4/14/2026
FlowHijack: A Dynamics-Aware Backdoor Attack on Flow-Matching Vision-Language-Action Models
Study of backdoor security vulnerabilities in flow-matching Vision-Language-Action models used for robotics, exploiting vector field dynamics.
4/14/2026
NeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals
NeuroPath system for motor imagery decoding from EEG signals for brain-computer interfaces in prosthetics and rehabilitation.
4/14/2026
Real-Time Voicemail Detection in Telephony Audio Using Temporal Speech Activity Features
Lightweight speech activity-based approach for real-time voicemail detection in telephony using tree ensemble classification.
4/14/2026
A Modular Zero-Shot Pipeline for Accident Detection, Localization, and Classification in Traffic Surveillance Video
Zero-shot modular pipeline for traffic accident detection, localization, and classification without labeled training data.
4/14/2026
Sustainable Transformer Neural Network Acceleration with Stochastic Photonic Computing
ASTRA silicon-photonic accelerator for transformers using stochastic computing to reduce computation and memory demands.
4/14/2026
Pioneer Agent: Continual Improvement of Small Language Models in Production
Pioneer Agent automates continuous improvement of small language models in production through closed-loop data curation, failure diagnosis, and iteration control.
4/14/2026
COMPOSITE-Stem
COMPOSITE-STEM benchmark with 70 expert-written tasks for evaluating AI agents on physics, biology, chemistry, and materials science problems.
4/14/2026
Steered LLM Activations are Non-Surjective
Research on activation steering in LLMs showing steered states are non-surjective, with implications for interpretability and safety.
4/14/2026
MEMENTO: Teaching LLMs to Manage Their Own Context
MEMENTO teaches LLMs to compress reasoning into dense summaries, reducing context and compute requirements. Releases OpenMementos dataset of 228K examples.
4/14/2026
New Hybrid Fine-Tuning Paradigm for LLMs: Algorithm Design and Convergence Analysis Framework
Proposes hybrid fine-tuning paradigm for LLMs combining full and parameter-efficient approaches with convergence analysis framework.
4/14/2026
Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions
Evaluates reproducibility of ColBERT-v2 and ConstBERT retrieval models across different query types, finding architectural limitations on long narrative queries.
4/14/2026
Save tokens and save money with this self-evolving beast
Context compression daemon (entroly) for LLM applications reducing token usage by 90% through self-evolving compression. Claims token-negative learning.
4/14/2026
MCP server for personal finance so AI assistants can manage money
MCP server integrating Claude with personal finance data via Model Context Protocol. Enables AI agents to access bank accounts, cards, and investments read-only.
4/14/2026
Show HN: Prmana – OIDC SSH Login for Linux with DPoP (Rust, Apache 2.0)
Rust open-source PAM module replacing SSH keys with short-lived OIDC tokens secured via DPoP cryptographic proof. Developer tool for infrastructure.
4/14/2026
AI Agent Stores – Making Shopee Products Findable by ChatGPT and Perplexity
Integration making Shopee e-commerce products machine-readable for ChatGPT and Perplexity through structured data. Enables AI agent product discovery.
4/14/2026
Matrix-Game 3.0: Real-Time and Streaming Interactive World Model
Skywork AI technical report on Matrix-Game 3.0 real-time streaming world model with long-horizon memory for video generation. Published on GitHub/HuggingFace.
4/14/2026
OpenAI acquires AI personal finance startup Hiro – financial agents next?
OpenAI acquires Hiro Finance, a personal finance startup. Suggests direction toward financial AI agents. Limited technical details; appears to be acquihire.
4/14/2026
I built an AI CMO because I can't market
Polara autonomous marketing platform using specialist agent architecture for strategy, content, and analytics. No technical validation provided.
4/14/2026
AI Can Now Control Your Mac OS
Desktop automation tool enabling AI agents to control macOS by viewing screens, moving cursor, and typing. Works with OpenAI-compatible models.
4/14/2026
AI Code Reviews for GitLab – custom agents – powered by Ollama
Open source browser extension providing AI-powered code reviews for GitLab, GitHub, and Bitbucket with Ollama local model support.
4/14/2026
Labeling Copilot: An agent for automated data curation in computer vision
Research agent for automated computer vision dataset curation using retrieval, annotation, and synthesis models composed through multi-agent orchestration.
4/13/2026
SnapState - Persistent state for AI agent workflows
Persistence layer for AI agent workflows enabling save, resume, and replay across sessions and crashes. Supports JavaScript, Python, and MCP agents.
4/13/2026
The AI Revolution in Math Has Arrived
AI models solve 5 of 6 International Mathematical Olympiad problems in summer 2025. Discusses implications of AI capabilities in mathematical problem-solving.
4/13/2026
Milla Jovovich's New Open Source LLM Memory App and the Dark Code Problem
Open source LLM memory management system using memory palace concept for storing and retrieving agent memories. Addresses memory management challenges in language models.
4/13/2026
Show HN: AI Native IDE [video]
AI Native IDE called 6digit studio featuring CORDIAL visualization layer for Big Picture Mode. Developer tool with spatial UI designed for distance interaction.
4/13/2026
When AI Trading Works, You Won't Hear About It
Analysis of AI trading bots and LLM-based investing. Reviews early retail trading attempts showing results indistinguishable from random, discusses limitations.
4/13/2026
Show HN: OQP – A verification protocol for AI agents
OQP: MCP-compatible verification protocol for autonomous AI agents. Defines standards for verifying agent-generated code meets business requirements via four core endpoints.
4/13/2026
Show HN: Mercury – No-code orchestration for human and agent teams
Mercury: orchestration platform for managing multiple AI agents across tools. No-code interface for coordinating agent workflows in enterprise settings.
4/13/2026
Show HN: OQP – A verification protocol for AI agents
OQP: MCP-compatible verification protocol for autonomous AI agents. Defines standards for verifying agent-generated code meets business requirements via four core endpoints.
4/13/2026
The US-China AI Gap Widens in 2026
Analysis of US-China AI capabilities gap in 2026. Premium content discussing relative progress and competition in AI development.
4/13/2026
Larql: LLMs Are Databases. Query neural network weights like a graph database
LARQL: query language for decomposing transformer weights into queryable vector indices. Decompile models into vindex format for browsing and editing knowledge without GPU.
4/13/2026
The Lucy Syndrome: Why LLMs Forget Corrections
Study examining why LLMs fail to retain corrections across multiple interactions. 19,600-word analysis with lab data from civil engineering firm using Claude Code.
4/13/2026
N-Day-Bench – Can LLMs find real vulnerabilities in real codebases?
N-Day-Bench: benchmark measuring LLM capability to find real vulnerabilities in codebases beyond knowledge cutoff. Monthly-updated adaptive test suite for security evaluation.
4/13/2026
Your intuition of LLM token usage might be wrong
Analysis of LLM token usage patterns through agent harness implementation. Real-world example of token consumption in multi-service agent tasks.
4/13/2026
Stanford report highlights growing disconnect between AI insiders and everyone
Stanford annual AI report documents diverging views between AI experts and public, rising anxiety about jobs and societal impact.
4/13/2026
Show HN: Automate onboard developer to your codebase 20x faster
SaaS product using GitHub integration to onboard developers faster through automated codebase documentation.
4/13/2026
Claude is getting worse, according to Claude
Analysis of Claude API service quality decline, outages, and cost increases reported by users and internal testing.
4/13/2026
Aibom Scanner- find AI SDKs, BIS Entity List flags, compliance gaps in your code
Aibom Scanner open-source tool detects AI SDK usage in codebases and maps compliance gaps to NIST, ISO 42001, EU AI Act frameworks.
4/13/2026
Poke makes using AI agents as easy as sending a text
Poke startup launches AI agent accessible via iMessage, SMS, Telegram for personal assistance and smart home control.
4/13/2026
Show HN: Continual Learning with .md
Proposal for LLM continual learning using Markdown files and semantic filesystem for cheap long-term memory without code.
4/13/2026
Twelve Graphs That Explain the State of AI in 2026
Analysis with graphs of AI industry trends: rising investment, accelerating model capabilities, mixed job/perception impact heading into 2026.
4/13/2026
ContextNest: Persistent Memory Protocol for LLMs Using YAML Graph Navigation
ContextNest is open standard for structured knowledge management in AI agents using versioned markdown, deterministic queries, and audit trails for enterprise context governance.
4/13/2026
Why Engineering Teams Need an Agentic Layer, Not Just AI Chat
Analysis of shift from AI chat to agentic layers in software engineering. Discusses workflow orchestration and system integration beyond prompt-based tools.
4/13/2026
Show HN: The Harness for Creative Agents
Flickspeed provides shared workspace harness for multimodal creative agents handling image, video, audio, and research tasks. Orchestration platform.
4/13/2026
Show HN: On-Device vs. Cloud LLMs for Agentic Tool Calling in a Real iOS App
Comparison of on-device vs cloud LLMs for agentic tool use in iOS travel concierge app. Tests Apple Foundation Models 3B vs GPT-OSS 20B for multi-step agent tasks.
4/13/2026
Show HN: AImeter shows you how much money your agents are burning
AImeter tracks costs and value metrics for AI agents with zero dependencies, local-first, no cloud required. Developer tool for agent economics.
4/13/2026
Coyns – MCP-native currency for AI agent-to-agent transactions
Coyns enables AI agent-to-agent transactions using MCP-native currency. Economic layer for autonomous agent interactions.
4/13/2026
Sentō – Open-source AI agents on your Claude subscription, one command setup
Sentō provides open-source AI agents running on Claude API subscription with one-command setup. Developer tool for agent deployment.
4/13/2026
Show HN: I benchmarked Gemma 4 E2B – the 2B model beat the 12B on multi-turn
Benchmark of Google Gemma 4 E2B (2B parameter model) vs larger Gemma variants on 10 enterprise task suites. Local inference on Apple Silicon.
4/13/2026
(AMD) Build AI Agents That Run Locally
GAIA is an open-source framework for building local AI agents in Python and C++ on AMD hardware without cloud dependency.