Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about how AI is making research papers talk to each other, AI-powered digital twins, and more.
AI in Action: Making the papers talk
Scientific journals have worked the same way since the 1600s: someone writes findings down, and someone else reads them. A team at Stanford Medicine wants those pages to answer back. Led by postdoctoral scholar Jiacheng Miao and biomedical data scientist James Zou, the group built a system called Paper2Agent that converts a published paper, along with its figures, code and data, into an AI agent. Ask it about the study and it answers; hand it fresh data and it runs the paper's methods on it. It can also hold a conversation with other paper agents, which is where things get productive.
Building one takes more than reading the PDF. A crew of worker agents tries to reproduce the research from scratch in a virtual environment, picking up the practical know-how (reagents, setup, execution steps) that readers usually have to dig out on their own. That knowledge gets organized through the Model Context Protocol, with each section of the paper filed where other agents can reach it. The human authors still have a role: the agent interviews them about failed experiments and judgment calls that never made it into print. The work was published in Nature on September 16.
For the demo, the team paired two papers with nothing in common. One described a tool that predicts how genetic mutations affect the genome; the other was a genome-wide study of ADHD risk. The prediction agent applied its method to the ADHD dataset and flagged a variant near a gene called MPHOSPH9 associated with higher risk, a connection Zou says hadn't been reported before. Normally, two labs would have to stumble across each other's work for that to happen. The team has built more than 100 paper agents so far. Zou's longer-term picture is millions of them finding overlap on their own, with any discoveries credited back to the original human authors.
🔥 Rapid Fire
Analysis: Where Are All The AI Chips?
Analysis: The Hater’s Guide to AI Debt (Part 1)
Oracle invokes force majeure over troubled data center
“Oracle is moving to shield itself from racking up expenses on a massive data center being built in New Mexico, adding a fresh wrinkle to a project beset by opposition and regulatory setbacks.”
$18 billion of debt tied to this data center slid into stressed territory
SoftBank’s data center company SB Energy delays IPO
Anthropic delays IPO again to November
Addendum: Anthropic and OpenAI are not delaying their IPOs because of safety. They are delaying their IPOs because their finances are wretched.
Investors are concerned about Anthropic's growth prospects
OpenAI expects to burn $280 billion by 2030
OpenAI is assuming $840 billion in revenue by the end of 2030
To be clear, this will never happen
Tracing the origins of AI doomerism
Podcast: Stopping The AI Safety Cult
Hallucinating LLM (aka bad software) almost starts war with China
“The intelligence report, circulated across the US military this spring in the midst of the war with Iran, immediately set off alarm bells: A Chinese ship in the Middle East was transporting components of a nuclear weapons program [...] It was only just before the planned operation that officials dug deeper into the report put together by a special operations command analyst and found it had been generated with the help of artificial intelligence (AI) — and that a chatbot the analyst had used inaccurately identified the material the ship was carrying.”
Shopify CEO and others backtrack on AI enthusiasm
OpenAI fires contractors for using AI to train AI
Most founders are one system away from turning LinkedIn into their best sales channel.
Engagement is easy to mistake for pipeline. On Sep 30, watch how a founder turns LinkedIn content into real outreach. Live. You'll walk away with a repeatable system: what to post, who to reach out to, and how to sequence it.
Eligible startups also get the LinkedIn-to-Leads Toolkit: ad credits, Apollo, Captions, and HubSpot's Prospecting Agent.
📖 What We’re Reading
A digital twin used to mean a detailed 3D replica that updated slowly and told you what had already happened. That definition no longer holds. From 2025 into 2026, digital twins have shifted from passive visualization tools to systems that predict, reason, and, in some cases, act on behalf of the physical assets they represent.
The change didn't come from better graphics or faster sensors alone. It came from embedding artificial intelligence and machine learning directly into how a twin interprets data and makes decisions.
