HN manas95 5/21/2026

Show HN: Real-time virtual try-on using hand gestures and live video diffusion

Real-time virtual try-on application using hand gestures and live video diffusion for dynamic retail interfaces.

HN scolvin 5/21/2026

Fast local spreadsheet viewer based on GPUI

Fast local spreadsheet viewer in Rust using GPUI, built primarily with AI coding tools. Supports XLSX and Parquet formats.

HN doener 5/21/2026

Agent Bazaar: Enabling Economic Alignment in Multi-Agent Marketplaces

Agent Bazaar framework for simulating and evaluating multi-agent LLM systems in economic marketplaces for stability and integrity.

HN vitalysim 5/21/2026

Show HN: The-knowledge-guy – ask, walk, and skim every book on your shelf

Tool converting PDFs/EPUBs into queryable Claude Code skills, enabling cross-document question answering with efficient token usage.

HN takira 5/21/2026

What it was like working on LLMs and security at Meta (2022-2026)

Personal account of working on LLMs, security, and open source AI at Meta (2022-2026), including Llama and CodeLlama development.

HN SVI 5/21/2026

AI red teaming agents change how LLMs get tested

Red teaming agents for LLM testing using adversarial techniques and tools like PyRIT, Garak, and Promptfoo frameworks.

HN aledevv 5/21/2026

Google search AI mode, the biggest update

Google Search update adding AI Mode with Gemini 3.5 Flash for conversational search experience similar to ChatGPT.

HN calcifer 5/21/2026

Twelve Ways to Be Wrong About AI-Assisted Coding

Stub post title only, no content provided.

HN moisacow 5/21/2026

Linki – open-source AI SDR for LinkedIn sequences and cold email

Open source AI sales development representative tool for automating LinkedIn sequences and cold email outreach.

HN joelolawanle 5/21/2026

How to scrape any website and get structured data with a single API call

Guide on web scraping and converting unstructured HTML to clean structured JSON data without custom parsers.

HN dohguy 5/21/2026

Are TypeScript back end frameworks ready for AI Agents?

Benchmark comparing Claude Code agent performance building TypeScript backends across five frameworks with same tasks and evaluation rubric.

HN levmiseri 5/21/2026

Show HN: Asciidia – LLM-Powered Game

LLM-powered text adventure game where players conjure objects and the AI generates functionality and properties dynamically.

HN gaojiuli 5/21/2026

Show HN: Handsets – A high-performance Android control CLI, built for agents

High-performance Android device control CLI built for AI agents. Provides low-latency command execution via binary protocol and state mirroring.

HN enzo_ferrari 5/21/2026

Show HN: A SQLite graph that captures why AI-generated code exists

Reference architecture using SQLite graph to preserve reasoning and context in AI-generated code beyond individual sessions.

HN anhydrite 5/21/2026

LLM: Documentation driven exploration for big codebase

Project providing structured documentation skills for AI coding agents, maintaining docs in sync with code through hierarchical dependency matrices.

HN Thenoahai 5/21/2026

Growing an AI orchestration platform to $3k MRR in 4 weeks

Founder built AI orchestration platform (Meerkats.ai) reaching $3k MRR in 4 weeks. Case study on GTM and growth strategy.

HN doppp 5/21/2026

Dust raises $40M Series B to scale multiplayer AI for human-agent collaboration

Dust raises $40M Series B funding to scale multiplayer AI platform for human-agent collaboration.

HN hpcaitech 5/21/2026

New API Models Launched

DeepSeek V4 Pro and V4 Flash models now available via HPC-AI.COM API with competitive pricing for LLM applications.

HN geox 5/21/2026

Merlin to Bring Military Autonomy Platform to Commercial Air Cargo

Merlin Labs extends autonomous pilot system from military to commercial cargo aircraft operations.

HN AsteroidHunter 5/21/2026

Put yourself in a loop with your coding agents

Companion app for managing and switching between multiple Claude Code AI agent sessions.

HN kmdupree 5/21/2026

Show HN: FKS2G – LLM-backed metrics for deciding how closely to review code

LLM-backed CLI tool that analyzes code changes to determine review thoroughness needed.

HN v4d1mv 5/21/2026

No JavaScript. No npms. Make realtime web apps in modern Java

Java-based toolkit for building stateful, server-rendered web applications without JavaScript.

HN javatuts 5/21/2026

Tokenspeed – How fast is 10 tokens per second really?

Interactive tool to visualize LLM token generation speeds across different hardware platforms and models.

Ax Bingnan Liu, Chenhang Cui, Rui Huang, Jiani Luo, Zhirong Shen, Tinghao Wang, Xiande Huang, Lingbei Meng, Fei Shen, An Zhang 5/21/2026

WildRoadBench: A Wild Aerial Road-Damage Grounding Benchmark for Vision-Language Models and Autonomous Agents

Benchmark coupling vision-language models with LLM-driven autonomous agents for aerial road-damage detection from UAV imagery.

Ax Zhaohui Zheng, Chenhang He, Shihao Wang, Yuxuan Li, Ming-Ming Cheng, Lei Zhang 5/21/2026

DEL: Digit Entropy Loss for Numerical Learning of Large Language Models

Novel loss function for LLMs to improve numerical prediction in math and code generation tasks, addressing limitations of standard maximum likelihood estimation.

Ax Yingwei Li, Xin Huang, Yang Liu, Yang Fu, Alex Zihao Zhu, Chen Song, Junwen Yao, Anant Subramanian, Hao Xiang, Weijing Shi, Yuliang Zou, Tom Hoddes, Zhaoqi Leng, Govind Thattai, Dragomir Anguelov, Mingxing Tan 5/21/2026

STELLAR: Scaling 3D Perception Large Models for Autonomous Driving

Study on scaling 3D perception models for autonomous driving, analyzing impact of model scale on multi-sensor fusion and spatial understanding.

Ax Takato Yasuno 5/21/2026

Understanding Deterioration Random Effects for Causal Discovery in Infrastructure Management

Bayesian hierarchical model for identifying heterogeneous deterioration patterns in infrastructure equipment using causal discovery.

Ax Jesse Campbell, Daniel Ibaibarriaga, Lev Reyzin 5/21/2026

Contradiction Graphs Determine VC Dimension

Theoretical work showing contradiction graphs determine VC dimension for binary concept classes.

Ax Animesh Maheshwari, Divyansh Sahu, Nishit Verma 5/21/2026

Do Vision--Language Models Understand 3D Scenes or Just Catalogue Objects?

Research on whether vision-language models understand 3D spatial scenes or only catalog objects.

Ax Yifeng He, Ethan Wang, Jicheng Wang, Xuanxin Ouyang, Hao Chen 5/21/2026

Code Generation by Differential Test Time Scaling

DiffCodeGen: test-time scaling method for code generation using coverage-guided approach with reduced token consumption.

Ax Jianmin Liao, Lixin Shen, Yuesheng Xu 5/21/2026

Oracle Supervision Transfers for Hyperparameter Prediction in Model-Based Image Denoising

Research on hyperparameter prediction for model-based image denoising using oracle supervision transfers.

Ax Lautaro Estienne, Erik Ernst, Mat\'ias Vera, Pablo Piantanida, Luciana Ferrer 5/21/2026

\ECUAS{n}: A family of metrics for principled evaluation of uncertainty-augmented systems

Research on metrics for evaluating uncertainty-augmented systems in automated decision-making contexts.

Ax Parsa Mazaheri, Kasra Mazaheri 5/21/2026

AgentAtlas: Beyond Outcome Leaderboards for LLM Agents

AgentAtlas provides unified benchmark beyond single-metric leaderboards for evaluating LLM agents across multiple dimensions including task success, tool validity, safety, and robustness.

Ax Slim Barkallah, Luke Bailey, Kaiyue Wen, Mohammed Abouzaid, Tengyu Ma 5/21/2026

Pseudo-Formalization for Automatic Proof Verification

Pseudo-Formalization method for automatic proof verification translating informal proofs into formal languages to enable AI mathematical reasoning evaluation.

Ax Haoyang Cao, Xin Guo, Wenpin Tang, Guan Wang 5/21/2026

Sample Complexity of Transfer Learning: An Optimal Transport Approach

Theoretical analysis of transfer learning sample complexity using optimal transport, addressing efficiency gains for LLMs and generative AI models.

Ax Tom\'a\v{s} Koc\'ak, Michal Valko, R\'emi Munos, Branislav Kveton, Shipra Agrawal 5/21/2026

Spectral bandits for smooth graph functions with applications in recommender systems

Bandit algorithm for smooth graph functions applicable to online content-based recommendation systems where item ratings correlate with neighbors.

Ax Hamza Golubovic, Matthew Shen, Genevera I. Allen, Tarek M. Zikry 5/21/2026

Group-Aware Matrix Estimation and Latent Subspace Recovery

Matrix completion method for heterogeneous data with multiple group memberships, preserving subgroup-specific variation in recommendation and neuroscience applications.

Ax Mengyang Liu, Taozhi Chen, Zhenhua Xu, Xue Jiang, Yihong Dong 5/21/2026

Multi-agent Collaboration with State Management

Framework for multi-agent collaboration addressing concurrent codebase edits through state management to prevent silent conflicts and integration failures.

Ax Soichiro Nishimori, Shinri Okano, Keigo Habara, Sotetsu Koyamada, Eason Yu, Masashi Sugiyama 5/21/2026

Mahjax: A GPU-Accelerated Mahjong Simulator for Reinforcement Learning in JAX

GPU-accelerated Mahjong simulator in JAX for reinforcement learning research on multi-player imperfect-information games with high-dimensional state spaces.

Ax Ming Liu 5/21/2026

Self-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax Dies

Analyzes how self-training on LLM outputs restructures language rather than flattening it, showing surface markers amplify while deep syntax diminishes across five model variants.

Ax Yifei Wang, Tianlin Li, Xiaohan Zhang, Yida Yang, Xiaoyu Zhang, Li Pan 5/21/2026

Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs

Backdoor attacks on LLMs via inference optimization/compilation exploiting numerical side effects; proposes attack framework.

Ax Yuxi Liu, Zekun Zhang, Yixiang Cai, Renjia Deng, Yutong He, Kun Yuan 5/21/2026

RoPeSLR: 3D RoPE-driven Sparse-LowRank Attention for Efficient Diffusion Transformers

Sparse-LowRank attention for diffusion transformers using 3D RoPE; addresses efficiency bottleneck in long-sequence video generation.

Ax Seungone Kim, Dongkeun Yoon, Kiril Gashteovski, Juyoung Suk, Jinheon Baek, Pranjal Aggarwal, Ian Wu, Viktor Zaverkin, Spase Petkoski, Daniel R. Schrider, Ilija Dukovski, Francesco Santini, Biljana Mitreska, Yong Jeong, Kyeongha Kwon, Young Min Sim, Dragana Manasova, Arthur Porto, Biljana Mojsoska, Makoto Takamoto, Marko Shuntov, Ruoqi Liu, Hyunjoo Jenny Lee, Niyazi Ulas Din\c{c}, Yehhyun Jo, Sunkyu Han, Chungwoo Lee, Huishan Li, Esther H. R. Tsai, Ergun Simsek, Khushboo Shafi, Yeonseung Chung, Jihye Park, Aleksandar Shulevski, Henrik Christiansen, Yoosang Son, Elly Knight, Amanda Montoya, Jeongyoun Ahn, Christian Langkammer, Heera Moon, Changwon Yoon, Nikola Stikov, Mooseok Jang, Edward Choi, Junhan Kim, Yeon Sik Jung, Woo Youn Kim, Jae Kyoung Kim, Ishraq Md Anjum, Hyun Uk Kim, Drew Bridges, Carolin Lawrence, Xiang Yue, Alice Oh, Akari Asai, Sean Welleck, Graham Neubig 5/21/2026

On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists

Study of AI peer reviewers on Nature papers with 45 expert scientists; evaluates capabilities, limitations, and credibility.

Ax Dongfang Zhao 5/21/2026

DIVE: Embedding Compression via Self-Limiting Gradient Updates

DIVE embedding compression via self-limiting gradients; reduces dimensionality for vector search systems without severe overfitting.

Ax Reese Levine, Rithik Sharma, Nikhil Jain, Abhijit Ramesh, Zheyuan Chen, Neha Abbas, James Contini, Tyler Sorensen 5/21/2026

Llamas on the Web: Memory-Efficient, Performance-Portable, and Multi-Precision LLM Inference with WebGPU

WebGPU backend for llama.cpp enabling memory-efficient, multi-precision LLM inference directly in browser.

Ax Miaobo Hu, Shuhao Hu, Bokun Wang, Rui Chen, Xin Wang, Xiaobo Guo, Daren Zha, Jun Xiao 5/21/2026

SAVER: Selective As-Needed Vision Evidence for Multimodal Information Extraction

Selective multimodal fusion for information extraction; decides when vision evidence is needed and which images to use.

Ax Xuehui Yu, Fucheng Cai, Meiyi Wang, Xiaopeng Fan, Harold Soh 5/21/2026

Conflict-Aware Additive Guidance for Flow Models under Compositional Rewards

Conflict-aware guidance for composing multiple constraints in diffusion/flow models without fine-tuning.

Ax Victoria Lin, Taedong Yun, Maja Matari\'c, John Canny, Arthur Gretton, Alexander D'Amour 5/21/2026

The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study

LLM-simulated experiments suffer from user drift; interventions cause unintended latent attribute shifts making results observational.

Ax Yosuke Miyanishi, Tetsuro Morimura 5/21/2026

Interaction Locality in Hierarchical Recursive Reasoning

Framework for measuring information flow locality in hierarchical reasoning using sparse autoencoders and activation patching.

Ax Connor Pedersen, Dong H. Ahn, Michel Migdal, Collin Neale, Nik Konyuchenko 5/21/2026

Instant GPU Efficiency Visibility at Fleet Scale

GPU efficiency metric (OFU) for AI workloads on HPC systems using performance counters; no application instrumentation needed.