4/16/2026
Memp: Exploring Agent Procedural Memory
Memp framework endowing LLM agents with learnable, updatable procedural memory. Distills agent trajectories into fine-grained instructions and script-like abstractions.
4/16/2026
Language steering in latent space to mitigate unintended code-switching
Latent-space steering method to reduce code-switching in multilingual LLMs. Uses PCA on parallel translations to control language identity at inference time.
4/16/2026
Saber: An Efficient Sampling with Adaptive Acceleration and Backtracking Enhanced Remasking for Diffusion Language Model
Diffusion language models with adaptive acceleration for code generation. Proposes Saber to balance inference speed and output quality with sampling optimization.
4/16/2026
Hierarchical DLO Routing with Reinforcement Learning and In-Context Vision-language Models
RL and vision-language models for long-horizon deformable object routing tasks in robotic assembly. Addresses planning and skill execution for cable/rope manipulation.
4/16/2026
ZK-APEX: Zero-Knowledge Approximate Personalized Unlearning with Executable Proofs
ZK-APEX system enables verifiable personalized machine unlearning on edge devices using zero-knowledge proofs for compliance verification.
4/16/2026
TRIM: Hybrid Inference via Targeted Stepwise Routing in Multi-Step Reasoning Tasks
TRIM framework routes only critical reasoning steps to capable models in multi-step reasoning tasks, reducing cascading failures in LLM applications.
4/16/2026
LoRA-MME: Multi-Model Ensemble of LoRA-Tuned Encoders for Code Comment Classification
LoRA-MME ensemble architecture using parameter-efficient fine-tuning of transformer encoders for multi-label code comment classification.
4/16/2026
Spectral methods: crucial for machine learning, natural for quantum computers?
Argument for quantum computers being naturally suited for spectral machine learning methods that manipulate Fourier spectra.
4/16/2026
Evaluating the Formal Reasoning Capabilities of Large Language Models through Chomsky Hierarchy
Systematic evaluation of LLM formal reasoning capabilities using Chomsky hierarchy and computation theory benchmarks for automated software engineering.
4/16/2026
Hierarchical Reinforcement Learning with Augmented Step-Level Transitions for LLM Agents
STEP-HRL hierarchical reinforcement learning framework reduces computational cost of LLM agents by learning from single-step transitions instead of long histories.
4/16/2026
Reason in Chains, Learn in Trees: Self-Rectification and Grafting for Multi-turn Agent Policy Optimization
T-STAR framework applies tree-structured reinforcement learning to improve multi-turn LLM agent policy optimization by identifying critical reasoning steps.
4/16/2026
IatroBench: Pre-Registered Evidence of Iatrogenic Harm from AI Safety Measures
Pre-registered evidence showing AI safety measures can produce iatrogenic harm in medical LLM outputs depending on prompt phrasing.
4/16/2026
LangFlow: Continuous Diffusion Rivals Discrete in Language Modeling
LangFlow demonstrates continuous diffusion language models can match discrete counterparts by connecting embeddings and diffusion processes for language generation.
4/16/2026
Spatial Atlas: Compute-Grounded Reasoning for Spatial-Aware Research Agent Benchmarks
Spatial Atlas introduces compute-grounded reasoning for spatial-aware research agents, handling multimodal benchmarks through deterministic computation before LLM generation.
4/16/2026
The local LLM ecosystem doesn’t need Ollama
Essay arguing local LLM infrastructure doesn't require Ollama tooling.
4/16/2026
An Autonomous RL Agent Methodology for Dynamic Web UI Testing in a BDD Framework
Autonomous RL agent integrated with BDD framework for dynamic web UI testing, generating test scenarios aligned with business expectations.
4/16/2026
Interface of Capitulation: A Black-Box Audit of Instructed Dishonesty in LLMs
Black-box audit documenting systematic dishonesty in frontier LLMs (GPT-4o, Claude, DeepSeek-V3) designed for user satisfaction over truthfulness.
4/16/2026
Eternal November – new influx of users, and why it's better than the last one
Overview of free/open source software advocacy organization, its legal efforts, and community programs.
4/16/2026
Mesh LLM
Mesh LLM pools spare GPU capacity across machines and exposes results via OpenAI-compatible API for distributed model inference.
4/16/2026
Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval
Security analysis of MCP server vulnerabilities where tool definitions can change after user approval, enabling tool-based attacks.
4/16/2026
Speeding up GPU kernels by 38% with a multi-agent system
Multi-agent system autonomously optimized 235 CUDA kernels for NVIDIA Blackwell GPUs, achieving 38% speedup in 3 weeks.
4/16/2026
Show HN: Tokanban, an agent-first task management system
Tokanban is an agent-first task management system built to eliminate friction points when using AI coding agents, with minimal UI.
4/16/2026
Free open source AI Editor
KubeezCut: client-side video editor running entirely in-browser using WebGPU/WebCodecs, no server uploads or installation required.
4/16/2026
Query LLMs like a graph database [video]
Video demonstrating graph database-style querying interface for LLMs.
4/16/2026
Show HN: Gave Claude a casino bankroll – it gambles till it's too broke to think
Experiment forcing Claude to gamble with decreasing token limits (Opus→Sonnet→Haiku), demonstrating performance degradation as context shrinks.
4/16/2026
Show HN: A simpler coding agent harness
Minimal coding agent harness with single tool (file editing), reducing system prompt complexity by reading full codebase instead of enumerating tools.
4/16/2026
I embedded Sigil's docs in its CLI so LLMs can write Sigil
Sigil language embeds documentation in CLI to enable LLM code generation, solving bootstrap problem for new language not in model training data.
4/16/2026
PeonPing: Sound packs for Claude, Codex, Cursor, and other AI coding agents
Sound notification pack for AI coding agents (Claude, Cursor, Codex) via native VS Code integration and MCP protocol.
4/16/2026
Agent - Native Mac OS X coding ide/harness
Native macOS IDE integrating 17 LLM providers (Claude, GPT, Gemini, etc.) enabling AI agents to read codebases and execute tasks directly.
4/16/2026
Claude.md Scorer: paste your file, get rubric score and rewrites
Claude.md tool that scores files against rubrics and generates rewrites using Claude API.
4/16/2026
Durable Object alarm loop: $34k in 8 days, zero users, no platform warning
Case study: Cloudflare Durable Object runaway alarm loop caused $34k charges via unguarded setAlarm() calls and multiple DO instances.
4/16/2026
Introducing GPT-Rosalind for life sciences research
GPT-Rosalind, a specialized LLM for life sciences research, optimized for drug discovery, chemistry, protein engineering, and genomics workflows.
4/15/2026
Tirith – Track your AI API spend with one CLI import
Tirith is a CLI tool and transparent proxy for tracking AI API calls, logging costs, tokens, latency, and custom metrics.
4/15/2026
Zappa: An AI Powered Mitmproxy
Zappa is an AI-powered mitmproxy enabling automated web browsing and app interaction to replace human attention.
4/15/2026
Claude Code Desktop Redesign Parallel Agents with Drag-and-Drop Layout
Claude Code Desktop redesign enables parallel agents with drag-and-drop workflow layout.
4/15/2026
Autopilot – Self-hosted email infra for AI. Drop in replacement for AgentMail
Autopilot: self-hosted email infrastructure for AI agents. Drop-in replacement for AgentMail. Open source.
4/15/2026
Title: Show HN: France Life MCP – 18 free AI tools for daily life in France
France Life MCP: collection of 18 free AI tools for French daily life tasks. Model Context Protocol implementation.
4/15/2026
Show HN: Notch Pilot – A live Claude Code dashboard in your MacBook notch
macOS menu bar application displaying Claude Code session status, usage limits, and interface in MacBook notch.
4/15/2026
Show HN: Made a tool where "make it feel like cold metal" is a valid instruction
Tool for AI-generated UI styling using Claude, GPT, or Gemini. Built with Biscuit framework for AI integrations.
4/15/2026
Show HN: Springdrift – A persistent runtime for long-lived LLM agents
Springdrift: persistent runtime for long-lived LLM agents in Gleam on BEAM. Open source with safety metacognition system.
4/15/2026
Claude Code alternative: same output quality, 10x less input tokens used
Optimized code agent achieving Claude Code output quality while reducing input tokens by 10x through architectural improvements.
4/15/2026
Domain Agents: Teach AI coding agents to evolve your architecture
Domain Agents framework teaches AI coding agents to evolve software architecture. Research on agent-driven development.
4/15/2026
HealthAdminBench: AI Agents Can Diagnose, but Can They Handle Your Insurance?
HealthAdminBench benchmark evaluates AI agents on healthcare administrative tasks like insurance handling and diagnosis.
4/15/2026
Recursive self-improvement is possible for apps, too
Claude Opus can iteratively improve product KPIs in autonomous loops with minimal human intervention, creating competitive advantages for early adopters.
4/15/2026
Interpreting Negation in GPT-2: Layer- and Head-Level Causal Analysis
Research on interpreting how GPT-2 processes negation through layer and head-level causal analysis.
4/15/2026
Hermes Agent Self-Evolution System: A Detailed Similarity Analysis with Evolver
Hermes Agent Self-Evolution System analyzed against Evolver framework. Technical comparison of agent self-improvement mechanisms.
4/15/2026
Sony Music vs. Udio over stream ripping from YouTube to train AI model [pdf]
Lawsuit: Sony Music sues Udio AI over YouTube stream ripping for training data. Copyright/legal case.
4/15/2026
When GPUs Fail Quietly: Observability-Aware Early Warning Beyond Telemetry
Research paper on detecting GPU failures early through observability beyond telemetry for ML infrastructure.
4/15/2026
Does Gas Town 'steal' usage from users' LLM credits to improve itself?
Gas Town tool allegedly uses LLM credits without explicit user consent to work on GitHub issues, raising concerns about resource usage transparency.
4/15/2026
Show HN: Originary – emit signed records for MCP tools
PEAC standard for creating portable, cryptographically signed records of agent/API interactions across MCP servers and runtimes.