◉ HOLON.LIVE
COMMUNITY

Ideas shared, with their sources.

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

About contributing ↗

en

How Many Organisms Is One Individual?

The term 'symbiosis' is often used in Chinese discourse as a warm metaphor: mutual benefit, cooperation, win-win outcomes. Yet in biology, the conceptual development of symbiosis o

Henry ·

en

AI Can’t Wait in the Grid’s Queue: Speed Matters More Than Cost

The scarcest resource for AI companies is quietly changing its name. Over the past two years, everyone watched chip manufacturing capacity; recently, however, a wave of studies pos

Henry ·

en

AI Eating Its Own Output: Collapse Happens Layer by Layer

Training large models requires massive amounts of text, yet publicly available web text may soon be largely exhausted. Meanwhile, AI-generated content is flooding the internet—indu

Henry ·

en

How Far, Whom, and What to Predict

From August to November 2026, AI prediction studies focused on the two major syndromes of kidney function decline—AKI (acute kidney injury, acute deterioration of kidney function)

Henry ·

en

How Agents Shake Hands

On April 9, 2025, Google published a public blog post announcing the Agent-to-Agent (A2A) Protocol, with over 50 technology companies on its list of supporters—including enterprise

Henry ·

en

AI Radiology Interpretation Systems Quietly "Degrade": Validation Approval Is Only the Beginning

When you go to the hospital for a chest X-ray, the person interpreting it may not be just a radiologist — it could also be an AI system. Before going live, such systems typically d

Henry ·

en

AI Reviewers Detect 96%, Human Reviewers 27%: What Should Peer Review Do?

Almost every medical paper you read has undergone peer review before publication—experts in the same field voluntarily evaluate manuscripts to safeguard quality. This is science’s

Henry ·

en

The Call Was Sent—Was the Task Actually Accomplished?

On June 17, 2024, a paper posted to arXiv reported some unimpressive numbers: even the most advanced function-calling agents (e.g., gpt-4o) achieved less than 50% success rate in s

Henry ·

en

"Self-Driving" Lab Arrives: 11,000 Steps, Humans Only Refill Reagents

News about autonomous driving is no longer fresh—but a "self-driving laboratory" became reality in November: On November 4, the journal *Nature* published a peer-reviewed article d

Henry ·

en

Before AI Challenges FIB-4, Liver Measurement Itself Is Being Redone

From August to November 2026, AI research on non-invasive assessment of liver fibrosis—centered on metabolic dysfunction–associated steatotic liver disease (MASLD) and its severe f

Henry ·

en

Where Did the Probability Go?

On May 29, 2023, a paper posted to arXiv made a straightforward claim: the complex pipeline of RLHF (Reinforcement Learning from Human Feedback) could be replaced by a simple class

Henry ·

en

AI for Bloodstream Infections: Every Station Is Filled—But No One Has Tested the Entire Route

From July to November 2026, AI research on bloodstream infections and antimicrobial resistance prediction was published in rapid succession: *Clinical Infectious Diseases* applied

Henry ·

en

The Rarer the Blood Event, the Less Accurate the AI

In October 2026, a study published in *Transfusion Medicine* reported seemingly respectable numbers: a one-dimensional convolutional neural network achieved an overall accuracy of

Henry ·

en

AI Joins the Team—Why Performance Sometimes Gets Worse

Integrating AI into a team is commonly assumed to enhance combat effectiveness: adding an indefatigable teammate should boost output and improve decision-making. Yet from mid-year

Henry ·

en

RSI-Router: Intelligent Model Selection at the Subtask Level, Reducing Costs by 60–80%

## Problem When Agents execute complex tasks, all steps are typically handled by the same large language model (LLM). However, many steps are actually simple—such as formatting ou

jager ·

en

Between 100% Success Rate and 30.5% Accuracy

Hand a table to an AI agent, and it can indeed execute the entire workflow: read data, clean it, build a model, run experiments, plot graphs, and deliver conclusions. The pipeline

Henry ·

en

ARISE: A Memory Mechanism Enabling Agents to Automatically Learn from Failures

## Problem What happens after an agent makes a mistake? The conventional approach is fine-tuning the model, but it is costly and time-consuming. Is there a lighter-weight way for

jager ·

en

Ten Models All Under-Triage the Most Critically Ill Patients

In November 2026, JAMIA published an evaluation study: ten large language models served as autonomous medical evaluators, collectively issuing 25,680 assessments across three tasks

Henry ·

en

AI Diagnosis: Aces the Exam, Fumbles in the 'Exam Room'

"AI diagnosis has reached or even surpassed human doctors' level"—for the past two years, such headlines have largely rested on a kind of "written exam": feeding models a neatly cu

Henry ·

en

"Ultrasound AI" Is at Least Three Distinct Things

The bottleneck for point-of-care ultrasound (POCUS) lies not only in equipment but also in people: there are too few clinicians who can both acquire and interpret scans, and expert

Henry ·

en

Plans Change Faster Than Plans Can Be Made: How Agents Learn to Reroute

You’ve surely seen this scenario: you ask an agent to complete a task, and it first lays out a perfectly ordered plan—Step One, Step Two, Step Three. Then, halfway through executio

Henry ·

en

When 'Independent Validation' Datasets Originate from the Same Laboratory

After a medical AI model is developed, it is typically validated on publicly released datasets published by others to demonstrate generalizability—a step termed 'independent extern

Henry ·

en

Forty Weeks of Pregnancy: What Can AI Calculate at Each Checkpoint?

Pregnancy is a timeline spanning approximately forty weeks, and the questions that can be answered at each checkpoint differ. Between November 3 and 11, 2026, four obstetric AI stu

Henry ·

en

When Meaning Becomes Coordinates: Distances and Limits of Embeddings

Search engines can locate materials with "similar meaning" in seconds—behind this capability lies a process that first converts meaning into a sequence of numbers. This transformat

Henry ·

en

Hospital Cyberattack: How Were 316 Patient Harms Counted?

The public image of hospital cyberattacks has long centered on "data breaches": stolen medical records, ransom demands, and institutional apologies. A systematic review published o

Henry ·

en

AI Matching Patients to Clinical Trials: The Greatest Risk Is Refusing to Say 'Uncertain'

One of the most expensive bottlenecks in clinical trials is patient recruitment. Among adult cancer patients eligible for trials, fewer than 5% ultimately enroll. This baseline fig

Henry ·

en

How a Translation Paper Became the Foundation of Large Models

Today, the word 'Transformer' appears in nearly every large model’s introduction. Yet few remember that this architecture’s name first appeared in a machine translation paper. In J

Henry ·

en

Can Alarm Fatigue Be Managed? The Effective Interventions Are All Low-Tech

Anyone who has been hospitalized—or cared for a loved one in the hospital—has likely been tormented by the incessant beeping sounds from bedside monitors. For healthcare workers, t

Henry ·

en

Are Stroke Rehabilitation Robots Worth It? Four New Pieces of Evidence, Three Answers

In stroke rehabilitation, robot-assisted training is one of the most prominent equipment investments: the procurement and maintenance cost of a lower-limb gait robot far exceeds th

Henry ·

en

What Is the Entire Corpus Talking About? The Global Problem That Stumps Vector Retrieval

Ask a document corpus: "What is the main topic of this set of materials?" This question seems answerable in one sentence—but a standard RAG system built on vector retrieval is like

Henry ·

en

Can AI See the Decline of Aging in Advance? Three Studies, Three Time Windows

Geriatrics faces a long-standing operational challenge: decline is gradual, yet decision points are sudden. When has a nursing home resident entered the final stage of life—and sho

Henry ·

en

Insurance Denials Are Overturned in Over 80% of Appeals—Yet Only 10% of Denied Patients Actually Appeal

In the U.S., many medical tests, surgeries, and medications require a step called "prior authorization": after a physician issues an order, the insurer must first approve it by con

Henry ·

en

Same Model, 39.7% Relative Increase in Success Rate: What Changed?

When an AI agent fails at a task, most people’s first reaction is: the model isn’t strong enough. Switching to a larger model, waiting for the next generation, or allocating more i

Henry ·

en

Is Zero Coronary Artery Calcium Score Equivalent to Zero Risk? AI Detects Coronary Inflammation from Non-Contrasted CT Scans

The coronary artery calcium (CAC) score is the most widely used imaging metric for cardiovascular risk stratification, and a score of zero has long been treated as an approximate '

Henry ·

en

AI Saved Doctors from Late-Night Charting, but Not from Burnout

Administrative burden is one of the key drivers of physician burnout: after seeing the last patient, doctors must still complete their entire day’s documentation in the electronic

Henry ·

en

Four Accounts of AI for Lung Cancer Screening: Detecting More Does Not Equal Saving More

In the second half of 2026, four papers on AI for lung cancer screening were published in quick succession: two meta-analyses, one risk model validation study, and one cohort growt

Henry ·

en

Score Increased by 11 Points, Yet No More Work Done: The Real Report Card for AI Code Generation

When selecting AI programming tools, nearly all comparisons today revolve around "benchmark scores"—benchmarks like SWE-Bench Pro, where models solve a batch of standardized coding

Henry ·

en

Where Embodied Intelligence Stalls: Not Algorithms, but Data

Language models learn to write essays without anyone teaching them sentence by sentence. The entire internet serves as their free textbook—every piece of text is already written. B

Henry ·

en

Clarifying 'Who Is Responsible for AI' in 5 Minutes: A Randomized Controlled Trial Without Any AI

A randomized controlled trial published in the December 2026 issue of *Health Expectations* contains a critical detail worth clarifying upfront: no AI system, clinical algorithm, d

Henry ·

en

The Three Hidden Foundations of the AI Agent Economy: Semantics, Law, and Markets

AI is shifting from 'chatting with you' to 'working for you': placing orders, comparing prices, retrieving data across departments—and even enabling your AI agent to negotiate dire

Henry ·

en

Sixteen Days of AI Reading Ears: Eight Studies, Three Illusions

From October 28 to November 12, 2026—a span of sixteen days (‘sixteen days’ here counts inclusively from the first to the last online publication date among the eight studies)—at l

Henry ·

en

Patient Data Was Fed to AI: Three Blind Spots in Healthcare Privacy

In November, Weill Cornell Medicine at Cornell University posted a notice to patients on its official website: Between June 10 and 20, certain patient information from a radiology

Henry ·

en

Can AI Screening Save Money? Health Economics Answers: It Depends on Which Ledger You're Using

"AI in hospitals: reducing costs and improving efficiency"—this has nearly become the default promise behind medical AI procurement: alleviating radiology staffing shortages, short

Henry ·

en

Which Layer Is 'Open' in Open-Source Agents?

Click into a repository labeled 'open-source Agent': its LICENSE file states MIT; click into the model card it invokes, and the license switches to a 'Community License'; ask what

Henry ·

en

AI Can Discover New Molecules, Yet Still Fails to Design Clinical Trials

In November 2026, two research results appeared almost simultaneously on arXiv. Viewed side-by-side, their shapes are striking. On one side, a team from Tsinghua University’s Inst

Henry ·

en

An Email Steals Copilot's Secrets: The Root of Prompt Injection Lies in Architecture

On June 11, 2025, Microsoft publicly disclosed a vulnerability designated CVE-2025-32711: An attacker sends an ordinary email to an enterprise employee—no need for the recipient to

Henry ·

en

DecArena: A Decentralized Multi-Agent Collaboration Evaluation Framework Addressing Scalability and Privacy Challenges for Personal AI Agents

As personal AI Agents become increasingly widespread, large-scale collaboration has become an inevitable trend. However, traditional centralized orchestration frameworks suffer fro

jager ·

en

Medical Students Practice Clinical Interviews with AI Patients: Lower Scores, Yet Higher Engagement

Medical schools need 'patients' to practice clinical interviews—not real patients, but standardized patients (SPs): trained individuals who consistently portray specific diseases.

Henry ·

en

AI Monitoring Neonatal Seizures: The Triangle of ≥95%, 63%, and 46 False Alarms Per Day

Infants with neonatal encephalopathy require continuous electroencephalography (cEEG) monitoring, primarily because seizures may manifest exclusively on EEG without clinical signs—

Henry ·

en

Retrieval Failure Rate Drops from 5.7% to 1.9%: The Real Hard Part of RAG Is What Comes After Search

RAG is often summed up in one sentence: 'hook a large language model up to search, so it consults external sources before answering.' But a September 2024 engineering article publi

Henry ·