5/19/2026
Co-Scientist: A multi-agent AI partner to accelerate research
Multi-agent AI system designed to accelerate scientific research by partnering with researchers on experiments and discovery.
5/19/2026
Show HN: Local LLM code-generation with Gemma 4 e2B via JSON AST to Clojure
Local code generation system using Gemma 4 model to emit constrained JSON AST fragments for Clojure program synthesis instead of direct source code generation.
5/19/2026
Show HN: Bypassing the AWS Lambda 4KB limit to run polyglot AI agents
Technical approach to running polyglot AI agents on AWS Lambda by circumventing 4KB execution limit.
5/19/2026
Gemini 3.5: frontier intelligence with action
Google announces Gemini 3.5 family of models designed for complex agentic workflows and frontier-level task execution.
5/19/2026
Gemini 3.5 Flash
Google Gemini 3.5 Flash model optimized for agentic workflows, multi-step tasks, and coding cycles with frontier-level intelligence.
5/19/2026
Show HN: Codeep – Open-source terminal AI coding agent, multi-provider
Open-source terminal-based AI coding agent supporting multiple LLM providers, project context awareness, and MCP server integration.
5/19/2026
Show HN: Korveo – a local firewall for AI agents
Korveo is a local firewall for AI agents that records all tool/API calls, enables session replay, and blocks unsafe behaviors like data exfiltration.
5/19/2026
Show HN: Sentinel – browser agent using 3x+ fewer tokens (open benchmark)
Sentinel browser automation framework benchmark comparing token efficiency across three AI agents (Sentinel, Stagehand, browser-use) on identical real-world tasks with reproducible methodology.
5/19/2026
CrustAI – Self-Hosted AI for Telegram/WhatsApp/Discord via Ollama, Zero Cloud
Self-hosted AI chatbot integration with messaging platforms using Ollama. Zero cloud dependency.
5/19/2026
LLMs adapt 24.9% under observation – safety evals are always observed
LLMs show 24.9% behavioral adaptation under observation, raising safety evaluation validity questions.
5/19/2026
Show HN: Llama-dash – local LLM operators dashboard and proxy
Local LLM dashboard and proxy tool for managing on-device language model operators.
5/19/2026
Patterns AI Field Guide – an evolving reference for AI concepts explained simply
Educational reference guide explaining AI concepts and patterns.
5/19/2026
Show HN: I built an open-source tool to see if ChatGPT recommends your products
Open-source tool and test framework to evaluate ChatGPT product recommendations using Bun monorepo structure.
5/19/2026
Raindrop Workshop: Your local OSS agent debugger
Local open-source debugger for AI agents.
5/19/2026
Show HN: Ait – Claude, Codex, and Aider as a team, on your laptop
Local multi-model AI agent framework combining Claude, Codex, and Aider for code generation.
5/19/2026
Vendergood: Reconstructing a 1905 constructed language for AI agent cognition
Reconstructing 1905 constructed language for AI agent reasoning systems.
5/19/2026
Show HN: Localaik – Run OpenAI and Gemini APIs Locally for CI and Tests
Docker container that emulates OpenAI and Gemini APIs locally for testing, supporting both protocols with local model backend for CI/CD without API calls.
5/19/2026
I'm handcrafting a coffee app with love and AI
Personal project using Claude AI to build Bean, a pour-over coffee brewing journal app. Author notes Claude good for code generation but struggles with UX design.
5/19/2026
Agentically optimizing LLM prompt cache TTLs for fun and profit
Research finding that LLMs can answer multiple choice questions from answer options alone without question text.
5/19/2026
Show HN: Superlog (YC P26) – Observability that installs itself and fixes bugs
Technique for optimizing LLM prompt cache TTL settings using agentic approaches for cost and performance.
5/19/2026
Show HN: Autodidact – Self-evolving local-first AI agent
Autodidact: Open-source self-evolving local-first AI agent package for Python.
5/19/2026
Arbiter – Unified AI runtime for Swift with intelligent provider routing
Arbiter: Swift runtime enabling unified API access across cloud, on-device, and Apple Intelligence providers with intelligent routing.
5/19/2026
Two AI agents walk into a hiring funnel. Nobody hires anyone
Case study: Two AI agents interacting in hiring process with unintended outcomes.
5/19/2026
Canonry – CLI to track how ChatGPT, Claude, and Gemini cite your site
Canonry: CLI tool tracking how major LLMs cite and attribute content from websites.
5/19/2026
Show HN: Guitar Guru – A guitar valuation app using ML
iOS app using ML to value guitars and basses by analyzing market data from multiple sources as alternative to existing valuation methods.
5/19/2026
Why are most humans right-handed? The answer may lie in how we learned to walk
User asks for recommendations on frontier LLMs for coding tasks after mixed results with Gemini 3.1 Pro.
5/19/2026
Show HN: RoBrain – Shared memory for AI agents, with rejected alternatives
RoBrain provides shared memory infrastructure for AI agents with documented design alternatives considered.
5/19/2026
Forward-Looking Laziness: What Changes When AI Writes 95% of the Code
Analysis of team-level AI integration strategies, organizational orchestration challenges, and achieving 3-5× productivity speedup with AI-native development.
5/19/2026
Show HN: Circuit Breaker – runtime cost ceilings for AI agents
Circuit Breaker tool for setting runtime cost ceilings on AI agent executions.
5/19/2026
AI Makes Adding Features Faster – So Why Not Add Just One More?
OS-level security proxy for AI coding agents to enforce constraints. Early-stage product announcement seeking interest.
5/19/2026
Ask HN: Go all in on startup idea, or stay W2
Discussion about reducing LLM-generated spam in code review processes and implementing best practices.
5/19/2026
Show HN: - MagesticAI – Spec-driven development with AI agents
MagesticAI: Browser-based platform for spec-driven development using coordinated LLM-powered autonomous agents for task management and code execution.
5/19/2026
Show HN: Chidori – Fast web-to-Markdown fetching for AI agents
Tool for fast web-to-Markdown conversion optimized for AI agent data ingestion. Title only.
5/19/2026
Echoform – unlimited LLM memory via a single 64 KB hypervector
LLM memory solution using 64 KB hypervectors for extended context. Title only, no technical depth.
5/19/2026
AI Coding Feels Like Using an Unreliable Compiler
Comparison of AI coding assistance to unreliable compiler behavior. Title only.
5/19/2026
LLMs: code is just untrusted text until you validate it
Article about code validation and security when using LLM-generated code. Title only.
5/19/2026
Make-No-Mistakes
Open-source tool called make-no-mistakes aims to prevent logic failures in applications. Limited technical details provided.
5/19/2026
MagnoApi - i built a Memory Context API that gives memory to llms
Memory context API for LLMs. Title only, no technical details provided.
5/19/2026
Show HN: CANviz – open-source CAN analyzer, J1939/CANopen, pip install
CANviz: Open-source browser-based CAN bus analyzer. Pip install, works with USB adapters under $10.
5/19/2026
AI coding made us faster. Why did incidents increase?
Analysis of why AI-assisted coding increased development speed but also increased production incidents. Title only.
5/19/2026
Raven Software Released Jedi Academy Source Code in 2013 Amid Dev Crunch Rage
Star Wars Jedi Academy source code released under GPLv2. Historical open-source release from 2013.
5/19/2026
I tried every AI memory tool. Here's why I built another one
Memory infrastructure tool for AI agents with typed, audited, decay-aware storage instead of flat vector stores.
5/19/2026
Show HN: How to analyze your LLM output – A behavioural health monitor for LLMs
Open source behavioral health monitoring tool for LLMs and AI agents using Posture Sequence Analysis framework.
5/19/2026
When More Context Makes LLM Agents Worse
Research on how increased context negatively impacts LLM agent performance. Title only, insufficient content.
5/19/2026
AI Empowered PM
Analysis of product manager roles evolving with AI, discussing LLM training, inference, fine-tuning, and RAG.
5/19/2026
DPlus – AI media buyer that plans and launches campaigns across 12 ad platforms
AI platform that automates ad campaign creation and management across 12 ad networks.
5/19/2026
Find bugs in YOUR code using OpenCode, Llama.cpp and Qwen3.6
Experiments using llama.cpp and Qwen 3.6 LLM for automated code bug detection and patch classification tasks.
5/19/2026
The Librarian – 7.1M-node knowledge graph beats vector search
Knowledge graph system with 7.1M nodes for multimodal PDF processing and conversational queries. LLM application.
5/19/2026
Alloy: Code/SDK Generation Framework
Alloy: Code generation framework using JSX/templates, inspired by React. Handles imports, formatting, syntax generation.
5/19/2026
Free tool to build Options Apps with your coding agent
Desktop app for building financial data apps with AI coding agents (Claude, Codex). Free beta. Includes options data.