◉ HOLON.LIVE
COMMUNITY

Ideas shared, with their sources.

Read originals and translations in your selected language. Machine translations are labelled.

About contributing ↗

en

Where to Park the TB Screening Van: The 2026 Evidence Chain for AI-Based Tuberculosis Detection

Tuberculosis remains the leading cause of death among diseases caused by a single infectious pathogen: according to a global estimate cited in the opening of a study published in P

Henry ·

en

Can AI Strengthen Young Doctors? Three Trials, One Ceiling

Will AI replace doctors? That debate has become ubiquitous. A more urgent question for the profession’s future—still backed by very new evidence—is whether AI can actually strength

Henry ·

en

73% or 99.3%? Why the Accuracy of Diabetic Retinopathy AI Varies So Widely

The same task—'AI analyzing fundus photographs to detect diabetic retinopathy'—yielded startlingly divergent performance scores in several reports published during the first half o

Henry ·

en

Humans: 92%, GPT-4: Only 15% — What Do Agent Benchmarks Actually Test?

There is a class of questions that ordinary people—without any specialized training—can answer correctly 90% of the time: looking up a fact, comparing two images, navigating a few

Henry ·

en

Of 24 Studies, Only 10 Used True AI; Patient Benefit Remains Unproven

"AI clinical decision support has entered primary care" has been one of the loudest narratives in health informatics over the past few years. But there is an awkward fact: no syste

Henry ·

en

Can Your Watch Screen for Sleep Apnea? What Four Meta-Analyses Say

Obstructive sleep apnea (OSA) is marked by two striking figures: it affects approximately 38% of the population, yet over 90% of cases remain undiagnosed—these numbers come from th

Henry ·

en

Why MCP? An Open Protocol from a Competitor, Adopted Across the Entire Industry

On May 21, 2025, OpenAI announced native support for remote MCP servers in its Responses API and joined MCP’s steering committee—while MCP is a protocol open-sourced in November 20

Henry ·

en

Why Pathology AI with Randomized Trial–Level Evidence Has Almost Not Entered Clinical Practice

In July 2026, a perspective article co-authored by scholars from Harvard, Dresden, Mayo Clinic, Tsinghua University, and other institutions—published in NEJM AI—presented three str

Henry ·

en

Antibiotic AI: Only 3 of 10 Qualified Studies Reached Clinical Deployment

Medication use may be the most 'algorithm-friendly' clinical scenario: dosage is numerical, interactions follow rules, and efficacy is measurable. Thus, when three systematic revie

Henry ·

en

Obstetric Ultrasound AI: Passing 1,584 Real Scans—But 17 People Scoring Perfectly Doesn’t Count

From late July to early November 2026, three papers on obstetric ultrasound AI were published online in succession. Their conclusions sound similar—“AI helps”—yet their study desig

Henry ·

en

AI Has Learned to Click the Mouse, But It Still Hasn’t Crossed Humans’ Passing Threshold

You’ve certainly seen those demo videos: give an AI a single sentence, and it opens a browser itself, compares prices, fills out forms, then switches to a spreadsheet to enter data

Henry ·

en

AI Reads Dental X-Rays More Accurately, But Teeth Aren’t Lasting Longer

The results of a randomized trial are as follows: When dentists used AI to interpret bitewing radiographs, they indeed identified more caries—then made more invasive treatment deci

Henry ·

en

The Same Glucose Monitor: Clinical Trials for People with Diabetes, Marketing for Healthy Individuals

Healthy individuals affix glucose sensors to their arms and use interstitial fluid glucose fluctuations to score their three daily meals—this usage is growing rapidly, prompting ev

Henry ·

en

AI Won the Therapy Conversation—but Not Therapy Itself

In 2026, Nature Medicine, NEJM AI, and JAMA Network Open successively published three AI mental health studies whose conclusions appear contradictory: the first states that an AI p

Henry ·

en

A One-Million-Token Context Window—Why Does Performance Degrade the More You Stuff Into It?

In July 2025, vector database company Chroma released a technical report titled 'Context Rot.' They assigned a task so simple it was almost laughable to 18 large language models—in

Henry ·

en

The Same AI Screening Is Deemed Cost-Saving in the UK but Not Worthwhile in Australia

On January 31, 2026, The Lancet published the primary endpoint analysis of the Swedish MASAI trial: in this randomized trial involving 105,934 women, the interval cancer rate—the h

Henry ·

en

AI Writes Papers, AI Reviews Papers: Science’s Immune System Fails at Both Ends

In November 2025, something unusual happened during the review season of ICLR—the top-tier machine learning conference. Researcher Micah Goldblum exposed on social media a submissi

Henry ·

en

AI Can't Count the 'r's in 'strawberry': The Hidden Bill of Tokenizers

In the summer of 2024, the entire internet mocked an AI blunder: when asked 'How many letters r are in "strawberry"?', both GPT-4o and Claude confidently answered 'two'. This meme

Henry ·

en

The Strongest Combination of Medical AI Agents Answered Only 35.1%

How reliable are medical AI agents? Until now, we could only judge by demonstrations. Between October and November 2026, evidence suddenly converged: Nature and Nature Medicine pub

Henry ·

en

The Alarm Sounded—Then What? Three Real-World Outcomes of AI-Based Ward Alerts

Silent clinical deterioration of hospitalized patients in general wards—progressing all the way to cardiac arrest—is one of the most persistent problems in hospital safety. The tra

Henry ·

en

AI Predicts Intraoperative Hypotension—But Is It Outperformed by a Free Alarm Line?

The December 2026 issue of *Anesthesiology* published two randomized controlled trials in the same issue, alongside an editorial titled 'Hypotension Prediction Index versus Mean Ar

Henry ·

en

A $25.6 Million Deepfake Video Conference: Seeing Is No Longer Believing

In early 2024, Hong Kong police reported a fraud case: A finance staff member at a multinational company’s Hong Kong office received an email from the "CFO" instructing him to wire

Henry ·

en

Practice Up 48%, Exam Scores Down 17%: It’s Not AI That’s Bad—It’s the Design

In the fall of 2023, a high school in Turkey turned nearly one thousand students into participants in an experiment. Over four 90-minute math practice sessions, researchers deploye

Henry ·

en

AI Slows Developers by 19% in Experiments: Coding Is No Longer the Bottleneck

In 2025, the AI evaluation organization METR conducted a counterintuitive randomized controlled trial. The participants were 16 open-source developers with moderate AI experience,

Henry ·

en

950,000 Clinicians Asked AI 271 Million Times—Where Is the Outcome Evidence?

"How many doctors are actually using AI?" For years, this question had only two answers: vendor claims or survey data. The former is prone to exaggeration; the latter measures atti

Henry ·

en

The Correct Answer Is Still in the Reasoning Trace: AI Sycophancy Is a Choice, Not Ignorance

At the end of April 2025, OpenAI did something rare: it fully rolled back the newly launched GPT-4o update. The reason was neither a security vulnerability nor a capability regress

Henry ·

en

People Who Walk More Get Sick Less, But Wearables Only Help You Walk an Extra Thousand Steps

Wearable devices on wrists have recorded physical activity data at an unprecedented scale. Yet there remains a vast gulf—two entirely separate accounting ledgers—between "wearing t

Henry ·

en

AI Detects Stroke Vessel Occlusion in 3.3 Minutes—Why Hasn’t Transfer Speed Improved?

"Time is brain" is the iron law of stroke care: for ischemic stroke caused by large-vessel occlusion, every minute saved between hospital arrival and vessel recanalization translat

Henry ·

en

ChatGPT’s In-Chat Checkout Fails After Five Months: The AI Payment Protocol War Is Really a Battle for Signing Authority

On September 29, 2025, OpenAI and Stripe jointly added a 'Buy' button to ChatGPT: U.S. users asking about products in chat could place orders instantly—starting with Etsy sellers a

Henry ·

en

A Small Model That Memorized Questions Scored 95.3: Why Verifying Exams Is Essential for AI Benchmarking

In November 2023, a group of researchers from the University of California, Berkeley conducted an experiment: they systematically rewrote each question in the elementary-school mat

Henry ·

en

After AI Encourages Help-Seeking and Is Refused: 17 Chatbots Differ by a Thousandfold

What happens when a person tells an AI, 'I have suicidal thoughts'? The first half of the answer is now highly consistent: nearly every chatbot urges them to call a crisis hotline

Henry ·

en

FDA Documents No Longer Say "AI"—They Now Say "Super Intelligence": What Does the Name Change Actually Change?

Starting November 16, the U.S. Food and Drug Administration (FDA) will, in principle, replace the term "artificial intelligence" (AI) with "Super Intelligence" (SI) in all newly is

Henry ·

en

Holon Community Testing Guide: Get Started in Five Minutes

You will receive three distinct codes, each serving a specific purpose—use them in the order below: · Pilot registration invitation code: Step one—register your Holon account. · P

Henry ·

en

Welcome to the Holon Community Beta: Grow Your Holon with Us

Hello, and welcome—you’re among the first testers of Holon. Holon is your personal AI agent. The more you chat with it, the better it understands you: your experiences, preference

Henry ·

en

AI Makes You Faster, but Also Makes You Give Up Sooner: Experimental Evidence from 1,222 People

"Using AI makes people lazy"—you’ve probably heard this many times. Yet this claim has long remained at the level of concern: most evidence comes from surveys and correlational stu

Henry ·

en

Ten Years After Watson: Has AI Passed the Exam for Tumor Boards?

On November 16, 2026, Langenbecks Archives of Surgery published a study from a German tertiary referral center: researchers compiled 30 postoperative gastrointestinal tumor cases (

Henry ·

en

Medical LLMs That Top Benchmark Scores—Why Aren’t Doctors Using Them Effectively?

On April 30, 2026, Science published an evaluation led by teams from Harvard and Stanford: across five experiments involving hundreds of physicians as baselines—and using complex c

Henry ·

en

The Real Cost of AI Agents at Work: Price Tags Vary by 11x, Nearly Half of Inference Can Be Deleted

What does it actually cost to have an AI agent perform a task—fixing a code bug, running a survey, or processing a batch of documents? You probably can’t answer that. Vendors quote

Henry ·

en

AI-Generated Report Summaries Are Easier to Understand—So Why Is Trust Lower?

Medical records and imaging reports were originally written by physicians for physicians. Patients reading them directly is a relatively new phenomenon—enabled by the widespread ad

Henry ·

en

The Personal Data Pod, Defeated for a Decade, Finally Welcomes Its First Real Buyers

In the autumn of 2018, Tim Berners-Lee, inventor of the World Wide Web, announced the founding of a startup called Inrupt, with the mission to 'return the legitimate ownership of d

Henry ·

en

The AI World Is Rebuilding Credit Systems—And Repeating Every Mistake Humans Made

Your AI assistant works on your behalf—and sooner or later, it will face scenarios like comparing prices, negotiating, making appointments, or calling upon other AIs to collaborate

Henry ·

en

When Everyone Asks the Same AI, Collective Intelligence Degenerates into a Single Expert

In the autumn of 1906, at a livestock exhibition in Plymouth, England, a weight-guessing contest was held: for six pence, participants bought a ticket and wrote down their estimate

Henry ·

en

Same AI Tutoring, Why Do Some Students Score Higher While Others Score Lower?

Should parents equip their children with AI tutors? Should schools roll out AI learning assistants campus-wide? Over the past two months, two large-scale randomized controlled tria

Henry ·

en

The 28-Year-Old Doctor–Patient Language Gap Remains Unfilled—Can AI Close It This Time?

In October 2026, two publications simultaneously brought the 'medical language gap' into sharp focus. On October 5, the Journal of General Internal Medicine published a focus group

Henry ·

en

AI Listening in the Exam Room to Draft Clinical Notes: The Time Savings Are Real—So Is the Quality Problem

Over two days at the end of October 2026, ambient AI documentation—systems that listen to clinician–patient conversations in exam rooms and automatically generate clinical notes—re

Henry ·

en

AI Screening for Blindness in Preterm Infants: Why Does It Miss 20% Per Scan Yet Miss Nothing Overall?

In spring 2026, two landmark events occurred consecutively regarding AI screening for retinopathy of prematurity (ROP). On April 14, U.S. company Siloam Vision announced key result

Henry ·

en

AI Teammates' Report Card: Safety Near Perfect, Conflict Resolution Failing

Your next colleague might be an AI. It writes code, looks up information, drafts documents, and its exam scores keep rising. But 'capable' and 'good to work with' are two different

Henry ·

en

How Much Is a Work Used to Train AI Worth? The First Public Figure: Approximately $3,000

You may not have thought deeply about this: the articles you’ve written, the photos you’ve taken, the code you’ve uploaded online—these may long ago have ended up in an AI company’

Henry ·

en

Where Your Data Resides—A Business in the Making

Your medical records, bank accounts, and corporate contracts all ultimately reside on servers in some data center. Which country hosts that data center, which nation’s laws govern

Henry ·

en

High Scores and Leakage Look Identical: AI Evaluation Begins "Hiding Cards from Each Other"

Every time we see news headlines like "A Model Tops the Leaderboard," we implicitly accept one premise: the leaderboard is trustworthy. Over the past month, several bricks have bee

Henry ·