Techin' Geek

Gemini 3 represents Google’s most significant advancement in AI thus far, providing enhanced reasoning, improved multimodal functionality, and upgraded Pro-tier features. From simple responses to complex workflows, it guarantees quicker, more intelligent, and dependable support throughout Search, applications, and the Gemini ecosystem.

Monarch Tractor, a startup focused on autonomous electric tractors, alerted its workforce on Thursday about the potential need to reduce its staff by over 100 employees or possibly even consider a ‘shutdown,’ according to a company-wide memo acquired by TechCrunch.

This communication follows recent job cuts at Monarch Tractor’s California headquarters and among its remote teams based in India and Singapore, as reported by multiple former employees who spoke to TechCrunch under the condition of anonymity.

Established in 2018, Monarch Tractor was created by a group that included a former senior executive from Tesla’s initial gigafactory along with Carlo Mondavi, a descendant of the renowned winemaking dynasty. The firm successfully raised over $220 million, which included $133 million in 2024, pursuing the vision of developing “driver optional” autonomous tractors suitable for use in wineries and various fruit farms.

While Monarch Tractor asserts it has delivered around 500 tractors so far, the company revealed a restructuring plan in late 2024 aimed at diversifying their tractor applications, such as managing feed in dairy operations and maintaining golf courses. At that point, CEO Praveen Penmesta mentioned Monarch Tractor’s intention to place greater emphasis on software services sales and licensing its autonomous technology.

One client — a dealer among Monarch Tractor’s earliest partners — alleges that the autonomous technology was never operational as promised, according to a lawsuit first highlighted by TechCrunch this week. Burks Tractor, an Idaho dealership, claimed that Monarch delivered “defective” units that bore “significant issues” upon arrival in 2024. Burks specifically accused Monarch’s tractors of being “unable to operate autonomously.” (Monarch refuted these allegations in a court filing.)

In the memo to employees on Thursday, Monarch Tractor indicated a stronger shift away from manufacturing tractors — a strategic move that may seem logical considering the startup lost its contract manufacturer, Foxconn, earlier this year.

“The revised business strategy will allow Monarch customers to introduce fully commercialized software as a service (SaaS) autonomy along with other software products directly to consumers, thereby creating new revenue opportunities for OEMs,” the human resources department wrote. “Regrettably, the timeline for completing this transition places Monarch at risk of closure.”

The memo advised employees that the company may permanently reduce its workforce by “up to 102 employees.”

Current employee numbers at Monarch remain uncertain. The startup had approximately 300 employees as of late 2024 before laying off over 10% of its workforce during the restructuring process. Former employees familiar with the recent reductions were unable to specify the exact size of the layoffs. Penmesta did not promptly reply to a request for comment.

Throughout this year, Monarch Tractor has also seen the departure of several key personnel, including co-founder Mark Schwager from Tesla.

“We initiated Monarch with an ambitious vision: to achieve electrification, automation, intelligence, and increased profitability in farming — all simultaneously,” Schwager noted in a LinkedIn post in July, while confirming he would remain on the company’s board. “Monarch is well-positioned and in capable hands for the next phase of its journey – making this transition timely.”

Nvidia’s founder and CEO Jensen Huang conveyed an optimistic outlook during the firm’s third-quarter financial results. Judging by the results, there could be justification for this perspective.

Nvidia posted revenues of $57 billion in the third quarter, marking a 62% increase compared to the equivalent quarter last year. The company’s net income on a GAAP basis reached $32 billion, up 65% year-over-year. Both revenue and profit figures surpassed Wall Street estimates.

The revenue landscape reflects a company thriving, mainly driven by its data center division. Nvidia’s data center revenue hit a record $51.2 billion, a 25% rise from the prior quarter and a 66% increase from a year prior. The remaining $5.8 billion in revenue derived from Nvidia’s gaming division, with $4.2 billion, followed by contributions from professional visualization and automotive sectors.

Nvidia’s CFO Colette Kress emphasized in a statement to shareholders that the data center segment has been propelled by a surge in computing, robust AI models, and agent-based applications. During the company’s Q3 conference call, Kress mentioned that in the preceding quarter, Nvidia revealed AI manufacturing and infrastructure initiatives totaling 5 million GPUs.

“This demand encompasses every market, including CSPs, sovereigns, modern builders enterprises, and supercomputing centers, along with numerous significant build outs,” Kress stated.

Blackwell Ultra, a GPU introduced in March and available in various configurations, has shown remarkable strength and is now leading within the company’s product lineup. Prior iterations of the Blackwell architecture also experienced continued strong demand, as noted by the company.

Huang remarked that sales of its Blackwell GPU chips “are through the roof.”

“Blackwell sales are through the roof, and cloud GPUs are completely sold out,” Huang stated in the company’s Q3 earnings report. “Computational demand continues to accelerate and compound across training and inference — both growing exponentially. We’ve entered a positive feedback loop in AI. The AI ecosystem is scaling rapidly — with a surge of new foundational model creators, more AI startups, across various industries and countries. AI is permeating everywhere, executing everything, simultaneously.”

Kress did mention that the company’s shipments of H20, a data center GPU designed for generative AI and high-performance computing, totaled 50 million, a disappointing figure due to challenges in selling to China.

“Significant purchase orders did not materialize in the quarter due to geopolitical factors and the increasingly competitive landscape in China,” Kress commented during the earnings call. “While we were let down by the current situation that limits our capacity to ship more competitive data center compute products to China, we remain dedicated to ongoing engagement with U.S. and Chinese governments and will continue to champion America’s ability to compete globally.”

Crucially, Nvidia anticipates further growth with projected revenues of $65 billion for the fourth quarter, contributing to a more than 4% rise in its share price during after-hours trading.

The bottom line, at least in Huang’s perspective: dismiss discussions of a bubble, as growth is the only narrative.

“There’s been considerable dialogue regarding an AI bubble,” Jensen expressed during the company’s earnings call. “From our perspective, we observe something quite different.”

The year is 2030. A “notorious mystery hacker” called the Puppet Master is causing chaos on the internet, infiltrating the so-called cyber-brains of multiple individuals as well as “every terminal on the network.” It turns out, the Puppet Master is a product of Japan’s Ministry of Foreign Affairs.  

In essence, the Puppet Master is what we would refer to today as a government-supported hacker, or advanced persistent threat (APT). However, in this instance, the “ghost” hacker goes rogue and is sought after for “stock manipulation, espionage, political engineering, terrorism, and breaches of cyber-brain privacy.”

This forms the foundational storyline of the Japanese anime cult classic “Ghost in the Shell,” which celebrated its 30th anniversary this week since its initial release and was based on the chapters named “Bye Bye Clay” and “Ghost Coast” from the first volume of the namesake manga, which came out in May 1989.

It might be an understatement to say that the Puppet Master’s narrative was ahead of its time. The World Wide Web, essentially what emerged from the internet as we recognize it today, was invented in 1989, the same year when the first volume of “Ghost in the Shell’s” manga — including the Puppet Master’s story — appeared in Japan. (The World Wide Web was publicly launched in 1991.)

In the manga, when the Puppet Master is apprehended, an official from Public Security Section 6, a body under the Ministry of Foreign Affairs, clarifies that they had been pursuing the hacker “for a long time,” and had “profiled his behavioral patterns and code/tech tendencies.”

“As a consequence, we were ultimately able to create a specialized anti-puppeteer attack barrier,” the official states in the manga. 

Risking too much extrapolation from a few sentences, the reality is that what the official describes is essentially what cybersecurity firms, including antivirus companies, do daily to block malware. They not only generate what are termed signatures based on the malware’s code but also on its characteristics and behaviors, referred to as heuristics. 

There are various other aspects of the plot that turned out to be remarkably prescient. 

At the outset of the Puppet Master inquiry, Major Motoko Kusanagi, the main character and leader of the counter-cyberterrorism unit Section 9, hacks into the Sanitation Department’s network to track a garbage truck. (Nowadays, government hackers employed by intelligence agencies often infiltrate large networks to monitor specific individual targets rather than extracting data from the hacked network itself.)

During this sequence, one of the garbage collectors confides to his coworker that he hacked into his wife’s cyber-brain because he suspects she is unfaithful. Shortly thereafter, we discover he’s been using a computer virus he acquired from “some programmer.” This clearly exemplifies tech-facilitated domestic abuse, or even stalkerware, which TechCrunch has extensively investigated in recent years. 

Ultimately, the abusive garbage collector had no spouse. His memories were entirely fabricated. His ghost — essentially his mind or consciousness — was hijacked by the Puppet Master with the intent of leveraging him to hack into government officials. In a similar fashion, this mirrors what some sophisticated hackers do when they infiltrate networks they then use to target their actual goals, as a means to obscure their trails, creating a separation from themselves and the ultimate target. 

The Puppet Master as a government hacker, the breaching of networks to monitor targets or utilize them to subsequently attack other networks, and a jealousy-driven hack are not the only intriguing elements of speculative fiction connected to hacking in the anime. 

John Wilander, a cybersecurity expert who authors hacker-themed fiction, crafted a thorough analysis of the film that emphasized details referencing real-world situations. Wilander provided instances, such as hackers reusing known exploits or malware to complicate attribution, investigating malware without notifying the creators, and leveraging computers for industrial espionage.

Clearly, the manga and anime take the fundamental — and plausible — notion of the Puppet Master as a hacker into more fantastical realms. The hacker, which is ultimately revealed to be an advanced artificial intelligence, can exert control over humans via their cyber-brains and possesses self-awareness to the extent that — spoiler alert — it requests political asylum and ultimately proposes to Kusanagi to merge their “ghosts,” akin to their minds.

To grasp how forward-thinking “Ghost in the Shell” was, it’s essential to consider its historical context. In 1989 and 1995, cybersecurity was not even a term yet, although the phrase “cyberspace” was famously invented by sci-fi writer William Gibson in his renowned novel, “Neuromancer.”  

Computer security, or information security, however, was already a tangible reality and had been for several decades, although it was an extremely niche specialization within the field of computer science. 

The first computer virus is thought to be the Creeper worm, unleashed in 1971 on the Arpanet, the government-developed network that became the precursor to the internet. A handful of other viruses and worms caused disruptions afterward, before they became widespread once the internet and the World Wide Web became operational. 

Arguably, the very first documented government espionage operation on the internet was the one uncovered by Clifford Stoll, an astronomer by training who also oversaw the computers at the Lawrence Berkeley National Laboratory in California. In 1986, Stoll observed a 75 cent accounting discrepancy in the network, which eventually led him to uncover that a hacker had infiltrated the lab’s systems. Ultimately, the hacker was identified and found to have been transmitting information from the lab and other U.S. government networks to the Soviet Union’s KGB. 

Stoll immortalized his months-long meticulous investigation in the book “The Cuckoo’s Egg,” a firsthand account that resembles a very detailed and thorough report by security analysts examining a hacking campaign executed by government hackers. “The Cuckoo’s Egg” has since become a classic, though it probably did not reach the mainstream audience upon its release. 

To the best of my knowledge, “Ghost in the Shell” creator Masamune Shirow never disclosed what real-life events inspired the hacking plot points in the manga. However, it is evident that he was attuned to what, at that time, was an obscure world that was foreign to most people on the planet, who were still years away from accessing the internet, let alone being cognizant of the existence of hackers.

When Anthony Sardain initiated the creation of Cavela in 2023, an AI enterprise designed to assist brands in automating supplier sourcing, he did not anticipate that new tariffs would lead customers to become more cautious about manufacturing products in China.

“You can’t just enter Vietnam and establish a supply chain overnight,” Sardain remarked to TechCrunch. “Many brands discover one supplier and cling to them for years, as they really don’t wish to lose that connection.”

This is particularly true for small and medium-sized businesses that do not have specialized global sourcing departments. Sardain asserts that Cavela addresses this issue through its AI agents, software solutions that autonomously execute tasks and serve as a personal procurement team in this instance. These agents can identify potential suppliers across more than 40 nations and negotiate product specifications and pricing.

On Wednesday, Cavela unveiled that it obtained $6.6 million in seed funding co-led by XYZ Venture Capital and Susa Ventures, with participation from Crossover Capital.

Locating suppliers and negotiating terms with them is a fundamentally complex and time-consuming procedure, and according to Sardain, it was unfeasible to automate without generative AI.

“Creating products entails dealing with a lot of text data, image data, diagrams, sketches, and photographs. This type of data wasn’t compatible with technology until AI came along,” Sardain explained.

The newest LLM and image models enable brands to upload their entire product data, which includes specifications, blueprints, and additional specifics about the items they wish to produce, directly into Cavela’s AI agent. Based on this information, the agent identifies numerous potential manufacturers and promptly contacts those factories via WhatsApp, email, or text to ascertain production capabilities and lead times, as well as to gather pricing estimates.

Sardain states that the entire procedure is effortless, alleviating brands from exchanging countless messages. “They log in a few days later, and there are quotes waiting in their inbox,” he mentioned. Typically, companies then request a shortlist of potential suppliers to provide product samples, facilitating the selection of one or more factories for final production.

Cavela estimates that its clients not only considerably lessen the time spent seeking manufacturing partners, but also achieve an average savings of 35% on production expenses.

“If you receive 100 quotes, you’re likely to stumble upon a much lower price, while also finding a superior supplier in the process,” Sardain added.

Cavela’s clientele, which includes fire-resistant apparel brand Western Welder Outfitting and men’s grooming brand The Longhairs, indicate that the startup has enabled them to discover manufacturers offering goods at a considerably lower cost — even beneath pre-tariff prices.

Sardain, who previously led data science for the data analysis startup Tierra, attributes his expertise in trade and sourcing to three generations within his family. He spent his upbringing across key Asian trade hubs such as Malaysia, Hong Kong, Thailand, Singapore, and mainland China, providing him with crucial insights into the approaches of local manufacturers.

Cavela’s rivals include Alibaba, which links brands with thousands of manufacturers in China, and Pietra, a brand operations startup that similarly assists businesses in sourcing goods with the aid of AI.

The volume of health information generated by individuals is rapidly increasing, from electronic health records and blood tests to continuous data from wearable devices. However, numerous users find it challenging to effectively connect and utilize this wealth of data to enhance their health.

Function Health, which provides a routine lab testing service to assist individuals in monitoring their health, aims to improve this situation by unifying health data and facilitating its use with an AI model. Recently, the company secured $298 million in a Series B funding round led by Redpoint Ventures, bringing its valuation to $2.5 billion.

This funding round also included contributions from a16z; Aglaé Ventures; Alumni Ventures; NBA players Allen Crabbe, Blake Griffin, and Taylor Griffin; Battery Ventures; the investment firm NFDG, founded by Nat Friedman and Daniel Gross; and Roku founder Anthony Wood. With this round, the total capital raised by the company reaches $350 million.

Along with the funding, Function introduced the Medical Intelligence Lab, an initiative to create a “medical intelligence” generative AI model capable of delivering personalized health insights based on individual user data, content, and research. The company stated that doctors are involved in training the model. For its clients, Function offers an AI chatbot that can address inquiries based on their health data, utilizing previous lab results, physician notes, and imaging to provide customized recommendations.

“It is insufficient to exist in a world with AI and not utilize it for health purposes,” commented Jonathan Swerdlin, CEO and co-founder of Function, in a discussion with TechCrunch. “You ought to be able to manage your biology. Function Health’s goal is to employ the best available technology to enhance human health.”

Swerdlin pointed out that the platform complies with HIPAA standards, encrypts user data entirely, and never commercializes personal information. “Your data and identity are never for sale. Every piece of your information is comprehensively encrypted and secured. We are devoted to safeguarding you and your data.”

Function’s chief medical scientist, Dr. Dan Sodickson, alongside co-founder and chief medical officer Dr. Mark Hyman, is spearheading the development of MI Lab, along with a team of physicians, researchers, and engineers. The MI model is developed with training from physicians who remain actively engaged in the process, according to Swerdlin.

Despite the various players in the industry, Function distinguishes itself from competitors such as Superpower, Neko Health, and InsideTracker with its device-agnostic model, Swerdlin mentioned, noting that the platform combines lab testing, diagnostics, and clinical insights to deliver more than a conventional AI coach or wellness application.

Function operates 75 locations across the U.S., and members “can have lab tests performed at 2,000 Quest locations,” he stated. Function claims it has carried out over 50 million lab tests since 2023.

This article has been revised to confirm that Function members have access to tests at 2,000 Quest locations.

Jennifer Neundorfer, co-founder of January Ventures, appeared on the Equity podcast at TechCrunch Disrupt to discuss fundraising in the current AI-centric market.  

Both founders and investors are captivated by AI, with Neundorfer noting that her firm seeks to leverage AI to enhance efficiency, particularly in conducting due diligence on the market and competitors. She favors founders aiming to build something completely innovative.  

“My excitement often arises when I observe someone utilizing AI to create something that isn’t just 10x better. It’s about developing an entirely new experience or workflow or behavior,” she commented. “That’s our focus. We’re looking for less incremental changes and more entirely new behaviors.” 

However, this is becoming increasingly challenging for founders, as fatigue has set in with numerous AI concepts starting to seem repetitive, according to her.

“Founders tend to break through when they can articulate to investors what distinguishes their approach as truly different from the multitude of startups pursuing similar ideas and why they are the right team to tackle that,” she stated.  

Regardless of whether we are experiencing an AI bubble, Neundorfer believes a market adjustment is likely, suggesting that many companies currently enjoying substantial investment may not endure. The successful ones will adeptly navigate this period by creating “truly category-defining companies,” aligning with the future trajectory of technology. “Founders who can remain ahead of that curve, innovate at the forefront of what is feasible today, and prepare for what’s on the horizon,” she identified. “Those founders who can accurately gauge the market and comprehend what their customers truly desire, rather than merely what is feasible, will possess an advantage.”  

On another note during the podcast, she reflected on her prior experience before entering venture capital, having worked at YouTube and 21st Century Fox.  

“A significant portion of my role involved engaging with individuals who possessed remarkable technology,” she reminisced about her tenure at 21st Century Fox. Interacting with tech enthusiasts was the aspect of her job that brought her the greatest joy and made her realize how much she would likely enjoy collaborating with early-stage founders.

However, the transition to investing came with a steep learning curve. Initially, she frequently checked in with founders, providing them with detailed feedback on their ventures.  

“This approach works for some situations, but it’s really about cultivating a relationship with the founder, offering support that extends beyond business discussions to personal encouragement,” she added.  

Now, she feels at ease in her role. She acts as a mentor for various organizations, including Techstars, and has made over 50 investments at January Ventures, according to PitchBook, achieving several exits along the way.  

Throughout their discussion, Neundorfer addressed the evolving venture landscape, funding opportunities for minorities and women, and the success of venture markets outside of San Francisco. Her principal advice for diverse founders in the current climate also applies broadly: disregard the distractions and concentrate on creating a solid company.  

“Anything else becomes something they can’t control, and the anxiety isn’t worth it.”  

Spotify has announced its acquisition of the community-driven database that tracks sampled music, WhoSampled, on Wednesday. The streaming platform also included this information in its blog post regarding new song credits features released that same day.

While the specifics of the deal weren’t revealed, Spotify confirmed to TechCrunch that the acquisition encompasses both the WhoSampled team and its database. (According to data from PitchBook and LinkedIn, the company employed only a small team of around 10 individuals.)

Founded in 2008, WhoSampled, based in London, provides a comprehensive database of songs, samples, covers, remixes, and artists, among other content. Its website indicates that it now tracks over 1.2 million songs and nearly 622,000 samples. This information is fueling Spotify’s latest features, including the newly announced music discovery tool, SongDNA.

Notably, WhoSampled had been on Spotify’s radar for some time, having formed a partnership with the streaming service in 2016 that allowed users to link their Spotify playlists and saved tracks within its platform.

In its announcement regarding the acquisition, WhoSampled stated that its independent platform and brand would continue post-deal, with expectations for user-facing enhancements as it integrates into Spotify’s broader operations. For example, the company mentioned that it would expedite moderation times for submissions, eliminate display ads shortly, and make its mobile applications available for free downloads with complimentary subscriptions.

“During our recent talks with Spotify, it became apparent that we both strongly believe in the importance of musical context — and share a vision for helping listeners explore deeper into their favorite songs,” reads the statement from the WhoSampled team. “Spotify has demonstrated a true commitment to our mission, and we couldn’t have envisioned a better home for WhoSampled,” it concluded.