HN tar-ive 18d ago

Promptetheus – Trace, detect, and auto-repair AI agent failures

Promptetheus: Python SDK and debugging infrastructure for tracing, detecting, and auto-repairing AI agent failures.

HN mzubairtahir 18d ago

Ask HN: How much memory is useable by GPU in MacBook?

Lore: Open source version control system for projects with large binary assets, optimized for scalability and team collaboration.

HN sonabinu 18d ago

Modern GPU Programming for MLSys

Technical guide on modern GPU kernel optimization for ML systems: attention, LLM prefill/decode, fused operations.

HN coolwulf 18d ago

Coding agent harness written in native Golang with built-in file and Git viewer

Golang-based coding agent harness with file and Git viewer for agent execution.

HN StizzurpXDD 18d ago

Isn't US Government trying to monopolize AI as a super power?

Speculation on US government restricting advanced AI models to trusted organizations.

HN mattmerrick 18d ago

Find the right AI agents to build

Agent Idea Hub provides ranked blueprints for building AI agents for business workflows.

HN coolwulf 18d ago

TUI email client in native Golang with LLM based drafting functions

Terminal email client in Golang with LLM-powered reply drafting using DeepSeek V4 Pro.

HN swolpers 18d ago

US allows Anthropic to release Mythos AI to 'trusted' US organizations

Anthropic's Mythos AI approved for limited release to trusted US organizations.

HN mzubairtahir 18d ago

Ask HN: MacBook vs. Dedicated GPU for LLM

HN discussion on running open-weight LLMs on MacBooks vs. dedicated GPUs; GLM 5.2 mentioned.

HN modinfo 18d ago

Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding

Research comparing 67 LLM models shows ensemble approaches rarely outperform best individual models. Empirical ML analysis.

HN toomuchtodo 18d ago

How China Is Gutting Western Automakers (2025)

Analysis of China's strategy to dominate global auto industry through manufacturing scale and exports.

HN logn 18d ago

Show HN: Boot a tiny app-making OS inside an AI chat

BlueBookOS: A microkernel and rule set for building apps directly inside AI chat interfaces, enabling source-first development in live sandbox environments.

HN pro_methe5 18d ago

Show HN: Tested – AI Tools Scored by a Panel of LLMs (Claude, GPT, Gemini, Grok)

Tool ranking platform using panel of AI agents (Claude, GPT, Gemini, Grok) to independently score 40+ AI tools. Transparent scoring methodology.

HN sudo_cowsay 18d ago

Show HN: GetBlocked – a local-only Chrome extension for blocking web trackers

Chrome extension blocking third-party trackers and cleaning tracking links locally. Privacy tool unrelated to core interests.

HN euan21 18d ago

Show HN: I scanned 87 MCP servers for agent-authority hygiene – leaderboard

Security audit of 87 MCP servers using deterministic capframe rule engine, scoring agent authority and hygiene with public methodology and open-source rules.

HN PaybackTony 18d ago

Show HN: Hacker News on a Train Station Style Flip Board

Fun project displaying Hacker News feed on flip board animation. Unrelated to AI/ML interests.

HN insightits 18d ago

PrismLib – semantic LLM cache and cluster mesh that cuts token spend

In-process semantic LLM cache with tensor operations and distributed cluster support. Reduces token spend via paraphrase matching and multi-tenant isolation.

HN logn 18d ago

Show HN: Boot a tiny app-making OS inside an AI chat

BlueBookOS enables building apps through AI chat using RAu instruction language and artifact contracts for structured app development.

HN rghosh8 18d ago

Utility for Multi-GPU Node Configuration

Installer utility for configuring multi-GPU nodes (A100/H100) on Ubuntu. Infrastructure tool without AI/ML focus.

HN flashbrew 18d ago

Add MCP Apps to Your AI SDK Application

Tutorial on building MCP Apps host with AI SDK to render interactive tool UIs from Model Context Protocol tools in sandboxed iframes.

HN anonu 18d ago

AI Powered Photo Gallery Without the Cloud

Local photo curation tool that scores family photos on sharpness, lighting, faces, and composition without cloud uploads.

HN philosopherr 18d ago

I removed the vector database from my AI agent stack

Moss is a sub-10ms semantic search runtime for conversational AI agents with hybrid retrieval, built-in embeddings, and WebAssembly browser support.

HN krackers 18d ago

Thinking to recall: How reasoning unlocks parametric knowledge in LLMs

Google Research: Study on how reasoning helps LLMs recall facts through latent computation and related fact priming. ML research on LLM mechanisms.

HN mirhacker 18d ago

ComAI – An open-source Linux assistant for troubleshooting with local LLMs

Open-source Linux terminal assistant using local LLMs or APIs for command explanation, log analysis, and troubleshooting. Supports Ollama and OpenAI.

HN E-Reverance 18d ago

Forward Self Models

Research on forward self-models: small networks predicting neural network layer activations. Achieves 97% similarity with 1-3% parameter overhead on 30M-1B models.

HN phoronixrly 18d ago

Steam Controller auto-charge with computer vision tracking

Open-source project using computer vision and optical flow to auto-pilot Steam Controller into charging dock. Hardware project unrelated to core interests.

HN dataking 18d ago

Trump admin allows Anthropic to release Mythos AI model to some companies

U.S. government approves Anthropic to release Mythos 5 model to roughly 100 companies and federal agencies.

HN bobrenjc93 18d ago

U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

US government approves Anthropic to release Mythos AI model to trusted US organizations.

HN ra0x3 18d ago

Ask HN: Any OSS models as good as GPT-4o-mini?

Discussion thread asking for open-source LLM alternatives comparable to GPT-4o-mini. No substantive content provided.

HN erikqu 18d ago

Workbench: A TUI for parallel coding agents

TUI workbench for running multiple coding agents (Claude, Gemini, Goose) in parallel with integrated file explorer and viewers. Built with Bun and React 19.

HN edanm 18d ago

Assessing GPT-5.6 Sol Against Cybersecurity Benchmarks

Evaluation of hypothetical GPT-5.6 Sol model against offensive cybersecurity benchmarks. Contains future date (2026), likely fictional or speculative content.

HN bennyjiang 18d ago

Show HN: Skillmaxxing – make every agent self-evolving

Open-source agent plugin enabling self-improvement by reflecting on work and creating reusable skills. Reduces token usage and improves execution efficiency.

HN recsv-heredoc 18d ago

Ask HN: Model access depends on citizenship. What should Non-US founders do?

Discussion of LLM access restrictions based on citizenship affecting non-US founders and model availability across Anthropic and OpenAI.

HN longtermop 18d ago

Show HN: All-in-one memory for AI Agents

Show HN submission for all-in-one memory system for AI agents with minimal details provided.

LB cleverhans.io via zurtar 18d ago

AI Agents Enable Adaptive Computer Worms

Research paper on how AI agents enable adaptive computer worms with evasion capabilities.

HN Decag 18d ago

Skill for generating cheatsheet PDF optimized for the reMarkable eink

Agent skill that generates bookmarked PDF cheatsheets optimized for reMarkable e-ink tablets with single-page subject organization.

HN davidmckayv 18d ago

OpenTag: An open-source alternative to Claude in Slack

OpenTag is open-source self-hosted AI agent for Slack that reads threads, calls tools, and renders results without per-seat pricing.

HN gcanyon 18d ago

The open source DOCX editor submitted to HN a few weeks ago has been deleted

Report that an open-source DOCX editor GitHub repository and website have been deleted with unknown reasons.

HN sona-coffee11 18d ago

AI audio translator with speech-to-text, LLM translation, and text-to-speech

AI audio translation system combining speech-to-text, LLM translation, and text-to-speech for multi-language audio processing.

HN todience 18d ago

Show HN: Even, the terminal-first desktop workspace

Even: terminal-first desktop workspace for managing multiple development projects with built-in browser and designed for AI code agent workflows.

HN SuboptimalEng 18d ago

Ask HN: Can distributed data centers in individual households provide UBI?

Discussion proposal for distributed household data centers with GPU clusters as compensation model for UBI-like income.

HN SEC_Lense 18d ago

Show HN: Forensic stock analysis from SEC filings, no LLM guessing (free)

Deterministic financial analysis tool for SEC filings detecting fraud signals without using LLMs, with historical backtests on failed companies.

HN valisvalis 18d ago

Show HN: Deskmate Live – AI Desktop Pet Companions

Desktop AI pet companion app featuring language learning, homework help, and screen context awareness.

HN stoicstoic 18d ago

AgentKits – 60 production-ready AI agent blueprints with guardrails

Collection of 60 production-ready AI agent blueprints with prompts, architecture, and tools across 30 categories, implementation-agnostic.

HN joecobb 18d ago

Evan's Jujutsu Tutorial

Tutorial for Jujutsu version control system, a simpler and more powerful alternative to Git written by Evan Martin.

HN kkm 18d ago

Prediction: A Frontier open-source LLM Will Be Released On 3rd December 2026

Analysis of performance gap between open-source and closed-source LLMs, predicting frontier open-source model release by December 2026.

HN franze 18d ago

Show HN: I build an app for you to take a break from AI. RainBreak

macOS menu-bar app providing rain sounds and break timer for users working with AI agents.

HN datsci_est_2015 18d ago

Ask HN: Which AI concepts are here to stay, and which will churn?

Discussion asking which emerging AI concepts will persist versus fade, including agents, agentic workflows, and Model Context Protocol.

HN Bender 18d ago

Linux Foundation and Others Launch Akrites Defend Open-Source from AI Exploits

Linux Foundation and others launch Akrites initiative to defend open-source projects from AI-based exploits.

HN kristianp 18d ago

2027 Macs to Get AI-Focused M7 Chips as Apple Skips High-End M6

Apple plans M7 chips optimized for AI workloads launching 2027, skipping high-end M6 variants.