Developer Experience Engineer on OpenAI Codex, reposted the video with a very succinct, “FUCK.”

Husk’s made a habit of terrorizing OpenAI over its voice model for some time now, making viral videos out of how poorly it handles certain prompts. One of his videos—in which he tasks ChatGPT with setting a timer and watching it fail—made it all the way to CEO Sam Altman, who tried to laugh it off in a very “I’m not mad, please don’t put in the newspaper that I got mad” kinda way.

Husk is not alone in giving GPT-Live-1 the old stress test. Another user on X gave the model the “Strawberry” test, asking it to count the number of times the letter R appears in the word “strawberry.” It’s a classic, and one that most AI models can answer correctly by now (whether that’s because they really know there are 3 Rs or because they’ve been trained on how to respond to prevent embarrassment is another question), but GPT-Live-1 gets tripped up by it.

Another user noted a new annoyance that has arisen from the model’s full-duplex functionality: While the model now allows ChatGPT to say things like “mhm” and “yeah” while you talk to confirm it is listening, it apparently is incapable of simply shutting up and not doing that. It seems ChatGPT has built an interrupting machine. Just what we’ve all been waiting for.

READ MORE:

The Future Is Always Listening: OpenAI Says Its New Voice Assistant Is ‘One Step Closer to a Truly Accessible AGI’

OpenAI Is Tired of Seeing All Those Videos of People Clowning on Its Voice Mode

#OpenAI #Beat #TikTokerartifical intelligence,ChatGPT,OpenAI,voice model

Friendly home robots have been the dream of sci-fi for a long time. Docile, helpful machines that do your laundry, take out the trash, maybe make dinner, and clean up afterward too. But if robots are going to do all that, they’re going to need some hands to make all that happen.

1X, a Norwegian-American robotics company, today revealed details about the five-finger hands attached to its soft, helpful robot companion, Neo.

The hands are built with actuators designed to replicate how tendons in the arms move human hands. 1X says this gives Neo’s hands 25 degrees of freedom of movement, which is just a little less than the 27 degrees of freedom human hands usually have. Cameras and AI smarts help sort out the broader context of what the fingers are trying to grab. It’s a dexterous mix that gives the Neo bot a very broad range of motion. 1X says the hands can grip odd shapes and detect when something is slipping out of its grip. The fingers can also move extremely quickly and hyperextend in directions human digits can’t. They also have an IP68 waterproof rating, meaning the robot can wash its own hands.

Courtesy of 1X

It’s a range that the engineers have aimed to dial in to align with what a human can actually do. Jonathan Terfurth, 1X’s director of actuators and hands, says the range of motion might even be better than a human’s, enabling it to open doors, lift heavy objects, and even plug itself in when its battery starts to die.

“You want to be able to operate with a human who has never worked or interfaced with a robot, and you still want it to be safe and compliant and soft,” Terfurth says. “Range of motion can be a bit extreme, but we try to be very close to what humans can do so that we can live in the world.”

Together, this is part of a growing ChatGPT-esque moment for robotics, where bots have gone from clumsy claw machines to capable handlers, able to gently carry fragile objects and become increasingly helpful for managing menial daily tasks and boring office chores.

The humanoid robot market tends to be dominated by tough, hulking Terminators meant to be paraded out for some defense industry contract or another. 1X is taking a different tack for its robot. The Neo is a soft, supple robot wrapped in a 3D lattice shell. Its design is inspired by characters like Baymax, the affable robot from the Disney movie Big Hero 6. In limited quantities, early access pricing is $20,000 or $500 per month to have it in your home, though the lump sum will prioritize delivery for 2026.

#Neo #Robot #Freaky #Fast #Fingersrobots,robotics,artificial intelligence">The 1X Neo Robot Has Freaky Fast FingersFriendly home robots have been the dream of sci-fi for a long time. Docile, helpful machines that do your laundry, take out the trash, maybe make dinner, and clean up afterward too. But if robots are going to do all that, they’re going to need some hands to make all that happen.1X, a Norwegian-American robotics company, today revealed details about the five-finger hands attached to its soft, helpful robot companion, Neo.The hands are built with actuators designed to replicate how tendons in the arms move human hands. 1X says this gives Neo’s hands 25 degrees of freedom of movement, which is just a little less than the 27 degrees of freedom human hands usually have. Cameras and AI smarts help sort out the broader context of what the fingers are trying to grab. It’s a dexterous mix that gives the Neo bot a very broad range of motion. 1X says the hands can grip odd shapes and detect when something is slipping out of its grip. The fingers can also move extremely quickly and hyperextend in directions human digits can’t. They also have an IP68 waterproof rating, meaning the robot can wash its own hands.Courtesy of 1XIt’s a range that the engineers have aimed to dial in to align with what a human can actually do. Jonathan Terfurth, 1X’s director of actuators and hands, says the range of motion might even be better than a human’s, enabling it to open doors, lift heavy objects, and even plug itself in when its battery starts to die.“You want to be able to operate with a human who has never worked or interfaced with a robot, and you still want it to be safe and compliant and soft,” Terfurth says. “Range of motion can be a bit extreme, but we try to be very close to what humans can do so that we can live in the world.”Together, this is part of a growing ChatGPT-esque moment for robotics, where bots have gone from clumsy claw machines to capable handlers, able to gently carry fragile objects and become increasingly helpful for managing menial daily tasks and boring office chores.The humanoid robot market tends to be dominated by tough, hulking Terminators meant to be paraded out for some defense industry contract or another. 1X is taking a different tack for its robot. The Neo is a soft, supple robot wrapped in a 3D lattice shell. Its design is inspired by characters like Baymax, the affable robot from the Disney movie Big Hero 6. In limited quantities, early access pricing is ,000 or 0 per month to have it in your home, though the lump sum will prioritize delivery for 2026.#Neo #Robot #Freaky #Fast #Fingersrobots,robotics,artificial intelligence

ChatGPT-esque moment for robotics, where bots have gone from clumsy claw machines to capable handlers, able to gently carry fragile objects and become increasingly helpful for managing menial daily tasks and boring office chores.

The humanoid robot market tends to be dominated by tough, hulking Terminators meant to be paraded out for some defense industry contract or another. 1X is taking a different tack for its robot. The Neo is a soft, supple robot wrapped in a 3D lattice shell. Its design is inspired by characters like Baymax, the affable robot from the Disney movie Big Hero 6. In limited quantities, early access pricing is $20,000 or $500 per month to have it in your home, though the lump sum will prioritize delivery for 2026.

#Neo #Robot #Freaky #Fast #Fingersrobots,robotics,artificial intelligence">The 1X Neo Robot Has Freaky Fast Fingers

Friendly home robots have been the dream of sci-fi for a long time. Docile, helpful machines that do your laundry, take out the trash, maybe make dinner, and clean up afterward too. But if robots are going to do all that, they’re going to need some hands to make all that happen.

1X, a Norwegian-American robotics company, today revealed details about the five-finger hands attached to its soft, helpful robot companion, Neo.

The hands are built with actuators designed to replicate how tendons in the arms move human hands. 1X says this gives Neo’s hands 25 degrees of freedom of movement, which is just a little less than the 27 degrees of freedom human hands usually have. Cameras and AI smarts help sort out the broader context of what the fingers are trying to grab. It’s a dexterous mix that gives the Neo bot a very broad range of motion. 1X says the hands can grip odd shapes and detect when something is slipping out of its grip. The fingers can also move extremely quickly and hyperextend in directions human digits can’t. They also have an IP68 waterproof rating, meaning the robot can wash its own hands.

Courtesy of 1X

It’s a range that the engineers have aimed to dial in to align with what a human can actually do. Jonathan Terfurth, 1X’s director of actuators and hands, says the range of motion might even be better than a human’s, enabling it to open doors, lift heavy objects, and even plug itself in when its battery starts to die.

“You want to be able to operate with a human who has never worked or interfaced with a robot, and you still want it to be safe and compliant and soft,” Terfurth says. “Range of motion can be a bit extreme, but we try to be very close to what humans can do so that we can live in the world.”

Together, this is part of a growing ChatGPT-esque moment for robotics, where bots have gone from clumsy claw machines to capable handlers, able to gently carry fragile objects and become increasingly helpful for managing menial daily tasks and boring office chores.

The humanoid robot market tends to be dominated by tough, hulking Terminators meant to be paraded out for some defense industry contract or another. 1X is taking a different tack for its robot. The Neo is a soft, supple robot wrapped in a 3D lattice shell. Its design is inspired by characters like Baymax, the affable robot from the Disney movie Big Hero 6. In limited quantities, early access pricing is $20,000 or $500 per month to have it in your home, though the lump sum will prioritize delivery for 2026.

#Neo #Robot #Freaky #Fast #Fingersrobots,robotics,artificial intelligence

Post Comment