3/8/2026
SCRY 17-source research engine for Claude Code(no API keys, pure stdlib)
SCRY: multi-source research engine for Claude Code that searches 17 sources in parallel without API keys, using pure Python stdlib.
3/8/2026
Show HN: Cursor skill for Claude Code's /loop scheduler
Claude Code /loop scheduler for running prompts on recurring or one-time schedules with cron-style timing.
3/8/2026
Show HN: Go LLM inference with a Vulkan GPU back end that beats Ollama's CUDA
Go-based LLM inference engine with Vulkan GPU backend, 28% faster than Ollama CUDA on some models.
3/8/2026
New Research Reassesses the Value of Agents.md Files for AI Coding
ETH Zurich research paper questioning the effectiveness of AGENTS.md files in AI coding agents, recommending minimal context files.
3/8/2026
Show HN: TracePact – Catch tool-call regressions in AI agents before prod
TracePact: Open-source tool for detecting tool-call regressions in AI agents by comparing cassette recordings of execution traces.
3/8/2026
Add llms.txt and fix robots.txt for AI agent discoverability
Best practice for AI agent discoverability: adding llms.txt file and fixing robots.txt configuration to make websites visible to LLM crawlers.
3/8/2026
Show HN: TTS.ai - Text to Speech
TTS.ai: Aggregator of 20+ open-source text-to-speech, transcription, and voice generation models with 107+ voices across 32+ languages.
3/8/2026
MCP vs. CLI for AI Agents
Technical comparison of MCP (Model Context Protocol) vs CLI approaches for AI agent tool integration, analyzing tradeoffs in token usage and composability.
3/8/2026
Ivy – Bringing LLMs to 35M offline students in Ethiopia
Ivy: Edge-inference educational AI copilot optimized to run offline on $150 Android devices for 35M students in Ethiopia with no internet access.
3/8/2026
Show HN: Render Claude Code and Codex Transcripts as Browsable HTML
Tool to render Claude Code and Codex transcript sessions as interactive browsable HTML.
3/8/2026
What Production AI APIs Need Beyond Response = LLM(prompt)
Analysis of production requirements for LLM APIs beyond basic prompt-response patterns.
3/8/2026
Show HN: Booklet AI – AI-Powered Digital Flipbook Creator
AI-powered booklet generator using transformer models to research and create content automatically.
3/8/2026
Show HN: Strata – 31-43% cheaper Claude Code reads via entropy, no parser
Entropy-based optimization reducing Claude Code API costs by 31-43% without parsing.
3/8/2026
Experiment That Predicted How AI Agents Would Cooperate
Research on multi-agent system cooperation focusing on information topology as infrastructure primitive, with controlled experiments isolating information flow as independent variable.
3/8/2026
Show HN: Beecon Infrastructure as Intent, open source IaC built for AI agents
Open-source Infrastructure-as-Intent framework designed for AI agents to manage cloud resources.
3/8/2026
Show HN: Scan0tron – AI screen capture that auto-fills forms ($49)
Desktop automation tool combining computer vision and LLM for form-filling and screen interaction tasks.
3/8/2026
AI agent freed itself and started mining crypto
Alibaba research paper documents AI agent deviating from instructions by autonomously mining cryptocurrency, highlighting safety risks in agent autonomy.
3/8/2026
Good Vibes, Bad Vendors: AI Coding Works Now
Opinion piece arguing AI coding tools now work effectively, with anecdotal framing about high school bullying.
3/8/2026
Show HN: RotaGuide Spotlight – Markdown based UI guide system for web apps
Open-source markdown-based UI guide library for adding step-by-step tutorials to web applications.
3/8/2026
Patching minified Claude Code so it can hear webhooks
Technique for patching minified Claude Code to enable webhook listening capability.
3/8/2026
Show HN: Qry – CLI web search that always outputs JSON, with swappable back ends
CLI web search tool with JSON output and pluggable adapters, designed for composability with agents and scripts.
3/8/2026
Show HN: SafeAgent – exactly-once execution guard for AI agent side effects
Python library protecting AI agent side effects from retries, preventing duplicate actions in tool calls via idempotency mechanisms.
3/8/2026
Show HN: Open-source personal finance AI that runs locally on your laptop
Open-source personal finance application using Claude, OpenAI, or local Ollama for transaction categorization, tax estimation, and portfolio monitoring.
3/8/2026
Plenty of AI hype, but not much useful software?
Discussion questioning whether AI productivity gains translate to measurable increase in useful software projects and SaaS tools.
3/8/2026
All Bench Leaderboard for Comparing LLMs Across Benchmarks
Analysis of LLM evaluation landscape fragmentation due to benchmark saturation; proposes unified leaderboard comparing models across multiple hard benchmarks.
3/8/2026
Armin Ronacher on AI Agents and the Future of Programming [video]
Video interview with Armin Ronacher on AI agents and future of programming.
3/8/2026
Agentmarketpro
Platform where specialized AI agents handle tasks autonomously and escalate to humans when needed, demonstrating human-AI team collaboration.
3/8/2026
autoresearch: AI agents running research on single-GPU nanochat training automatically
autoresearch: Framework for autonomous AI agents to conduct machine learning research on single-GPU hardware automatically. Satirical but discusses agent autonomy.
3/8/2026
AI found us before Google did
Discussion of company receiving leads from Gemini before Google indexed site, suggesting LLMs may surface content through different discovery mechanisms.
3/8/2026
Show HN: Pappardelle, a TUI for Multi-Clauding
Pappardelle: TUI developer tool orchestrating Claude Code with Git, Linear/Jira, and tmux for multi-agent coding workflows.
3/8/2026
The Great AI Arbitrage
Analysis of economic arbitrage in software development where AI reduces production costs while client pricing remains unchanged.
3/8/2026
Show HN: A dynamic, crowdsourced benchmark for AI agents
Open-source crowdsourced benchmark arena for AI agents with Elo ratings, leaderboard, and community-authored challenges.
3/8/2026
Show HN: SiClaw – Open-source AIOps with a hypothesis-driven diagnostic engine
Open-source AIOps platform with AI agent for infrastructure diagnostics, read-only analysis with change management integration.
3/8/2026
AirLLM optimizes inference memory usage
AirLLM reduces LLM inference memory usage, enabling 70B models on 4GB GPU and 405B on 8GB without quantization.
3/8/2026
Show HN: DailyDefense – Daily tower defense for agents or humans
Tower defense game designed as benchmark environment for AI agents with daily seeded challenges.
3/8/2026
Ask HN: How to serve inference as we do with containes with cached token
Question about serving LLM inference with vLLM and token caching for GPU-constrained environments.
3/8/2026
Agent-town – A pixel-art AI agent online collaboration platform
Agent-town visualizes AI agent orchestration as collaborative pixel-art office environment. Novel UI for multi-agent systems.
3/8/2026
Show HN: Apc-CLI – sync AI memory across Claude Code, Cursor, Copilot
apc-cli tool syncs AI context/memory across Claude, Cursor, Copilot. Unifies multiple AI coding tools with shared configuration.
3/8/2026
ClawPurse Micropayment Ecosystem
ClawPurse micropayment system for DePIN and agentic AI with encrypted wallet and on-chain verification. Developer tool for agent payments.
3/8/2026
AI Powered Exploit Kit
CyberStrikeAI is a Go-based security testing platform with 100+ integrated tools, AI orchestration, and MCP protocol support for automated vulnerability assessment.
3/8/2026
Show HN: Joy – Open trust network for AI agents (AI-to-AI vouching)
Joy is a decentralized trust network enabling AI agents to build reputation through vouches and MCP server discovery with verified endpoint ownership.
3/7/2026
Why developers using AI are working longer hours
Study examining how AI-assisted development increases code output but extends developer work hours due to debugging AI-generated code issues.
3/7/2026
How to run Qwen 3.5 locally
Guide for running Alibaba's Qwen 3.5 LLM family locally, covering models from 0.8B to 397B parameters with multimodal reasoning and agentic coding capabilities.
3/7/2026
Show HN: PolicyCortex – AI agent that autonomously remediates cloud misconfigs
PolicyCortex is an AI agent platform automating NIST 800-171 compliance enforcement and MITRE ATLAS threat detection across cloud infrastructure for defense organizations.
3/7/2026
OpenAI GPT-5.4 Explained
Announcement of OpenAI GPT-5.4 flagship model with advanced reasoning, improved coding, tool use, and native computer-use capabilities across APIs and products.
3/7/2026
Show HN: Beam Protocol – SMTP for AI Agents (natural language agent-to-agent)
Beam Protocol is an SMTP-like standard enabling agent-to-agent communication with global addressing, authentication, and discovery for autonomous AI agents.
3/7/2026
AI SAd-ware
Discussion of Greywall, a sandboxing tool for controlling AI agent network access and file system permissions to block unwanted advertisements.
3/7/2026
Ask HN: How to scale agent systems when Layer 7 is unreliable?
HN discussion on scaling agent systems with unreliable Layer 7: rate limiting, retries, circuit breakers, resuming workflows.
3/7/2026
Show HN: Modulus – Run multiple coding agents with shared project memory
Desktop app enabling multiple coding agents to collaborate with shared project memory across repositories.
3/7/2026
Show HN: PolyClaude – Using math to pay less for Claude Code
PolyClaude: Tool optimizing Claude Pro account rotation to work around 5-hour rate limits through mathematically-timed activation.