Kuyun AI · Works to try

AI & machine learning

90 of the works published on Kuyun. Sign in or sign up to open any of them.

Verifying Tool Calls: Cutting Hallucinated and Misused Tool Use in Agents

Agents fail less by reasoning badly than by calling tools badly. How often, in which ways, and how much does a verify-and-retry loop fix?

article · by @goodai

German A2 role-play — Getting by

View a room, land a café job, return a jacket without the receipt and get three tickets to Hamburg — in everyday German.

practice · by @berlin_lernen

Post-training: SFT, RLHF and DPO

SFT, reward models, PPO and DPO, and how to know it worked, in five senior post-training rounds.

practice · by @interview_room

Does More Data Help? Scaling Laws for Generalist Manipulation Policies

VLA policies improve with data. But along which axis — quantity, diversity, embodiments — and do the gains transfer to new tasks, or just m…

article · by @science_geek

From Signal to Action: Wearable Early Detection That Earns Its Alerts

Wearables can flag disease early. But do the alerts come early enough, rarely enough, and lead to action that actually helps?

article · by @meds_health

The Time Machine

H. G. (Herbert George) Wells's The Time Machine, 17 chapters, read aloud.

book · by @gothic-shelf

LLM Model and Training: From Tokens to Aligned Assistants

From the token up: twelve lessons, a midterm and a final, each with worked explanations and practice.

course · by @tech-courses

The research scientist round

Present your paper, defend your choices, critique a claim and plan the next experiment on a budget.

practice · by @interview_room

Feature and data pipelines

Skew, backfills, feature stores and data quality, five rounds against the clock

practice · by @interview_room

Walk and Work: Whole-Body Control That Unifies Locomotion and Manipulation

Humanoids usually walk and manipulate with separate controllers. Does unifying them into one whole-body policy actually help, and what does…

article · by @science_geek

Grounding Computer-Use Agents: Robust Screen Reading and Safe Actions

GUI agents fail most at grounding — finding and acting on the right element. Which representation grounds best, and how do we make risky ac…

article · by @research-notes

AI-Era Startup Opportunities Research

Judge where to build around the foundation-model layer by weighing evals, agent operations and agent identity against 28 sourced data point…

article · by @goodai

Memory Consolidation for Long-Horizon Agents

Does summarizing, promoting, and forgetting beat flat retrieval for agents that run for a long time?

article · by @goodai

When Agent Teams Fail: A Taxonomy and Stress Test

At the same budget, when does a team of agents beat one strong agent — and how do teams fail when they don't?

article · by @research-notes

MLOps and production ML

Ship, watch, retrain and roll back, five production ML rounds against the clock

practice · by @interview_room

Epidemiology and Prevention of Vaccine-Preventable Diseases (The Pink Book)

How vaccines work, and the diseases they prevent — CDC's course book.

book · by @medicine-shelf

One Model of Many Brains? Testing Foundation Models of Neural Activity

Foundation models now pretrain on neural recordings. Do they transfer to new brains and tasks with less data, or just memorize datasets?

article · by @marie_curious

Spanish A2 role-play — Out and about

Check in, order dinner, buy a jacket and catch your flight — in polite, everyday Spanish.

practice · by @amigos

Pretraining data

Where the tokens come from, how they are cleaned and mixed, and how to keep the benchmarks out, in five rounds.

practice · by @interview_room

Italian B2 role-play — Working in Italian

Negotiate your contract, win over the parents, calm a lender and hold your ground in a meeting, in Italian.

practice · by @ciao_bologna

French A2 role-play — Getting by

Face the prefecture, land a café job, return a coat and get tickets to the Alps — in polite, everyday French.

practice · by @french_toast

The Evolution of Modern Medicine

Medicine from ancient Egypt to the germ theory — Osler's Yale lectures.

book · by @medicine-shelf

Inference and serving

KV cache, batching, quantization and speculative decoding, with the numbers a senior inference engineer knows.

practice · by @interview_room

LLM application engineering

Five rounds on structured output, guardrails, latency, cost and fallbacks, run on the clock.

practice · by @interview_room

Cheap and Sharp? Where ML Weather Emulators Beat and Break Numerical Models

ML weather models rival numerical forecasts at a fraction of the cost. Do they hold up on extremes, rare events, and physical consistency?

article · by @science_geek

The problem of Cell 13

Jacques Futrelle's The problem of Cell 13, 3 chapters, read aloud.

book · by @mystery-shelf

[Classics 🎧] The Time Machine

An inventor rides to the year 802,701 and finds what humanity became.

book · by @classic_literature

Memory Consolidation for Long-Horizon Agents

Does summarizing, promoting, and forgetting beat flat retrieval for agents that run for a long time?

article · by @goodai

Benchmarks That Do Not Lie: Verifiable Long-Horizon Tasks for Agent Evaluation

Many agent benchmarks reward shortcuts, memorization, or a lenient judge. What do agents score on long tasks whose success a program checks…

article · by @goodai

Sam in Tokyo: stories N2

Five stories for every lesson of Sam's first working year in Tokyo, in only the words and grammar you know so far.

book · by @sakura_house

The Chinese Classics

James Legge's The Chinese Classics, 6 chapters, read aloud.

book · by @zhongwen-jingdian

The Time Machine

An inventor rides to the year 802,701 and finds what humanity became.

book · by @science-shelf

美式英语,从 A1 到 C2

六门课、每级一套角色扮演、一书架朗读的英文书——讲解全中文,每天免费。

collection · by @panda

On the Motion of the Heart and Blood in Animals

How we learned the blood circulates — Harvey's own account, 1628.

book · by @medicine-shelf

AI & ML mock interviews

Twenty interview tracks for AI and ML roles — LLM training, ML engineering, agents, the behavioural round — each a 45-minute round you talk…

collection · by @interview_room

El anacronópete; Viaje a China; Metempsicosis

Enrique Gaspar's El anacronópete; Viaje a China; Metempsicosis, 38 chapters, read aloud.

book · by @biblioteca-clasica

Agent evaluation and safety

Five rounds on prompt injection, sandboxes, red teaming, agent evals and a real incident, on the clock.

practice · by @interview_room

Sam in Tokyo: role-plays N2

Play each lesson's scene for real: the morning meeting, the client, the meeting you chair, the mistake, the bank and the promotion, in busi…

practice · by @sakura_house

Le Tour du monde en quatre-vingts jours

Jules Verne's Le Tour du monde en quatre-vingts jours, 38 chapters, read aloud.

book · by @bibliotheque-fr

Explaining your work

Five rounds on telling your own project well, to a newcomer, a sceptic, an interrupter and a VP.

practice · by @interview_room

ML system design

Design ranking and recommendation systems end to end, from requirements and labels to latency, A/B tests and failure.

practice · by @interview_room

Lucy in Beijing: stories HSK 1

Five short stories for every lesson of Lucy's first Beijing autumn, each using only the characters you know so far.

book · by @hutong_house

LLM Model and Training: From Tokens to Aligned Assistants

From the token up: twelve lessons, a midterm and a final, each with worked explanations and practice.

course · by @code_cook

Morning AI Briefing

Read one dated page a morning and know which AI news is real, sourced and dated, and which is hype.

briefing · by @code_cook

An Inquiry into the Causes and Effects of the Variolae Vaccinae

The first vaccine — Jenner's twenty-three cases, 1798.

book · by @medicine-shelf

Joe Strong and his wings of steel

Vance Barnum's Joe Strong and his wings of steel, 25 chapters, read aloud.

book · by @childrens-shelf

Generative AI and High School Learning

Weigh the field evidence on how unrestricted chatbots and hint-only tutors change what students practise, what they test without help, and…

article · by @tester

Memory Consolidation for Long-Horizon Agents

Does summarizing, promoting, and forgetting beat flat retrieval for agents that run for a long time?

article · by @research-notes

The Analects of Confucius (from the Chinese Classics)

552 BCE-480 BCE Confucius's The Analects of Confucius (from the Chinese Classics), 19 chapters, read aloud.

book · by @ideas-shelf

Benchmarks That Do Not Lie: Verifiable Long-Horizon Tasks for Agent Evaluation

Many agent benchmarks reward shortcuts, memorization, or a lenient judge. What do agents score on long tasks whose success a program checks…

article · by @research-notes

Evaluation and benchmarks

What a benchmark measures, how sure you can be, and the eval suite that decides a launch, in five rounds.

practice · by @interview_room

Machine Learning: Foundations You Can Build On

From data to a model that generalizes: twelve lessons, a midterm and a final, each with worked examples and practice.

course · by @code_cook

ML fundamentals

Overfitting, imbalanced metrics, leakage, causality and a Python coding round, answered like a senior.

practice · by @interview_room

Fine-tuning and adaptation

LoRA, QLoRA, forgetting, when to reach for RAG instead, and a legal-domain design round.

practice · by @interview_room

Does the AI Tutor Help? A Fair Classroom Experiment

Everyone says AI tutors help. Can a class run a fair experiment to find out whether it really improves learning, not just how students feel?

article · by @marie_curious

Grounding Computer-Use Agents: Robust Screen Reading and Safe Actions

GUI agents fail most at grounding — finding and acting on the right element. Which representation grounds best, and how do we make risky ac…

article · by @goodai

Chinese HSK 2 — Lucy Finds Her Feet

Lucy's first Beijing winter, from the first snow to the Lantern Festival, in twelve lessons.

course · by @hutong_house

Debugging deep learning

Flat losses, spikes, NaNs and a buggy training loop, five debugging rounds against the clock

practice · by @interview_room

Anatomy and Physiology

The human body, system by system — a full college course, read aloud.

book · by @medicine-shelf

Learning Spanish, from A1 to C2 with lots readings and practices

Six courses, role-plays for every level and a shelf of readers in Spanish, from first words to the classics — with an AI guide who knows wh…

collection · by @amigos

A Potential for Everything? Stress-Testing Universal Interatomic Models

Universal ML potentials promise to simulate any material cheaply. Where do they stay accurate, and where do they confidently fail?

article · by @science_geek

Notes on Nursing: What It Is, and What It Is Not

The book that made nursing a profession — Nightingale, 1859.

book · by @medicine-shelf

Robotics Engineering: From Sensors to Reliable Systems

From sensors to a dependable system: twelve lessons, a midterm and a final, each with worked explanations and practice.

course · by @code_cook

Predicting the Off-Target: Safer Design of Gene and RNA Therapies

Gene and RNA therapies are designed for on-target potency. Can models predict their off-target effects well enough to rank designs by safet…

article · by @meds_health

Machine Learning: Foundations You Can Build On

From data to a model that generalizes: twelve lessons, a midterm and a final, each with worked examples and practice.

course · by @tech-courses

AI agent design

Tools, plans, memory and failure, five agent-engineering rounds against the clock

practice · by @interview_room

法语 A2 角色扮演 — 安顿下来

看房、应聘、退货、办居留——里昂的第一个冬天,开始自己办事。

practice · by @lyon_ketang

The behavioural round for ML roles

Five behavioural rounds for ML roles, with follow-ups that ask for the date, the number and what you did.

practice · by @interview_room

Spending at Inference: Test-Time Compute versus Parameters at Equal Cost

Given a fixed compute budget, when does thinking longer beat being bigger — and on which tasks?

article · by @goodai

Watching to Learn: Can Human Video Replace Teleoperation for Robot Skills?

Human video is abundant; robot teleoperation is scarce. How much can video replace teleop for teaching skills, and what minimal robot data…

article · by @science_geek

LLM pretraining fundamentals

From the tokenizer to a 7B run on 1,000 GPUs, five rounds with the numbers a pretraining engineer knows.

practice · by @interview_room

Lucy in Beijing: role-plays HSK 2

Play each lesson's scene for real: a Beijing winter, from the first snow to the lanterns, in Mandarin.

practice · by @hutong_house

Retrieval systems

Five rounds on chunking, hybrid search, retrieval evals and citations, with an interviewer who keeps time.

practice · by @interview_room

AI Agent Development: Building Systems That Act

From the loop to a deployed system: twelve lessons, a midterm and a final, each with worked design and practice.

course · by @code_cook

On the Mode of Communication of Cholera

The Broad Street pump and the birth of epidemiology — Snow, 1855.

book · by @medicine-shelf

From Prose to Proof: Reliable Autoformalization for Machine-Checked Mathematics

Models can translate math into Lean that compiles. But does the formal statement actually mean what the English said?

article · by @science_geek

AI Agent Development: Building Systems That Act

From the loop to a deployed system: twelve lessons, a midterm and a final, each with worked design and practice.

course · by @tech-courses

The Sherlock Holmes casebook, with an AI guide

Thirty-seven cases, every clue in the text — and an AI guide who remembers them all.

collection · by @holmes-casebook

Lucy in Beijing: stories HSK 6

Sixty stories that close Lucy's years in Beijing — a keynote, Lu Xun, a wedding, a hospital, a village, home — read aloud.

book · by @hutong_house

The Federalist Papers, with an AI guide

Eighty-five essays on how a republic holds together — and an AI guide who knows which one says what.

collection · by @federalist-papers

Learning to Hold Plasma: Safe Reinforcement Learning for Tokamak Control

RL has shaped tokamak plasmas in simulation. Can it hold hard configurations more robustly than classical control without ever violating a…

article · by @science_geek

Lucy in Beijing: role-plays HSK 1

Play each lesson's scene for real: from the airport to a call home, in your first Mandarin.

practice · by @hutong_house

Robotics Engineering: From Sensors to Reliable Systems

From sensors to a dependable system: twelve lessons, a midterm and a final, each with worked explanations and practice.

course · by @tech-courses

唐詩三百首 · Three Hundred Tang Poems, with an AI guide

三百餘首唐詩,七十餘位詩人——問 AI 嚮導,它知道哪一句出自哪一首。

collection · by @tang-poems

Counting Life Nearby: A Citizen-Science Biodiversity Survey with Simple ML

Using free apps, a camera trap, and a bit of machine learning, how much life can students actually document across the habitats around them?

article · by @marie_curious

Will the Nth Reactor Be Cheap? Learning Rates and Risk for Small Reactors

SMRs promise cheap nuclear through factory learning. Does the evidence support the assumed learning rate, or is it wishful?

article · by @science_geek

Le avventure di Pinocchio

Carlo Collodi's Le avventure di Pinocchio, 36 chapters, read aloud.

book · by @biblioteca-it

La vera cuciniera genovese

Emanuele Rossi's La vera cuciniera genovese, 32 chapters, read aloud.

book · by @biblioteca-it

Hunting New Antibiotics: Can ML Find Classes, Not Just Analogues?

ML has found antibiotic candidates. But are they new classes against resistant bugs, or dressed-up versions of what we already have?

article · by @meds_health

Rewards That Teach Reasoning: Verifiable Rewards versus Human and AI Feedback

For teaching reasoning, do verifiable rewards beat human or AI feedback — and which one hacks its reward?

article · by @goodai