5/22/2026
Show HN: Cq exchange: Hosted knowledge commons for AI coding agents
Hosted knowledge commons platform designed for AI coding agents to share and access information.
5/22/2026
Fixing LLM Writing with Distribution Fine Tuning
Technique for improving LLM writing quality through distribution fine-tuning (limited details provided).
5/22/2026
Show HN: Mercury v3 – Convert Python Notebooks to Web Apps
Framework converting Python notebooks to interactive web apps with support for chats, AI agents, dashboards and reports using reactive cell re-execution.
5/22/2026
Show HN: Runner – desktop app for running Claude Code, Codex as a crew
Desktop app for orchestrating crews of CLI coding agents (Claude Code, Codex, custom shells) in a unified workspace instead of multiple terminals.
5/22/2026
A Marketplace of Fine Tuned SLMs for Agentic Tasks
Marketplace platform offering fine-tuned small language models optimized for agentic tasks.
5/22/2026
AI Dev 26 x SF | Brandon Waselnuk: Building the Context Engine AI Agents Need
Talk on context engines for AI agents; insufficient technical detail provided.
5/22/2026
Ask HN: Are LLMs creating busy work?
Discussion of LLM-assisted work creating busywork rather than efficiency gains. Challenges whether token-burning equals progress in software development.
5/22/2026
Ceres, an open copilot for VS Code with budget and local LLMs support
Open source VS Code copilot supporting budget and local LLM models for code completion.
5/22/2026
AI Dev 26 x SF | Jerry Liu: My Agent Can't Read a PDF?
Talk on PDF reading limitations in AI agents; lacks detailed technical content in provided excerpt.
5/22/2026
Continuous Development still has prerequisites
Analysis of engineering challenges in rapid agentic code deployment; discusses prerequisites beyond generation speed.
5/22/2026
Ask HN: What is the least sycophantic frontier LLM?
User survey asking about sycophancy and objectivity differences between frontier LLMs.
5/22/2026
Building Context-Aware Search in Python with LLM Embeddings and Metadata
Tutorial on building context-aware search using Python, LLM embeddings and metadata filtering. Practical developer tool application.
5/22/2026
Ask HN: We need a standard way to say how much AI was used in a PR
Discussion proposing standardized notation for documenting AI involvement level in pull requests.
5/22/2026
Show HN: Videowright – we made our launch video with Claude Code (open source)
Videowright tool generating demo/explainer videos from descriptions using Claude Code agent.
5/22/2026
AI Edits a Classical Chinese Paper: Multi-Model Stress Test
Multi-model AI stress test editing a classical Chinese academic paper for quality and accuracy.
5/22/2026
Show HN: LLM-mock – Record real LLM API responses once, replay them in tests
Open source testing tool for recording and replaying real LLM API responses in tests.
5/22/2026
Diia - Ukraine gov app launched AI agent based on Google Gemini
Ukraine's Diia government app deployed an AI agent powered by Google Gemini for citizen services.
5/22/2026
Introducing Nova, our internal platform for coding agents
Dropbox's internal coding agent platform supporting monorepo context, validation, and engineering lifecycle integration.
5/22/2026
Show HN: Mnestica – a memory system for notes, reading, and cards
Learning tool converting notes/articles into spaced repetition flashcards with AI-generated cards.
5/22/2026
Which flavor of software are LLMs exactly?
Essay examining LLMs as a software paradigm, building on Software 2.0 concept by Karpathy.
5/22/2026
What 1,192 conversations taught us about knowledge base search in AI agents
Analysis of 1,192 real conversations revealing insights and best practices for knowledge base search functionality in AI agents.
5/22/2026
Show HN: Myco – coordinate Claude and DeepSeek and other LLMs in one agent swarm
Framework for coordinating multiple LLMs (Claude, DeepSeek, others) in a unified agent swarm.
5/22/2026
An Android UI dump for LLMs (10× fewer tokens, same actions)
Optimized XML representation for Android UI automation reducing token usage 8-10× for LLM agents while maintaining decision quality.
5/22/2026
Deep codebase context cuts Claude Code's token cost by 47%
Claude Code achieves 47% token cost reduction through improved deep codebase context implementation.
5/22/2026
Twelve Ways to Be Wrong About AI-Assisted Coding
Technical analysis of twelve common misconceptions when evaluating AI-assisted coding tools, with explanations of flawed assessment methods.
5/22/2026
Nyx Wave: an AI agent that extracted expert knowledge via email
Analysis of 'Nyx Wave' AI agent that conducted structured knowledge extraction via email over 10+ messages. Pattern extraction visible without model access.
5/22/2026
Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark
Benchmark comparing AI coding tools (Claude, Cursor, Google Antigravity) on OpenSCAD 3D model generation task. Antigravity 2.0 achieved best results.
5/22/2026
Thermo-Nuclear Code Quality Review (Cursor Team Kit skill)
Cursor IDE skill for strict code quality reviews emphasizing implementation quality, maintainability, and structural refactoring opportunities.
5/22/2026
Show HN: I made an open-source memory layer for agents
Open source memory layer for AI agents with persistent storage, graph relationships, semantic search, and optional self-hosted infrastructure.
5/22/2026
Proton Pass for AI Agents
Proton Pass adds credential sharing via AI access tokens, allowing secure delegation of account access to AI agents with activity monitoring.
5/22/2026
Live 204-node MoE visualization reveals emergent cognitive stratification
Ternary balanced-logic systems programming language and inference runtime with Mixture-of-Experts language model trained natively using trit logic.
5/22/2026
I built a Chrome extension to reduce ChatGPT UI lag
Chrome extension that reduces UI lag on ChatGPT by applying content-visibility optimization and managing CodeMirror listeners. No data transmission.
5/22/2026
Show HN: Interactive and Stylized AI Chat
Browser tool that renders AI chat responses as rich interactive components including cards, diagrams, foldouts, and styled layouts.
5/22/2026
Show HN: An open source tool that checks for post-quantum crypto
Open source tool to scan hosts and check cryptographic algorithms for post-quantum compliance readiness.
5/22/2026
Google makes Gemini 3.5 Flash the default AI model for billions of users
Google deploys Gemini 3.5 Flash as default model in Gemini app and Search, claiming faster performance and lower cost than 3.1 Pro.
5/22/2026
NVCF Is Now Open Source: Inside Nvidia's GPU Function Platform
NVIDIA open-sourced NVCF platform under Apache 2.0, including control plane, compute plane, CLIs, and Helm charts for GPU function infrastructure.
5/22/2026
AI coding agents and the evolution of developer skills by 2026
Analysis of agentic coding tools in 2026. Majority of code from agents at some orgs. Key skill gap: evaluating AI output rigorously like junior dev PRs.
5/22/2026
Lucy – pay-per-task AI agent in USDC, no subscription (A2A/MCP/x402)
Lucy: autonomous pay-per-task AI agent using USDC stablecoin. Works via Telegram for humans and A2A/MCP protocols for agents, no subscription model.
5/22/2026
AI dev tools: Cost, ROI, and budgeting for 2026
Survey of 900+ engineers (April 2026) on AI dev tool spending. 30% hit usage limits; most lack ROI frameworks; teams with success metrics see better results.
5/22/2026
A local-first multimodal knowledge platform for managing entities
Desktop application and CLI for managing entities, relationships, and spatial data as typed graphs with Prolog rules engine and blob storage. Master's thesis on knowledge representation.
5/22/2026
Why Svelte Is Better Than React in the Agentic Era
Analysis comparing Svelte vs React for AI agent-assisted development. Argues Svelte's simplicity better suits AI code generation and reasoning.
5/22/2026
CodeAlta – a terminal workspace for agentic coding
CodeAlta: terminal workspace for agentic coding with model provider setup, project navigation, prompt attachments, threaded sessions, delegated agents. CLI in .NET/C#.
5/22/2026
SEO Survival Kit – Claude Code Plugin for Core Update Recovery
Claude Code plugin for SEO recovery. Provides 6 skills and decision-maker-friendly reports for websites hit by search ranking drops.
5/22/2026
Bytewax: Stream processing library built using Python and Rust
Bytewax (Python/Rust stream processing library) transitioned to community maintenance. Original company ceased operations May 2025; seeking new maintainers.
5/22/2026
Show HN: I threw away my analytics dashboard and replaced it with 42 MCP tools
Developer rebuilt analytics tool using 42 MCP tools and Claude Code agents instead of traditional dashboard, focusing on headless architecture.
5/22/2026
CodeAlta an efficient agentic AI coding CLI assistant coded in C#/.NET
CodeAlta: keyboard-first terminal AI coding workspace for project/model/thread/plugin management. Requires .NET 10, emphasizes compact design.
5/22/2026
Cisco used AI to write security incident reports, with mixed results
Cisco applied AI to generate security incident reports with mixed effectiveness results.
5/22/2026
Roundtables: Can AI Learn to Understand the World?
MIT subscriber-only discussion on world models and AI's understanding of physical environments.
5/22/2026
OMLX v0.3.9 Stable Merges Native MTP (Multi-Token Prediction)
OMLX v0.3.9 stable release adds multi-token prediction for faster LLM decoding.
5/22/2026
A.I. Learns to Optimize Factorio Blueprints [video]
Video demonstrating AI optimization of game production blueprints in Factorio.