Save history with AI. Advances in nerve regeneration. Trick an AI into believing a fake disease. Find out all that and more in this week’s The Prototype. Find out all that and more in this week’s The Prototype. To receive it in your inbox, Sign up here.
aall over the countryhistorical societies, libraries, and universities have boxes of archival materials waiting to be reviewed and cataloged. It can take an archivist more than an hour to process the items in one box. As a result, some institutions have backlogs stretching back months, even years, and important parts of our history remain unknown to this day.
Dean Serrentino and his company Historiq aim to solve this bottleneck with Una, an AI platform designed to accelerate archiving. Instead of painstakingly cataloging items “with a clipboard and a pencil,” archivists will be able to verbally describe their observations as they examine the materials, Celentino said. These notes are then connected to the institution’s data system. Additionally, all found documents can be digitized simultaneously using a laptop or mobile phone camera.
This speeds up the cataloging process and improves accuracy. When archivists don’t have to write everything down, they “bring more context and historical description from their brains into the system,” he says. He added that AI only produces drafts that need to be reviewed and approved to ensure the human element is present.
Historiq was founded in 2025 with a $1.25 million investment from Rob Waldron, chairman of edtech company Curriculum Associates. Customers already include several institutions. One of these is Fort Ticonderoga, a Revolutionary War site, which will use Una to digitize its collection of rare books later this year.
This week’s discovery: Trick AI into believing in a fake disease
a A team of scientists invented the condition Dubbed “bixonimania”, it published two fake papers about it on a preprint server in early 2024 and posted several posts on Medium. The paper was published under the name of a made-up researcher named Razliv Izgubrjenović, who is said to be working at Asteria Horizon University, a fictional university located in Nova City, California.
According to natureleading AI chatbots began reporting on “diseases” as if they were real, and fake studies began to be cited in peer-reviewed papers, suggesting that scientists were relying on AI for background research when preparing their own studies.
The large language model incorporated these papers into the data, despite intentionally including multiple flags such as the phrase “This entire paper is a hoax,” and claimed that its “research” was funded by the “Professor Sideshow Bob Foundation for Research in Advanced Tricks.”
According to the researchers, Google Gemini, Microsoft Copilot, Perplexity AI, and OpenAI’s ChatGPT all quickly provided information about fake illnesses. That said, more recent versions of the model have had a more nuanced interpretation of condition.
The purpose of the experiment was to see how easily the AI chatbot uses misinformation as the basis for legitimate health advice to users. This builds on previous research by team leader Almira Osmanovic Thunström, who found that LLMs are more likely to accept misinformation presented in academic papers.
This experiment is reminiscent of a more informal one conducted by BBC technology journalist Thomas Germain. With just one well-crafted blog post, he was able to get a leading AI chatbot to report that he was a hot dog eating champion. But both of these examples highlight that even artificial intelligence is not immune to one of the golden laws of computing: garbage in, garbage out.
Nervgen Advances Spinal Cord Injury Treatment
LAround DecemberI spoke to Adam Rogers. His company, NerfGen, develops drugs to help heal injured spinal cords. In a small clinical trial, this treatment improved hand function, lower body mobility, and upper limb strength in patients with spinal cord injuries. Now the company is taking the next step. The company has just reached an agreement with the FDA to plan a larger clinical trial, this time involving 150 patients at up to 60 different sites in the U.S. and Canada.
The new study will begin this summer and hopefully conclude in the second half of 2027, Rogers told me. If this large trial shows similar initial results, the company hopes to begin the FDA approval process for its drug in 2028. Meanwhile, the company plans to release more data this year in a study on how well its drug improves outcomes for spinal cord patients, and is also considering announcing clinical trials for other diseases where its nerve repair treatments may help.
Featured opinion: We should talk about generative science.
Every week, I ask investors for their thoughts on technology trends within their industry. Today, we would like to introduce your thoughts. james joaquinCo-founder and Managing Partner of Obvious Ventures. He aims to make early investments in breakthrough technologies such as advanced materials, artificial intelligence and plant-based proteins.
james joaquin
obvious ventures
What technologies are being overhyped right now?
humanoid robot. I think the technology surrounding bipedal humanoid robots is overhyped. The real unsolved problem in robotics is manual dexterity and fine motor skills. Solving this would open up a multi-trillion dollar robot market without requiring robots to have legs or mobility. Building a superhuman hand will be a ChatGPT moment for robotics.
What technology should more people be talking about today?
generative science. Let’s talk more about the positive impact AI is having on human health. Since the Nobel Prize-winning work of AlphaFold, modern AI has led to new advances in health and medicine. Hospitals are using AI to identify risks, detect infections early, and save lives in the process. Researchers are leveraging AI co-scientists and open-source biomolecular interaction models like Boltz to dramatically accelerate the development of new treatments.
What technologies will we be talking about in five years?
personal AI. In five years, each of us will have our own personal AI assistant that we can trust. We will talk to human doctors about health insights and diagnoses uncovered by AI, talk to human colleagues about business strategies co-created by AI, and hopefully spend more time with our loved ones thanks to tasks outsourced to AI.
on my radar
Anthropic vs Pentagon, CTD: Last month, a federal judge blocked the Department of Defense from designating human-related issues as a supply chain risk. But this week, an appeals court overturned that decision. This means that defense contractors cannot use Anthropic’s software in any work they do for the federal government, at least while the case is heard in court.
Spatial imaging movement: This week’s many announcements highlight that the business of Earth imaging is still alive and well. Geospatial data company Xoople has announced a partnership with L3Harris to develop a satellite constellation with optimized data output for AI models. Meanwhile, video company Vantor announced that it is building a constellation of satellites capable of both high-resolution footage and continuous surveillance. Focusing on one usually means trading off the capabilities of the other, which is a difficult prospect. Finally, Firefly Aerospace announced a deal with Nvidia to provide processing power to Ocula, a project to develop a commercial lunar imaging and mapping system.
What are you looking forward to this week?
Reviving a sitcom is always a risky prospect, and it’s difficult to recapture the old magic and chemistry that made the original. So it’s pretty impressive scrub I was able to accomplish that. The original main cast are likely returning to their old roles, playing older and slightly wiser versions of themselves, but the new additions are great. Joel Kim Booster in particular steals every scene he’s in as Dr. Kevin Park. If you’re a fan of the show, you won’t be disappointed.
