3/31/2026
Show HN: Claude Code's Source Code Got Leaked, Let's Talk About It
Discussion of leaked Claude Code source code and its implications.
3/31/2026
Anthropic investigating Claude Code usage limits hitting faster than expected
Anthropic reports Claude Code hitting usage limits faster than expected. News about LLM application deployment.
3/31/2026
Introducing Kern: AI agents built for coworking
AI agent platform designed for cross-application deployment across terminal, browser, Slack, and mobile interfaces.
3/31/2026
Show HN: Pardus Browser- a browser for AI agents without Chromium
Pardus Browser is a lightweight browser for AI agents that doesn't rely on Chromium.
3/31/2026
Vercel: Updates to Terms of Service
Vercel updates ToS to support agentic features for proactive incident investigation and app performance monitoring.
3/31/2026
Ask HN: What do you use for local embeddings?
Discussion of local embedding alternatives to OpenAI for RAG systems, referencing Reminder project using open-source embeddings.
3/31/2026
Show HN: DreamGraph – An autonomous cognitive layer for software systems
DreamGraph is an autonomous cognitive layer for software systems that discovers, verifies, and resolves problems through structured reasoning loops without user prompts.
3/31/2026
Claude Code Source Code Leaked (2026-03-31)
Archived snapshot of Claude Code source code exposed via npm source map for security research and supply-chain analysis.
3/31/2026
Claude Code full source code leaked on NPM
NPM source map exposure revealed Anthropic's Claude Code CLI TypeScript source code via unpacked bundled files.
3/31/2026
I built 5 AI agents that scrape every platform 24/7 to find dev tools you'd miss
Project that uses 5 AI agents to autonomously scrape platforms 24/7 and curate developer tools.
3/31/2026
Cursor rules that prevent AI from hallucinating bad Next.js code
Next.js 15 boilerplate with 22 architecture rules preventing AI coding assistants from generating insecure auth, deprecated packages, and broken patterns.
3/31/2026
Prompt intensity threshold effect on AI-generated invention quality (preprint)
Study on Knowledge Innovation System prompt framework testing LLM-generated invention quality across ChatGPT, Gemini, and Claude with intensity levels and pre-structuring variants.
3/31/2026
Does RAG Help AI Coding Tools?
Benchmark comparing Claude Code and GitHub Copilot with/without RAG semantic search across 60 queries. RAG improved accuracy and reduced token consumption by 28%.
3/31/2026
LibreTranslate: Free and Open Source Machine Translation API
Free open-source machine translation API powered by Argos Translate library, self-hosted alternative to commercial services.
3/31/2026
Tell HN: If your agent can create a PR, it can merge it too
Feature request to split GitHub PAT pull-request permissions for AI coding agents and human developers to enforce separate access controls.
3/31/2026
Show HN: HolyCode – OpenCode in Docker. Use your Claude subscription. 30 tools
Docker image wrapping OpenCode AI coding agent with 30+ dev tools pre-installed, headless browser stack, and process supervision for consistent environments across machines.
3/31/2026
Memory Continuity: Zero-Dep Markdown Session Persistence for OpenClaw
OpenClaw memory plugin using Markdown for session persistence across restarts, model fallbacks, and context compaction without external dependencies.
3/31/2026
Bravos AI
No-code chatbot builder that learns from URLs, PDFs, and documents to answer customer questions.
3/31/2026
Show HN: Instant E2E encrypted (large) file sharing from your terminal
File.kiwi is an open-source CLI tool for end-to-end encrypted large file sharing with client-side encryption and no authentication.
3/31/2026
Show HN: Free AI Coding Skills for Rails
Free installable AI coding skills for Rails development that teach agents senior-level patterns, compatible with Claude Code, Cursor, and Windsurf.
3/31/2026
QuantumLeap: 2.3× faster MoE inference with intelligent expert caching
QuantumLeap optimizes MoE LLM inference with intelligent expert caching and adaptive prefetching, achieving 2.3× faster speeds on any hardware via llama.cpp integration.
3/31/2026
Show HN: TaskTrace, a local-first activity log that gives LLMs context
Desktop app that records user activity locally and surfaces it to LLMs via MCP for context-aware agent assistance, no cloud storage.
3/31/2026
Show HN: Claudebase: Sync your Claude Code env across machines via GitHub
Claudebase syncs Claude Code environments (agents, skills, rules, memory) to GitHub for multi-machine profile management and backup.
3/31/2026
Sandflare – I built a sandbox that launches AI agent VMs in ~300ms
Sandflare launches Firecracker microVMs for AI agent code execution in ~300ms with integrated Postgres for persistent state management.
3/31/2026
EchoMind – AI Voice Notes
iOS app using AI to convert voice recordings into transcripts, summaries, and action items.
3/31/2026
Show HN: Raincast – Describe an app, get a desktop app (open source)
Raincast is an open-source AI-powered tool that generates native Tauri desktop applications from natural language descriptions.
3/31/2026
Call.md
Call.md is a tool that records, transcribes, and analyzes meetings with real-time AI intelligence using agent-based processing.
3/31/2026
Vector Databases Explained in 3 Levels of Difficulty
Explains vector database functionality across three difficulty levels, covering similarity search and indexing strategies for large-scale retrieval.
3/31/2026
Would you use a GitHub App that auto-generates changelogs from commit diffs?
GitHub App that auto-generates changelogs from commit diffs using LLMs, providing team-wide standardized documentation.
3/31/2026
Read-Only vs. Action AI: Why Most Odoo AI Tools Stop at the Report
Analysis of Odoo mobile app limitations and comparison with action-oriented AI tools that write data, not just read.
3/31/2026
Screenpipe
Local desktop application that records screen activity, provides search, and enables automated agents based on user behavior.
3/31/2026
Claude usage limits hitting faster than expected
Anthropic investigating Claude Code usage limits hitting faster than expected, top priority for team.
3/31/2026
How fast can algorithms advance capabilities?
Data aggregation resource tracking AI systems by compute, organizations, satellite imagery, and chip performance.
3/31/2026
Bitboard version of Tetris AI
High-performance Tetris implementation using bitboards for efficient RL agent training with improved simulation speeds.
3/31/2026
Multiverse: Language-Conditioned Multi-Game Level Blending via Shared Representation
Language-conditioned multi-game level generation capturing structural relationships across game domains via shared representations.
3/31/2026
Concerning Uncertainty -- A Systematic Survey of Uncertainty-Aware XAI
Systematic survey of uncertainty-aware explainable AI examining methods for incorporating and evaluating uncertainty in explanations.
3/31/2026
Neuro-Symbolic Learning for Predictive Process Monitoring via Two-Stage Logic Tensor Networks with Rule Pruning
Neuro-symbolic learning using Logic Tensor Networks with rule pruning for sequential event prediction and compliance.
3/31/2026
Compliance-Aware Predictive Process Monitoring: A Neuro-Symbolic Approach
Neuro-symbolic approach for predictive process monitoring incorporating domain-specific constraints and logical rules.
3/31/2026
FormalProofBench: Can Models Write Graduate Level Math Proofs That Are Formally Verified?
FormalProofBench evaluates whether AI models can produce graduate-level math proofs formally verified in Lean 4.
3/31/2026
When Verification Hurts: Asymmetric Effects of Multi-Agent Feedback in Logic Proof Tutoring
Benchmark study evaluating LLM-provided step-level feedback for propositional logic proofs with knowledge-graph grounding.
3/31/2026
The Price of Meaning: Why Every Semantic Memory System Forgets
Theoretical analysis proving geometric tradeoff between semantic generalization and interference/forgetting in memory systems.
3/31/2026
MediHive: A Decentralized Agent Collective for Medical Reasoning
Decentralized multi-agent system using LLMs for medical reasoning with collaborative intelligence handling uncertainty and conflicting evidence.
3/31/2026
daVinci-LLM:Towards the Science of Pretraining
Study of LLM pretraining phase examining how initial training determines model capabilities and exploring transparent pretraining research.
3/31/2026
Aligning LLMs with Graph Neural Solvers for Combinatorial Optimization
AlignOPT combines LLMs with Graph Neural Networks to improve combinatorial optimization by capturing relational structures.
3/31/2026
AutoMS: Multi-Agent Evolutionary Search for Cross-Physics Inverse Microstructure Design
Multi-agent evolutionary search system for inverse microstructure design in materials science using collaborative agent-based optimization.
3/31/2026
Quantification of Credal Uncertainty: A Distance-Based Approach
Distance-based approach to quantify aleatoric and epistemic uncertainty in credal sets for multiclass classification problems.
3/31/2026
Self-evolving AI agents for protein discovery and directed evolution
VenusFactory2 self-evolving multi-agent framework for protein discovery via dynamic workflow synthesis, outperforming existing agents on benchmarks.
3/31/2026
EpochX: Building the Infrastructure for an Emergent Agent Civilization
EpochX infrastructure for coordinating AI agent work at scale through credits-native market mechanisms, addressing agent delegation and verification.
3/31/2026
CounterMoral: Editing Morals in Language Models
CounterMoral benchmark dataset for evaluating model editing techniques' ability to modify moral judgments across ethical frameworks.
3/31/2026
Beyond Completion: Probing Cumulative State Tracking to Predict LLM Agent Performance
Research on measuring LLM agent state-tracking ability via Working Memory Fidelity probe, evaluated across 20 open-weight models to predict agent performance beyond task completion.