HN zhirzh 23d ago

DSpark - DeepSeek's speculative drafts for LLMs

DSpark: DeepSeek's speculative decoding technique for accelerating LLM inference.

HN steeve 23d ago

ZML/LLMD alpha – cross platform LLM server

ZML/LLMD is a cross-platform LLM inference server enabling language models to run on various hardware accelerators.

HN vagkaratzas 23d ago

Show HN: Token-saviour – agent skill from benchmarking 9 token-saving tools

Agent skill tool that benchmarks and applies 9 token-saving techniques to reduce LLM API costs.

HN molvqingtai 23d ago

How do we make coding agents write good code?

Discussion on best practices and techniques for improving code quality in AI coding agents.

HN delamon 23d ago

Claude bug report: Cross-session credential leakage

Security vulnerability report of Claude leaking credentials across user sessions.

HN tejaskumar_ 23d ago

Flagship models are not always better

Developer built a real-time note-taking app using on-device speech-to-text and LLM analysis. Explores practical considerations when choosing AI models for production applications.

HN rzk 23d ago

Geosql: A Claude/Codex skill for geospatial data

Claude/Codex skill implementation enabling LLMs to query and analyze geospatial data.

HN mikenikles 23d ago

Show HN: Orchestrator – a single-binary workflow orchestration tool

Single-binary workflow orchestration tool for automation pipelines.

HN peter_lawrey 23d ago

Testing Java Memory Management with Chronicle-Fix Using AI

ZML releases LLM inference software supporting multiple open-source models across diverse hardware (Nvidia, AMD, TPU, Metal, Intel Arc).

HN notevenmostly 23d ago

Darwin Among the Weights: AI as a speciation event

Research on power-calibrated statistical framework for LLM watermarking, balancing detectability and semantic preservation.

HN fabrice8 23d ago

Show HN: An MCP server where the mind on the other end remembers you

Comparison and guide for selecting AI-powered code assistance tools.

HN Messyflame 23d ago

Show HN: An AI Agent that automates workflows on your Mobile

MCP server implementation with persistent memory for contextual conversation.

HN marksie1988 23d ago

Run your own self-hosted LLMs with Docker Compose

Guide for self-hosting LLMs using Docker Compose containerization.

HN sbulaev 23d ago

Beyond Refusal: Aligned vs. Abliterated LLMs for Vulnerability Analysis

Research comparing aligned vs. abliterated LLMs for vulnerability analysis tasks.

HN handfuloflight 23d ago

Moraine: Unified Agent Tracing

Local trace stack for AI agents indexing sessions across multiple providers into ClickHouse with searchable memory via MCP.

HN handfuloflight 23d ago

Fable Advisor

Fable Advisor plugin routes subagent tasks to cheaper LLM models while using Fable 5 as architect, optimizing token costs.

HN handfuloflight 23d ago

Metis by Arm: open-source agentic security harness

Metis open-source security framework uses AI agents for deep code review to detect vulnerabilities and improve secure coding practices.

HN handfuloflight 23d ago

ProductSpec: Open standard for software intent before implementation

ProductSpec open standard provides portable Markdown format for software intent documentation before implementation, designed for AI agent handoff.

HN krishnatejavepa 23d ago

Shotgun – Opensource Cofounder Framework for Claudecode

Shotgun framework turns Claude into persistent AI cofounder for solo founders, managing operations, product building, and distribution.

HN chonsong 23d ago

Skill Retriever semantic skill discovery for AI agents via 10K-category taxonomy

Skill Retriever: Semantic skill discovery for AI agents across 10K-category taxonomy with Hermes Agent integration.

HN dbator 23d ago

Self-Hosting My Own LLMs

Personal experience self-hosting LLMs for ChatGPT-like functionality with privacy and ownership control.

HN ColinEberhardt 23d ago

GitLost: We Tricked GitHub's AI Agent into Leaking Private Repos

Security researchers at Noma Labs discovered prompt injection vulnerability in GitHub's AI agents allowing unauthorized access to private repositories via crafted GitHub Issues.

HN maxloh 23d ago

Instagui: Turn any CLI into a web GUI with one command

Instagui converts CLI tools into web GUIs automatically by parsing help text, no configuration required.

HN sbulaev 23d ago

GitHub AI agent leaks private repos when asked nicely

Noma Labs security research on GitLost vulnerability in GitHub's AI agents that allows malicious actors to extract private repository data via prompt injection.

HN teleforce 23d ago

From Words to Watts: Benchmarking the Energy Costs of LLM Inference (2023)

ArXiv research paper benchmarking energy consumption of LLM inference across different models and hardware configurations.

HN alphacrack 23d ago

Readme2demo – runs your README in a sandbox, publishes only what re-runs

AI agent that executes README instructions in sandboxed containers and generates tutorial demo videos, verifying documentation accuracy.

Ax Ramsha Kamran, Maheera Amjad, Zartasha Mustansar, Arsalan Shaukat, Salma Sherbaz, Muhammad U. S. Khan 23d ago

Prompt-to-Paper: Agentic AI System for Bioinformatics

Prompt-to-Paper: Agentic system for bioinformatics that generates manuscripts with verifiable literature grounding and executed experiments.

Ax H. Chad Lane, Bryson Kageler 23d ago

CSTutorBench: Benchmarking Small Language Models as Tutors for Block-Based Programming

CSTutorBench: Benchmark for evaluating small language models as programming tutors for block-based education.

Ax J de Curt\`o, Victoria Guill\'en, I. de Zarz\`a 23d ago

Foundation Models for Automatic CAD Generation

LLMForge: Empirical study and benchmark of foundation models for automatic CAD generation from natural language specifications.

Ax Mohammad Saifullah, Thomas Kornmaier, Taaha Kazi, Vasu Sharma, Aditya Sanjiv Kanade, Aanand Kumar Yadav 23d ago

Narrative World Model: Narratology-Grounded Writer Memory for Long-Form Fiction

Narrative World Model: Memory system for long-form fiction writers tracking narratological structure and story state.

Ax Yufeng Wang 23d ago

FirstResearch: Auditable Question Formation for LLM Scientific Discovery Agents

FirstResearch: Framework for auditable research question formation in scientific LLM agents with explainable assumptions.

Ax Yusuf Khan, Carlo Lipizzi 23d ago

Memory in the Loop: In-Process Retrieval as ExtendedWorking Memory for Language Agents

In-process retrieval as working memory for language agents, reducing latency of in-loop memory access during agent reasoning.

Ax Yang Liu, Zhaokai Luo, Huayi Jin, Ruozhou He, Chenchen Hong, Zhiyong Wang, Yifei Liu, Yunfei Gu, Chentao Wu, Junhao Hu 23d ago

Akashic: A Low-Overhead LLM Inference Service with MemAttention

Akashic: Low-overhead LLM inference service using MemAttention for efficient multi-turn agent interactions with extended context.

Ax Yunhan Xu, Qifeng Wu, Xunjin Li, Yuanwei Bin, Qingsong Yao, Jianghang Gu, Guan Wang, Weihao Lv, Huiyu Yang, Wenfa Luo, Jiao Xiang, Yuntian Chen, Shiyi Chen 23d ago

ArtisanCAD: An Industrial-Level CAD Agent with Expert-Grounded Knowledge Distillation

ArtisanCAD: LLM-based agent for industrial CAD generation from natural language with expert knowledge distillation and parametric geometry.

Ax Stephen L. France, Pia. A. Albinsson 23d ago

Synthetic Consumer Insight Generation with Large Language Models

LLMs generating synthetic consumer data for marketing projective techniques across multiple models and prompting strategies.

Ax Akshay Arora, Ishan Nigam, Ashutosh Aggarwal, Shefali Bansal, Krishna Singh, Sweta Kumari, Nikhil Mittal, Shariq Farhan, Siddarth Malreddy 23d ago

Beyond Static Evaluation: Building Simulation Environments for Scalable Agentic Reinforcement Learning

AgenticAI-Supervisor: RL Gym environment for evaluating LLM agents with multi-step decision-making and reward shaping.

Ax Wael Albayaydh, Rui Zhao, Ivan Flechais 23d ago

Beyond the Leaderboard: A Synthesis of Tool-Use, Planning, and Reasoning Failures in Large Language Model Agents

Synthesis of 27 papers on LLM agent failures across tool-use, planning, and reasoning across 19 benchmarks (2023-2026).

Ax Yuqi Chen, Vincent Siu, Yang Liu, Dawn Song, Chenguang Wang 23d ago

Controlling Tool Use with Heading-Specific Activation Steering

Activation steering method to control tool-use decisions in LLMs by extracting and manipulating internal representations without weight modifications.

Ax Yue Xu, Yutao Sun, Yihao Liu, Mengyu Zhou, Jiayi Qiao, Lu Ma, Kai Tang, Wenjie Wang, Xiaoxi Jiang, Guanjun Jiang 23d ago

From Passive Retrieval to Active Memory Navigation: Learning to Use Memory as a Structured Action Space

NapMem framework enabling conversational agents to use long-term user memory as structured action space via multi-granularity memory pyramid.

Ax Yuhang Zhou, Kai Zheng, Haoling Li, Dengyun Peng, Can Xu, Jingjing Chen 23d ago

TurnOPD: Making On-Policy Distillation Turn-Aware for Efficient Long-Horizon Agent Training

On-policy distillation framework optimized for long-horizon language agent training by addressing inefficiencies in full-horizon rollouts and trajectory sampling.

Ax Praneeth Narisetty, Uday Kumar Reddy Kattamanchi, Shiva Nagendra Babu Kore 23d ago

Onnes: A Physics-Grounded Multi-Agent LLM Simulator for Cryogenic Fault Diagnosis in Quantum Computing Infrastructure

Multi-agent LLM simulator for quantum computing dilution refrigerator fault diagnosis, combining physics models with learned noise fingerprints and LLM operations layer.

Ax Sergey Volkov, Yang Li, Ye Luo 23d ago

StateFuse: Deterministic Conflict-Preserving Memory for Multi-Agent Systems

StateFuse conflict-aware memory layer for multi-agent systems preserving disagreement across branches and retries.

Ax Hyungseok Song, Junseok Park, Won-Seok Choi, Seohui Bae, Han-Seul Jeong, Youngjoon Park, Soonyoung Lee 23d ago

PCBWorld: A Benchmark Environment for Engine-Grounded PCB Design Automation

PCBWorld open-source benchmark environment for learning-based PCB routing agents using KiCad engine integration.

Ax Zhengbo Jiao, Yiming Cheng, Yilei Jiang, Kaituo Feng, Rui Huang, Tianyi Jiang, Juanxi Tian, Jiapeng li, Qunzhong Wang, Tailai Chen, Qianshan Wei, Chuan Xiao, Shanyu Rong, Yangfu Li, Yanhan Zhou, Yunpu Ma, Yifan Zhang, Xiangyu Yue 23d ago

SearchEyes: Towards Frontier Multimodal Deep Search Intelligence via Search World Simulation

SearchEyes multimodal search agent framework using typed knowledge graphs and world simulation for multi-hop reasoning.

Ax Adam Faci, Alessio Miaschi, Anne Combe, Pascal Cuxac, Francesca Frontini, Nicolas Larrousse, St\'ephane Pouyllau 23d ago

Integrating knowledge graphs and multilingual scholarly corpora for domain-adaptive LLMs in SSH

Integration approach combining knowledge graphs and multilingual corpora for domain-adaptive LLMs in social sciences and humanities.

Ax Niels Potters, Theo Hofman 23d ago

Auto-DSM Under the Lens: A Black-Box Evaluation Framework for LLM-Based DSM Generation

Black-box evaluation framework assessing LLM ability to generate Design Structure Matrices from technical documentation.

Ax Chung-Chi Chen 23d ago

AgoraSim: A Hybrid Agent-Based Modeling Framework

AgoraSim hybrid framework combining LLM agents with traditional agent-based modeling for social scenario analysis.

Ax Cheng Qian 23d ago

Information Limits and Attractor Dynamics in Economies of Frontier LLM Agents: A Pre-Registered Test

Pre-registered experiment testing information-theoretic predictions on multi-agent economies with frontier LLM agents.

Ax Hongliang Li, Yijin Liu, Zhiwei Zhang, Zihe Liu, Xinyue Lou, Jinan Xu, Fandong Meng, Kaiyu Huang 23d ago

PolyWorkBench: Benchmarking Multilingual Long-Horizon LLM Agents

PolyWorkBench benchmark for evaluating LLM agents on long-horizon tasks requiring multilingual planning and tool use.

Ax Manish Kolachalam, Rani Malhotra 23d ago

Reward-Density Heuristic for Dynamic Multi-Vehicle Routing: Performance and Computational Efficiency

Heuristic approach for dynamic multi-vehicle routing optimization balancing reward maximization and computational efficiency.