Techin' Geek

The newsletter service Beehiiv is venturing into fresh engagement realms with a new feature named Community, enabling a space for creators’ subscribers to communicate with one another. Additionally, the company has introduced an AI Copilot designed to assist creators in audience management and growth.

These enhancements are part of Beehiiv’s strategy to establish itself as a creator-centric platform that extends beyond newsletters. Recently, Beehiiv has rolled out podcasts, webinars, and adaptable paywalls. Some of these initiatives have already yielded favorable outcomes, such as 50% of podcast users transferring their shows from other platforms.

With the new Community feature, users can initiate discussions within the platform. Currently, creators often utilize separate Discord, Slack servers, or Facebook groups for member chats, but Beehiiv aims to centralize these discussions back on its platform. Creators can establish paid membership levels for exclusive chatroom access and manage discussions.

“Individuals engaged with your content share a common interest in your creations, yet they’re unable to converse with one another. Whether the focus is sports, the World Cup, or politics, having a space where your audience can interact is incredibly beneficial,” stated Beehiiv CEO Tyler Denk in a discussion with TechCrunch.

The platform is also adding a new revenue stream through programmatic advertising, allowing users to monetize ad slots in their newsletters. They can select ads that potentially yield the highest returns based on their audience, content, and metrics.

Beehiiv already offers functionalities like metered paywalls, paid trials, and a sponsorship storefront for selling ad slots in packages. Moreover, the company reported that publishers on its platform generate over $1 million monthly via its ad network.

Additionally, Beehiiv is unveiling an AI assistant called Copilot, which comprehends contextual elements such as content, audience, subscribers, and performance to guide users on managing their newsletters and expanding their readership. The assistant can review the effectiveness of various newsletters and podcasts, draft outreach campaigns, and identify new revenue opportunities.

This assistant is part of a broader suite of AI initiatives. Earlier this year, Beehiiv launched a model context protocol (MCP) server, allowing users to connect their Beehiiv accounts to other assistants like ChatGPT and Claude for inquiries and insights. The company is also improving AEO (Answer Engine Optimization), which enhances visibility for newsletters in AI assistant responses.

Alongside these innovations, the company is delivering a revamped editor that allows users to view editing and preview modes simultaneously, facilitating a better understanding of how their written content will appear to readers.

Denk mentioned that in the upcoming quarter, Beehiiv intends to focus on educating users about these tools and demonstrating how leading newsletters are leveraging them for growth.

Competitors in the platform space are also adapting by introducing new features. For example, Riverside initiated a newsletter publishing function last month, while Substack unveiled a built-in recording studio tool in March.

X is adopting a stricter approach towards individuals who attempt to exploit its creator revenue-sharing program by soliciting engagement and appropriating content from others.

It is not unusual for viral material to be redistributed, particularly on platforms such as X, where that material can garner likes, followers, and, in some situations, actual cash rewards. Instagram, Facebook, and Reddit also face this issue and have introduced various technical measures to mitigate such actions — for example, tools that recognize when a user has re-shared someone else’s work without providing credit.

X has recently joined in on this effort, adding an enhanced video editing and recording feature to its platform to motivate creators to utilize X’s own tools rather than appropriating the work of others.

According to Nikita Bier, X’s latest iteration of the Grok AI model can identify duplicated content at three times the speed of its former version. Moreover, adding watermarks, intros, and other modifications designed to mislead viewers into believing that appropriated content is original will lead to the monetized impressions being credited to the original uploader. This also applies to replicated viral text posts. (Bier mentions that a frequent example of such content is, “Twitter is like the smoking section of the internet”—demonstrating that people are still referring to the app as Twitter!)

Bier indicated that the company has identified 1.5 million instances of content theft in its most recent cycle, although he did not specify the duration of that cycle.

He did, however, mention that as a result of these changes, over $1 million in creator earnings will now be redirected to the original creators of the appropriated content.

Given that many of these issues are facilitated by AI, X has also been working to suspend bots more rapidly. In April, Bier stated that the platform was identifying and suspending “208 bots per minute and increasing.”

X now asserts that repeated or deliberate efforts to bypass its new policy will lead to users being excluded from the creator program, as will attempts to solicit engagement and followers. Specifically, if a user is caught three times or more (for instance, by stating something like “I’ll follow everyone who replies”), their account will be removed from the creator program and referred to the platform’s policy team for suspension.

Bier has previously expressed concerns about engagement bait on the platform, even calling out top creator MrBeast for frequently using financial incentives to attract viewers to his videos.

Google might introduce an AI product under one name in its testing phase, but it will ultimately connect everything to Gemini. In the latest instance of this pattern, the firm is rebranding its AI-based research tool NotebookLM to Gemini Notebook. The corporation is also incorporating features to enhance the interactivity of the tool by integrating code execution for data analysis.

The company initially showcased NotebookLM at Google I/O in 2023 as Project Tailwind, and since its launch, it has evolved into a product that is utilized by 30 million users and over 600,000 organizations. Over the past three years, the firm has introduced functionalities such as interactive podcast creation, curated notebooks, video summaries, support for additional file formats, and an enterprise solution.

Thanks to NotebookLM, other businesses and startups have incorporated features for podcast creation from foundational material and research instruments.

In conjunction with the rebranding, Google is releasing a new update that transforms each notebook into its own secure container, allowing users to produce code that makes outputs interactive. It emphasized that with the code execution feature, users can access multiple sources and perform intricate data analysis directly within the tool.

The company indicated that the update is accessible to users on the Google AI Ultra paid plan, as well as Workspace business clients with AI Ultra Access and AI Expanded Access. Pro users will gain access to this feature in the upcoming weeks.

Google stated that users can currently view their notebooks in the Gemini app, and soon, they will have the ability to access them via AI Mode in search.

On Thursday, Google declared that it now permits users to connect and engage with several of their favorite applications directly in AI Mode, the company’s conversational search feature. Initially, the supported apps feature Instacart, Canva, and YouTube.

With this enhancement, Google is broadening AI Mode from merely answering inquiries to performing tasks using the applications they frequently utilize. The tech giant is likely aiming for users to depend on AI Mode more frequently for activities such as organizing and shopping. Furthermore, this rollout will enable Google to enhance its competition against rivals like OpenAI’s ChatGPT and Anthropic’s Claude, both of which incorporate app integrations.

For instance, Google mentions that if you’re organizing a barbecue and using AI Mode to build a grocery list, you can link your Instacart account to directly include ingredients in your shopping cart and swiftly check out on the Instacart application or website.

Alternatively, if you’re engaged in a project and seeking design inspiration, such as for a flyer, you can request Canva to present a variety of templates. In another scenario, Google states you could utilize AI Mode to assemble a playlist for your upcoming party and promptly save it to YouTube Music.

The update is being implemented for users in the U.S. Google indicates it is collaborating with a range of partners and intends to introduce support for additional applications shortly.

Today’s announcement enhances a function that Google introduced earlier this year during Google I/O that allows users to link third-party applications to the Gemini app for swifter task completion. Supported applications feature Canva, OpenTable, Spark, Instacart, among others.

Since its debut in early 2025, Google has been progressively enriching AI Mode with further capabilities. Recently, the tech giant revealed that AI Mode can now assist in confirming whether an item you require is available at a nearby store. Google has also added the ability for users to navigate the web concurrently with AI Mode to accomplish tasks such as comparing details and posing follow-up questions while maintaining the context of their inquiry.

Earlier this year, Google unveiled “Personal Intelligence” within AI Mode, empowering it to access users’ Gmail and Google Photos to provide more personalized answers.

Sudo prepare me a sandwich. The future is here! DoorDash has launched a limited beta version of DoorDash CLI, a command-line utility for developers that enables ordering DoorDash straight from your AI assistant. According to the company, the tool allows users to search for stores, discover discounts, and complete transactions.

Referred to as “dd-cli,” this innovative tool is available to U.S. and Canadian macOS developers via a waitlist, as stated by DoorDash co-founder and CTO Andy Fang in a post on X. DoorDash was approached for feedback regarding the new feature.

The announcement is attracting considerable interest because, at first glance, it’s rather amusing. Command-line interfaces are typically linked with coding, not lunch orders. An AI assistant executing commands to order your salad or sandwich may seem quite ludicrous initially.

However, the DoorDash CLI is not merely a joke; it exemplifies what agentic commerce could resemble.

Through this initiative, the company is making DoorDash’s ordering system accessible to AI agents, giving developers the opportunity to enhance functionality within their own applications and services. This implies that rather than using the DoorDash app, developers could create their own solutions for food ordering, grocery shopping, or locating local lunch discounts, among other aspects, or utilize those features as foundational elements in conjunction with other tools.

DoorDash has also tried offering its service through iMessage and currently has its AI chatbot, “Ask DoorDash” — providing two instances of how agentic commerce can function. The service is also made available to AI chatbots, such as OpenAI’s ChatGPT and Claude.

The company’s registration form for access to the new CLI tool includes a section asking developers what they would create if granted access to the beta.

The launch carries a humorous undertone, reminiscent of that classic XKCD comic about programmers automating absurd tasks — like making a sandwich. In the comic, a programmer requests “make me a sandwich,” to which the other person replies, “What? Make it yourself,” prompting the programmer to say “sudo make me a sandwich,” and the other person responds, “OK.” (It’s programming humor, after all!)

The video included in the X post amplifies the over-engineering theme, as it interacts with Slack, retrieves memories, processes JSON, inspects menu layouts, executes Python scripts, recovers from issues, and computes totals, all to accomplish something as straightforward as ordering three salads. During the operation, the interface displays “Flibbertigibbeting,” making the scenario even more comical.

On Thursday, police in the U.K. announced that the imprisonment of two hackers has significantly hindered the operations of the notorious cybercrime group known as Scattered Spider.

Owen Flowers, 18, and Thalha Jubair, 20, admitted guilt earlier this year for hacking Transport of London (TfL), the government entity managing the public transit system of the U.K. capital, in 2024. They received sentences of five years and six months in prison on Thursday. 

The incarceration of Flowers and Jubair serves as a reminder that often the most dangerous and proficient hackers are not employed by advanced government organizations with substantial budgets. Instead, they frequently comprise young and intelligent individuals driven by financial gain and notoriety among their peers.

Groups like Scattered Spider and ShinyHunters, another criminal network, typically focus on targeting and manipulating individuals and employees rather than solely compromising computer systems, a tactic that proves both effective and challenging to combat.

While members of hacking groups may fluctuate, the entities themselves can rebrand. However, U.K. authorities believe the imprisonment of Flowers and Jubair constitutes a major setback for Scattered Spider, a fluid organization tied to numerous notable attacks, including those against the casino giant MGM, airline WestJet, and cybersecurity company Okta. These breaches subsequently provided the hackers access to many customers of these firms. 

“Scattered Spider has posed the most considerable cybercrime risk to the U.K. in recent times. This investigation has significantly undermined that risk and brought important perpetrators to justice,” remarked Paul Foster, head of the U.K. National Crime Agency’s National Cyber Crime Unit. 

The two hackers executed the cyberattack against TfL in the summer of 2024, resulting in the shutdown of the system’s infrastructure, including the ticketing system and the online real-time train arrival information. The disruptions persisted for several weeks.

Flowers and Jubair were apprehended a year later. At the time, the FBI accused Jubair of engaging in attacks on over 120 companies employing social engineering techniques. 

Authorities stated that the attack on TfL led to losses estimated at approximately £29 million (around $47 million). The hackers gained such extensive access to TfL’s systems that they “could have completely locked out and disabled TfL,” holding “the keys to the kingdom” to the company’s infrastructure, according to The Guardian.

“Your information is confidential. Period,” states the period tracking application Stardust on its website. However, new research from Mozilla indicates that some users may consider this assertion to be questionable.

According to Mozilla’s recent research that analyzed the privacy measures of period-tracking applications, Stardust was discovered to be sharing sensitive health information of users with the analytics firm RudderStack. This information encompassed the user’s birthdate, type of birth control, reproductive aims, and particular symptoms being experienced by the user, all linked to a unique identifier instead of the user’s name. (The FTC has previously cautioned that this does not render the data anonymous or safeguard it from being traced back to an individual.)

Mozilla’s investigation highlights the privacy and security hazards associated with period tracking and other health applications that distribute data to third parties. Frequently, this occurs as background processes within the application and remains unnoticed by the user. It is not unusual for applications to send data to other services for purposes such as storage, analysis, and payments, but sharing individuals’ information with external entities inherently introduces risks such as possible security failures, data breaches, or having the data pursued by law enforcement.

TechCrunch had previously reported on Stardust in 2022, when the app experienced a spike in downloads following the reversal of the constitutional right to seek an abortion within the United States. Stardust asserted that it utilized end-to-end encryption — signifying that even the company could not access the data of its users — but TechCrunch determined through network traffic analysis that this claim was inaccurate.

Mozilla security researcher Shoshana Wodinsky employed a similar method of analyzing network traffic for several period trackers, including Stardust, to investigate how the applications gathered and shared data (if at all) with third parties. Wodinsky discovered that Stardust was the only application among the six examined that transmitted users’ sensitive health data to another organization.

As reported by BBC News, a spokesperson from Stardust remarked that RudderStack is “legally forbidden from selling or utilizing it for its own ends.” As companies based in the U.S., both Stardust and RudderStack remain subject to requests for users’ information from law enforcement regarding health information stored on their servers.

Rachel Moranis, the founder of Stardust, did not reply to TechCrunch’s inquiry for comments on Thursday, nor to questions about whether the company had faced requests for its users’ data. A representative acknowledged the receipt of an email but did not furnish any comments.

Out of the six applications evaluated by Wodinsky, Mozilla identified Euki as “squeaky clean,” as the app was not found to be sharing any data with third parties concerning its main features, and the user’s health data remained on their device.

You might have seen that OpenAI unveiled its initial hardware this week: a $230 mini keyboard. What you may not know is that along with this “command center for agentic work,” OpenAI also introduced a ChatGPT basketball.

“This basketball originates from the Pause. Play. Prompt. initiative, serving as a tangible reminder that creativity isn’t confined to our screens,” the product page states. I could not locate any additional references to the “Pause. Play. Prompt.” initiative on OpenAI’s site, but it seems to signal OpenAI’s intent for people to not be glued to Codex all day.

Who claims tech firms aren’t considering our mental well-being?

The basketball is priced at $70, or roughly 56 million input tokens for GPT-5. It’s composed entirely of rubber, making it more suitable for outdoor play because of its weather durability compared to the pricier leather balls you’ll encounter on professional courts. I appreciate that OpenAI is imagining a scenario where outdoor sports are feasible, even while the surge of generative AI boosts tech companies’ carbon footprints.

It’s hard to picture who the ideal consumer is for the ChatGPT basketball. Who is this meant for?

Step outside the protective bubble of an AI-saturated, token-maximizing Silicon Valley, and one might fear being teased for bringing a ChatGPT basketball onto the court. You couldn’t pay me $70 to show up at a community court in Philadelphia with this ChatGPT basketball. (If it were complimentary merchandise from a conference, it might be seen as ironically cool — I hold dear my “#FACEBOOK” tote, styled like a 2000s bar mitzvah party favor.)

In defense of the ChatGPT basketball, the AI sector isn’t precisely recognized for its savvy regarding product-market alignment. May the Humane Ai Pin rest in peace.

Along with that $70 item, OpenAI is also offering a collection of merchandise adorned with motivational phrases, such as “Good research takes time,” which I’d argue is ideal apparel for a startup founder discussing with investors demanding accelerated growth. There’s an undeniable je ne sais quoi about the $175 quarter-zip that features “research” in cursive. The product description claims it “boasts a sharp collar reminiscent of our academic days,” which might exclude those in the “I never went to college because I’m a coding genius” demographic. (Additionally, can an object truly reminisce about academic experiences? Should I expect grammatically correct phrases from individuals who compose their emails using ChatGPT?)

But there’s nothing wrong with a bit of corporate swag. If OpenAI is seeking to commission ceramic artisans to celebrate the company’s history through practical tableware, I’d like to express my interest.

Andrew Dai departed Google DeepMind with a clear intention to establish himself in the realm of visual AI. He executed a rapid fundraising effort that yielded a valuation-to-capital ratio more aggressive than Thinking Machines, known for securing one of the largest funding rounds in U.S. history.

In this installment of Build Mode, host and Startup Battlefield lead Isabelle Johannessen converses with Andrew Dai, founder and CEO of Elorian and a former Google DeepMind researcher, about how his company secured a $55 million seed round at a $300 million valuation just shortly after departing Google.

With over ten years of experience aiding in the development of some of the most significant AI systems globally, including research that later contributed to ChatGPT, Andrew articulates his belief that visual AI represents one of the upcoming critical frontiers in artificial intelligence. “We have models excelling in mathematics, advancing new physics concepts, and, of course, coding has gained popularity … However, progress has been particularly inconsistent in visual comprehension and visual reasoning,” stated Dai. “At Elorian, our goal is to create models that lead us towards visual AGI.”

Andrew outlines the fundraising journey from the founder’s perspective, detailing how he transformed a highly intricate vision into a compelling narrative for investors. He emphasizes the importance of choosing strategic partners like Nvidia and Menlo Ventures over even higher valuation offers, and how selecting investors who grasp the realities of advancing frontier AI turned out to be more beneficial than merely maximizing his company’s valuation.

The discussion also imparts practical insights for founders maneuvering through today’s swiftly changing AI environment. Andrew reveals how startups can convey complex technical concepts without resorting to jargon, why speed has become a key competitive edge in AI, and what it requires to entice top-tier researchers away from Big Tech.

In this episode, you’ll discover:

• What leading venture capital firms seek when investing in frontier AI startups.
• Why the highest valuation isn’t always the most favorable fundraising outcome.
• How to present highly technical products to nontechnical investors.
• What founders should consider when selecting venture capital partners.
• How startups can attract exceptional AI talent away from Big Tech.
• Why speed has emerged as a significant competitive advantage in AI.
• How founders can construct sustainable moats as AI technology progresses.

This season on Build Mode, we’re exploring all facets of fundraising with experts who possess firsthand experience in raising considerable pre-seed rounds, making substantial investments, bootstrapping, going public, and navigating the unpredictable market conditions that can alter the landscape.

Subscribe to Build Mode on Apple Podcasts, Spotify, or your preferred listening platform. Watch the complete videos on YouTube. New episodes of Build Mode are released every Thursday.