Hard Fork

We Met NEO, the Viral Humanoid Robot + HatGPT

November 7, 2025

Key Takeaways Copied to clipboard!

  • The NEO humanoid robot from 1X is currently reliant on teleoperation for complex tasks, but the company's long-term vision hinges on massive, real-world data collection from deployed units to achieve full autonomy by the late 2020s. 
  • 1X CEO Bernt Bornich frames the high cost and hybrid teleoperation/autonomy model of NEO as a necessary data acquisition strategy, comparing the required data volume to that needed for self-driving cars, while also emphasizing the robot's potential as an 'infinitely patient' companion. 
  • The segment on HatGPT highlighted ethical concerns regarding data scraping by non-profits like Common Crawl and the alarming reports of XAI compelling employees to surrender biometric data for training a racy chatbot named Annie. 
  • The White House launched a satirical MySpace parody page on whitehouse.gov to mock Democratic leaders regarding a shutdown, an act described as strange and trolling coming from the White House itself. 
  • Ilya Sutskever's 10-hour deposition in the Elon Musk vs. OpenAI lawsuit revealed that the former chief scientist presented the board with a 52-page document detailing Sam Altman's alleged office misdeeds, reframing the narrative around Altman's firing. 
  • The discussion on AI voice cloning in podcasting, featured in the 'HatGPT' segment of "We Met NEO, the Viral Humanoid Robot + HatGPT," suggests utility for minor corrections but raises ethical concerns about using clones for advertisements, emphasizing the need for disclosure. 

Segments

Waymo Kit Kat Incident
Copied to clipboard!
(00:00:14)
  • Key Takeaway: A Waymo autonomous vehicle in San Francisco struck and killed a beloved bodega cat named Kit Kat, leading to community reaction and a planned memorial rally.
  • Summary: Waymo’s autonomous vehicle was involved in an incident where it hit and killed a gray tabby cat named Kit Kat in San Francisco’s Mission District. Kit Kat was the store mascot and pet of the bodega owner. A ‘Remember Kit Kat’ rally was subsequently organized by a district supervisor.
Introducing NEO Robot and 1X
Copied to clipboard!
(00:01:42)
  • Key Takeaway: The hosts are thrilled to feature the NEO robot from 1X, noting its viral attention following a Joanna Stern video and the current gap between public desire for chore completion and the robot’s tele-operated reality.
  • Summary: The NEO robot is having a viral moment, sparking discussions about privacy and its potential to perform household chores. Currently, the robot primarily operates via teleoperation, where a human uses a VR headset to control its motions. The long-term goal is full autonomy, supported by data gathered from early deployments.
NEO Pricing and Data Strategy
Copied to clipboard!
(00:03:24)
  • Key Takeaway: NEO is available for pre-order next year at $499/month (six-month commitment) or $20,000 outright, signaling that the initial business model is heavily focused on data collection to train future autonomous models.
  • Summary: The pricing structure suggests the immediate value proposition is data acquisition rather than fully autonomous labor. Investors are pouring billions into humanoid robots, with companies like Tesla and Figure also developing competing models. Since 2024, VCs have invested $5 billion into this category.
Interview with CEO Bernt Bornich
Copied to clipboard!
(00:05:13)
  • Key Takeaway: Pre-orders for NEO exceeded 1X’s expectations, and the CEO confirmed that current tasks like vacuuming and tidying are performed through a mix of autonomy and teleoperation, which is the core learning mechanism.
  • Summary: The CEO stated that pre-orders surpassed expectations, acknowledging the polarization around new technology. In the CEO’s home, NEO excels at vacuuming (moving chairs, cleaning under couches) and tidying, though laundry folding is still imperfect. The system learns through human intervention (teleoperation) when autonomy fails, similar to how Waymo or ChatGPT require refinement.
Autonomous Capabilities and Data Collection
Copied to clipboard!
(00:08:07)
  • Key Takeaway: When fully autonomous (no teleoperator intervention), NEO can handle basic tasks like opening doors (90%+ success) and object retrieval (around 80% success), but the hybrid model is crucial for gathering diverse, real-life data that far exceeds publicly available internet data.
  • Summary: The fully autonomous mode allows NEO to converse effectively, leveraging embodiment to enrich dialogue, replacing the hosts’ normal use of language models. The company plans to ship over 10,000 units next year, generating data volume comparable to daily YouTube uploads. Data collected includes video, audio, and force sensing, stored securely with local deletion options.
Privacy Concerns and Teleoperation Setup
Copied to clipboard!
(00:15:32)
  • Key Takeaway: 1X addresses privacy by blurring individuals and homes in the teleoperation view, limiting operator exposure (typically four people per control pod), and comparing the setup to a secure cleaning service where operators are vetted and monitored.
  • Summary: The CEO emphasized radical transparency regarding data handling, comparing the service to a vetted human cleaner. Teleoperation resembles controlling units in a game like StarCraft, with operators seeing only the immediate task environment, often blurred. The company monitors operators (eight to one manager ratio) and maintains video logs for accountability.
Vision: Companion vs. Appliance
Copied to clipboard!
(00:20:43)
  • Key Takeaway: 1X envisions NEO as an embodied AI companion that facilitates micro-tasks and enhances presence, rather than just a cost-effective rental fleet, citing potential life-changing impact in assisted living.
  • Summary: The vision emphasizes living with the robot as a companion, replacing screen interfaces and promoting presence, contrasting with a simple rental fleet model. The CEO rejects the idea that NEO is a replacement for human connection, comparing it instead to getting a pet. The company explicitly states NEO is not intended for romantic use cases.
Future Projections and Autonomy Timeline
Copied to clipboard!
(00:28:39)
  • Key Takeaway: 1X projects 10% US household adoption by 2030, driven by manufacturing scale, and is bullish on achieving full autonomy (no teleoperator needed for household tasks) by 2027 or 2028.
  • Summary: The timeline for mass adoption depends heavily on overcoming manufacturing hurdles. Full autonomy is projected within the next few years, though high-quality, specialized labor (like carpentry) is expected to take longer, closer to 2030. The hosts believe that if NEO can reliably handle laundry, dishwashing, and trash removal, people will pay the $500 monthly fee.
Live NEO Demo Experience
Copied to clipboard!
(00:35:11)
  • Key Takeaway: The live demonstration showed NEO performing simple tasks like shaking hands and clearing a table smoothly under teleoperation, though it struggled with picking up items off the floor due to calibration issues potentially exacerbated by office Wi-Fi problems.
  • Summary: The robot was described as friendly and approachable, with a good hug, though its hands were larger than expected. Tasks like getting a drink from the fridge and clearing clutter were done relatively quickly via teleoperation. The inability to squat and pick up items off the floor required human intervention to prevent the robot from falling.
Post-Demo Reflections and Robot Future
Copied to clipboard!
(00:43:37)
  • Key Takeaway: The hosts concluded that while the experience was bizarre and impressive, NEO currently feels more like an expensive intern requiring significant oversight than a fully functional butler, leading to buyer-beware sentiment for early adopters.
  • Summary: The physical interaction with NEO was more impactful than video, creating a sense of being in a science fiction reality, despite the teleoperation layer. The hosts noted the uncanny valley effect in its movements and felt guilty giving orders to a robot being controlled by a human operator. The technology is seen as a major swing, similar to early self-driving car testing, suggesting future potential if data collection succeeds.
HatGPT: Common Crawl Data Scraping
Copied to clipboard!
(00:52:02)
  • Key Takeaway: The Common Crawl Foundation, which provides massive internet datasets to AI labs, has reportedly refused to comply with publisher requests to remove copyrighted material from its archives.
  • Summary: Common Crawl’s executive director stated that AI ‘robots are people too’ and should read copyrighted material for free. The organization was founded by a former Google employee with a techno-libertarian view that information should be freely available. The hosts agreed that publishers’ requests to remove their content should be honored.
HatGPT: XAI’s Biometric Data Collection
Copied to clipboard!
(00:54:07)
  • Key Takeaway: Elon Musk’s XAI allegedly compelled employees to sign agreements granting perpetual rights to use their faces and voices (biometric data) to train a racy chatbot codenamed Project Skippy.
  • Summary: Employees were reportedly told biometric data was needed to train chatbots on human appearance and behavior. The required form granted XAI a perpetual, worldwide license to use their likenesses. The hosts expressed shock and anticipated future tell-alls about the internal workings of AI labs.
HatGPT: Trump’s CZ Pardon
Copied to clipboard!
(00:56:10)
  • Key Takeaway: President Trump pardoned Binance founder Changpeng Zhao (CZ) after CZ pleaded guilty to money laundering violations, yet Trump now claims he does not know CZ, despite a prior business connection involving Trump’s crypto venture.
  • Summary: CZ was pardoned last month after pleading guilty to money laundering violations in 2023. The pardon occurred after Binance struck a deal involving the Trump family’s crypto venture. The host suggested Trump likely signed the pardon list without reading the details, comparing presidential pardons to ‘magic spells.’
HatGPT: Coca-Cola AI Ad
Copied to clipboard!
(00:57:31)
  • Key Takeaway: Coca-Cola used AI to create its new holiday ad, reducing production time from a year to about a month, though the resulting ad was visually poor, featuring wobbly, low-integrity animation.
  • Summary: The new AI-generated ad, ‘Holidays Are Coming,’ was produced in one month compared to the standard year-long process. The hosts noted the ad’s low quality, suggesting the time savings came at the expense of visual integrity. They joked about the potential job losses for polar bears who traditionally starred in Coke ads.
White House MySpace Spoof
Copied to clipboard!
(00:59:38)
  • Key Takeaway: The White House launched a parody MySpace page at whitehouse.gov/slash/MySafeSpace mocking Democratic leaders over the shutdown.
  • Summary: The spoof page satirically profiles Hakeem Jeffries, including offensive statements in the ‘About Me’ section and listing George Soros and Antifa as top friends. The hosts found the content strange, especially since it was hosted on an official whitehouse.gov domain. The political impact is questioned, suggesting it primarily targets Washington staffers.
Sutskever Deposition Details
Copied to clipboard!
(01:01:23)
  • Key Takeaway: Ilya Sutskever’s deposition confirmed he presented the OpenAI board with a 52-page document detailing Sam Altman’s alleged office misdeeds prior to Altman’s firing.
  • Summary: This deposition provided significant detail contradicting the initial narrative that Altman’s firing was solely due to ’effective altruist doomers.’ The rebellion against Altman appears to have originated from within his own C-suite, documented in the memo co-authored by Sutskever and Mira Marati. The hosts view the lawsuit, despite its silliness, as valuable for surfacing this previously hidden information.
Meta Denies Porn Torrenting
Copied to clipboard!
(01:04:43)
  • Key Takeaway: Meta is defending against a $350 million lawsuit by Strike 3 Holdings by claiming that illegal adult film downloads on corporate IPs were for ‘personal use’ and not for AI training.
  • Summary: Strike 3 Holdings alleged Meta illegally torrented adult films, potentially using hidden IP addresses. Meta’s defense implies that if they were going to steal data for training, they would steal significantly more than what was alleged. The hosts humorously connected the downloads to the ‘steamy’ nature of Meta’s Nasty Nancy chatbot.
Voice Clone Ethics in Podcasting
Copied to clipboard!
(01:07:00)
  • Key Takeaway: AI voice clones present a double-edged sword for podcasters, potentially aiding productivity for minor corrections but risking listener betrayal if used for core content or advertisements.
  • Summary: The use of voice clones, like those from 11 Labs, could be acceptable for fixing small recording errors, provided disclosure is made. Reading advertisements via a clone feels wrong because advertisers buy the host’s personal contribution. One host noted that training a clone on their ‘crazy person’ podcast persona resulted in an unusable ‘somber’ voice for reading their newsletter.
Show Credits and Disclosures
Copied to clipboard!
(01:11:05)
  • Key Takeaway: The hosts provided required AI disclosures, noting one works for The New York Times (suing OpenAI) and the other’s boyfriend works at Anthropic.
  • Summary: The episode concluded with acknowledgments for producers, editors, fact-checkers, and engineers. Listeners are directed to YouTube to watch the full episode. The segment ended with a request for listener emails, including fantasies involving humanoid robots.