Science Corporation, a startup focused on innovative brain-computer interfaces (BCI), received approval from Europe’s medical device authority to commence sales of a device that restores vision impaired by age-related macular degeneration.

The firm announced that the device, named PRIMA, has also obtained a designation from the US Food and Drug Administration, marking the initial step toward a fast-tracked regulatory review, which may allow the device to be utilized for treating two rare forms of blindness.

Countless individuals globally experience age-related macular degeneration, which deteriorates the light-sensitive cells at the rear of the eyes, complicating reading and facial recognition.

For utilizing the device, patients experiencing this vision loss undergo a one-hour outpatient operation that inserts a small chip at the back of their eye. Subsequently, they wear glasses equipped with a camera that relays their surroundings to the chip. Max Hodak, the founder and CEO of Science Corp., states that this product restores functional vision to individuals who have lost it.

“One of our patients in France recently completed a 300-page novel and sent us the book,” Hodak mentioned to TechCrunch. “We have a sketch on the wall [that] one of our patients drew of the Sydney Opera House. There are videos of patients engaging in crossword puzzles and completing Sudoku.”

Hodak is recognized as the co-founder and former president of Neuralink, Elon Musk’s BCI startup. He departed in 2021 to establish Science Corp., with ambitions to invent a new BCI that combines silicon chips with living cells. However, the company first needed to validate its methods and create a viable business.

“What this field requires is a company generating $100 million per year in revenue,” Hodak commented. “There’s a risk of a downturn affecting the entire industry, so we believe it’s crucial to establish a sustainable business as we advance these long-term technologies.”

Hodak and his team are confident that this sustainable business will come from restoring sight to the blind, particularly patients whose conditions arise from issues with the light-sensitive cells in the back of the eye. After investigating various methods, they concluded that Pixium, a French company that created the PRIMA technology, was on the right track and acquired the company in 2024. Science Corp. utilized its internal platform to develop the documentation and enhance the product in preparation for regulatory approval and market entry.

Each PRIMA device is anticipated to be priced in the hundreds of thousands of dollars; Science Corp. and its medical collaborators in Europe are currently negotiating with healthcare providers regarding reimbursement. The company is setting the stage to start offering PRIMA in Germany, where its clinical trials took place, and the first procedure could occur in September.

Science Corp. plans to keep enhancing the vision capabilities of PRIMA with a new chip, alongside modifying the design of its glasses, which presently necessitate a battery pack to function. The objective is to create something akin to Meta’s AR glasses, although the power and processing demands for PRIMA are greater.

Science Corp. is also collaborating with Dr. Murat Günel, chair of Yale Medical School’s neurosurgery department, to devise procedures for human trials of a directly implanted bio-hybrid brain sensor.

Hodak asserts that launching PRIMA is “the most critical endeavor for the company, because we can’t pursue the bio-hybrid initiatives in the long term without a robust vision business. That’s what mainly finances everything else — that’s the aspect investors understand how to create financial models around.”

With the transformation of social media from a means to engage with friends to expansive entertainment platforms, a startup called Yope is quietly shaping what it envisions as the future of online connections: social networking devoid of algorithms, advertisements, and public content.

Currently supported by a $12.3 million investment round led by Northzone, Yope’s app is competing against tech titans such as Facebook, TikTok, Snapchat, and Instagram by creating a social platform focused on what it terms “micro communities” — small groups where friends and family communicate privately, sharing photos, videos, messages, and soon, playing games.

The concept of establishing an algorithm-free, private social media experience isn’t innovative, yet few apps have succeeded in scaling without integrating creator-driven content to maintain user engagement. Yope aims to shift this trend by fostering connections among real-world friends, positioning it equally as a communication tool and a platform for photo sharing or life updates.

In Yope, user accounts are private by default, and it lacks an algorithmic feed. Additionally, the platform does not incorporate advertisements, choosing instead to cultivate a premium, subscription-based experience for its dedicated users.

Bahram Ismailau, Yope’s co-founder and CEO based in London, mentions that the team recognized the prospect of a new kind of social network after conducting over 100,000 interviews — assisted by AI tools — to better comprehend how youth globally utilize social networks. They discovered that nearly 30% were employing what are often termed photo dump or “spam” accounts, where individuals share more honest photos with a smaller circle of actual friends.

Moreover, numerous individuals don’t post anything publicly, opting instead to rely on messaging to stay connected with friends.

“[Young people] don’t possess a space for self-expression unless they are prepared to become influencers,” Ismailau stated during an interview with TechCrunch. Nevertheless, teenagers still wish to express themselves; they simply want enhanced privacy.

“We identified this significant opportunity to establish a new environment for youth to socialize. We named it Yope, and it represents an AI-native social platform tailored for the younger generation and what’s to come.”

He emphasized that the group doesn’t endorse using AI to produce content; instead, it sees AI as a means to facilitate better connections. This includes employing AI to develop mini-games that users can enjoy with friends, a feature anticipated to debut in approximately a month. The platform also plans to harness AI to assist users in meeting physically, possibly through purchasing event tickets or finding a venue to watch a sports game.

On Yope, users create profiles by sharing photos, which may be transformed into cut-out stickers. Rather than being organized into albums, the photos appear in a collage-like, seemingly disordered layout on each user’s “wall.” Shortly, Yope users will be able to further tailor their space by including their interests, favorite songs, mini-games, as well as customizing aesthetics with colors and wallpapers of their choice.

This concept is reminiscent of Myspace, where users used to cultivate their own areas on the internet, customizing them to mirror their identities.

The app also draws heavily from current social platforms, providing features that have now become standard, such as in-app messaging, highlights of top moments (generated by AI), and lock screen widgets that display friends’ photos. During the onboarding process, the app guides users through the setup of its various functionalities and authorizing the necessary permissions to access their photos and connect with friends, motivating them to add contacts and initiate private conversations.

This blend of features seems effective — Yope currently boasts nearly 15 million users, who collectively share between 10 million and 20 million pieces of content every week, including photos, videos, or stickers. These figures indicate consistent activity. Ismailau asserts that over 50% of Yope’s users access the app at least five times a week, categorizing them as power users.

Furthermore, it’s not solely young individuals adding their friends. Ismailau mentions that roughly 20% of active users have encouraged an older family member to join the platform.

This usage has captured investor interest, as Ismailau states that Northzone approached the firm, rather than the reverse. Northzone, which has previously supported other consumer applications like Spotify and Klarna, led the $12.3 million round, with backing from Inovo, Redseed, and Geek Ventures. This funding round elevates the company’s total financing to $20 million. (Yope results from two earlier pivots — TechCrunch reported on a prior iteration of the company. The existing form has been in development for around two years.)

“Yope represents a modern, empowering, and secure approach to social media, where users hold complete control over their experience, contrasting with the exploitative strategies of Meta, TikTok, or X,” stated Pär-Jörgen Pärson, partner at Northzone. “We are thrilled to collaborate with Bahram, Paul [Rudkouski, co-founder], and their team to expand the service well beyond its current millions of users and the half a billion moments shared.”

The capital will be utilized to enhance the product, expand the team — which now comprises around 35 individuals — and establish a presence in the United States.

Yope is available for free on both iOS and Android.

Correction: Users are sharing 10 to 20 million pieces of content weekly, not daily; the founder made a mistake. The post has been updated accordingly.

As the capabilities and appeal of Chinese open-weight AI models escalate, discussions regarding the appropriate responses to them have surged once more.

There’s speculation that the Trump administration may attempt to impose a ban (although no actions have been taken yet). In the meantime, creators of proprietary models, particularly OpenAI and Anthropic, seem increasingly apprehensive about these developments.

Open-weight models like Moonshot AI’s Kimi K3 or Alibaba’s Qwen provide inference at a fraction of the cost per token compared to the closed source models produced by these major U.S. labs. The concern is that they might represent a form of threat. Undoubtedly, they jeopardize the profit margins of the large proprietary AI laboratories.

But should companies utilizing these models in their own data centers give in to the anxiety that they could become a conduit for Chinese hackers?

No, asserts Lucas Atkins, the Chief Technology Officer of Arcee, which is developing open models to provide U.S. firms with a domestically produced alternative to Chinese variants.

If any startup would benefit from a prohibition on Chinese models, it would be Arcee. However, Atkins claims that China’s open models are not any more perilous than any other open source software a company might employ. In fact, he mentions, they even offer advantages to his own enterprise.

“Many perceive this as akin to a Chinese software application. Like, it was developed with these x, y, z objectives” that a malicious actor could potentially manipulate, he remarked.

“That is fundamentally not how these models undergo training. There is essentially no way for an Arcee, or an Alibaba, to create a model, have it executed in someone’s setting and have us gain any access to it at all,” he clarified.

Although most of these models are referred to as “open weight” and aren’t entirely open source software, the source code (the portion that will actually operate on servers), if acquired from open source platforms like Hugging Face, is predominantly transparent and subject to scrutiny. (What remains inaccessible is the methodologies and data utilized to train the models.)

Large organizations ought to subject any model core to their security evaluations and inspection protocols, and they often retrain the models for their specific applications, examining factors such as bias, toxicity, hallucinations, and sensitivity to particular subjects. Thus, they analyze, optimize, and comprehend the models before users commence submitting prompts.

Could a model designed for coding potentially insert harmful backdoors into the code it generates? While that is theoretically feasible, it would necessitate intricate maneuvers to achieve.

“There’s no reason a sufficiently skilled actor couldn’t train a model to be an outstanding coding resource in every situation, but when faced with a specific type of codebase… some concealed training would be triggered,” speculated Atkins, who dedicates his time to training models. However, he adds: “I’m not sure how one would accomplish this.”

Given that large language models are inherently creative, the likelihood of eliciting a contemporary model to produce malware in reaction to a meticulously planned perfect storm of context and prompt is minimal. Even more unlikely is that any organization would opt to use that code.

Could such an event occur in the future? That remains uncertain. However, organizations are also constructing their AI applications to be model-agnostic and to integrate multiple models. Thus, even if Chinese models are the most cost-effective today, firms won’t be confined to using them indefinitely.

“I believe rather than the dialogue focusing on how to ban Chinese models, it should shift to how we can cultivate a robust, open ecosystem here in the U.S.,” remarks Atkins.

Arcee also receives benefits from Chinese models. Their openness allows the startup to “gain from those models being effective because we can learn from what they have accomplished. We can build upon them. Then they can learn from our advancements,” he states. “We hold immense respect for the individuals developing those models, the individual researchers.”

Ultimately, the approach to compete with Chinese models “is to release a model that surpasses theirs,” asserts Atkins. “We need to provide them with something worth discussing.”

Substack has introduced a new feature that reveals which of your preferred newsletters are generated using AI technology.

This week, the newsletter and writing platform unveiled an integration with Pangram, an AI writing detection tool, enabling users to analyze posts, comments, and replies on Substack’s app to estimate the proportion of content authored by humans versus AI.

In the immediate future, this initiative might negatively impact Substack’s business, as it could reveal numerous newsletters on its platform that are not entirely crafted by humans. This might weaken trust in the platform’s collection of independent news and blogs, potentially harming its standing as a provider of high-quality content.

However, in the long run, AI-detection capabilities could help keep Substack devoid of “AI slop” and foster greater user confidence that what they are reading is penned by real individuals or, at the very least, help them understand when it’s not.

Substack follows several platforms that are beginning to label AI-generated content, particularly as AI assumes a larger role in the creation of such content. Social media sites now mark photographs and videos created with AI, while music streaming services have recently started labeling and sometimes penalizing AI-generated music.

“This is a positive application of AI,” stated Substack CEO Chris Best.

“When I used to present Substack to authors, I would say… we’ll handle everything except the challenging part,” he elaborated during an online discussion with Pangram’s founder, Max Spero. “You need to possess something — an idea that’s worth reading, worth caring about, worth sharing. That single element is quite difficult and immensely valuable… [S]oftware should manage everything else, but the hard part should fall to the individual.”

This feature will be accessible in Substack’s app for any post, note, reply, or comment exceeding 100 characters. Substack will also permit its writers to incorporate an optional AI author note, allowing creators to disclose their use of AI, as informed to TechCrunch by the company.

The corporation clarified that this tool is not intended to hinder or punish AI-assisted writing, but rather to motivate writers to add a “how I create this” explanation, detailing their methods.

Additionally, publishers can run Pangram on their own drafts prior to publication and can report and remove scans of their work that they believe contain errors.