HN garymiklos 19d ago

Show HN: NoMac – let your AI agent ship iOS apps without a Mac

Show HN: NoMac builds, tests, and ships iOS apps via cloud Macs without local hardware. Agent-first with TestFlight/App Store integration.

HN jrud 19d ago

The fix for my AI agent wasn't making it smarteR

Title only. Personal experience improving AI agent performance without scaling model size.

HN EGreg 19d ago

Generalizing LLMs to PLTs: Probabilistic Language Tries

Research generalizing language models to probabilistic language tries as alternative architecture.

HN hackeratrandom 19d ago

Litert.js, Google's High Performance Web AI Inference

Google releases Litert.js, a JavaScript library for efficient AI inference in web browsers.

HN spillcoffee 19d ago

Lessons from designing MCP tools for AI agents

Design patterns and best practices for building Model Context Protocol tools for AI agents.

HN oli266 19d ago

I made an LLM think in latent space. The model never read a single thought

Research on LLM internal reasoning in latent space without generating intermediate text outputs.

HN diptanu 19d ago

Zero-copy TLS ingress with kTLS and splice(2) for sandboxes

Performance optimization for sandbox infrastructure using kTLS and zero-copy techniques to support agent workloads.

HN Aarav03790 19d ago

Show HN: I made AI agent that deploys your thing on diff plats automaticaly

AI agent that automatically deploys applications across multiple platforms (GitHub, DigitalOcean, Vercel, Render, Stripe).

HN mavster26 19d ago

Show HN: Token Governor – The cheapest LLM token is the one you never send

Tool to optimize LLM token usage by reducing unnecessary token sends, focusing on cost reduction strategies.

HN LorenDB 19d ago

OpenAI's Atlas browser doesn't make it to its first birthday

OpenAI discontinued Atlas AI browser after 11 months, integrating browser agent features into ChatGPT Work platform.

HN antinucleon 19d ago

What Muse Spark 1.1 Taught Us About Enterprise Agent Architecture

Design patterns and architecture lessons from Muse Spark 1.1 enterprise AI agent deployment.

HN skillrail 19d ago

Skillrail: One versioned source for AI agent skills, synced to every tool

Platform for versioning and syncing AI agent skill definitions across multiple tools (Claude, Cursor, Copilot, Windsurf).

HN abelaer 19d ago

Truth is not a direction: a Tarski attack on LLM probes

Research paper analyzing LLM probes using Tarski's work, questioning whether truth is directional in model representations.

HN rcarmo 19d ago

AI as a weapon of mass cognitive destruction

Critical perspective on AI-generated content proliferation and its cognitive impact on organizations.

HN skyphusion 19d ago

Show HN: Postern, an email system for both humans and agents

Email system enabling AI agents to have addresses and respond to emails with unified inbox and optional Vectorize integration on Cloudflare Workers.

HN Brajeshwar 19d ago

Patreon Blocks Crawlers from Stealing Creators' Work for AI Training

Patreon implements crawler blocking to prevent AI training data theft. Data protection measure.

HN lumli 19d ago

Show HN: Tiny ONNX Runtime Web Worker Starter

Starter template for running ONNX Runtime in browser Web Workers to keep UI thread responsive during model inference on mobile and desktop.

HN edifil 19d ago

Ask HN: Thoughts on a MCP to manage cloud and AI spend?

Discussion about MCP protocol for monitoring and controlling AI agent spending and cloud infrastructure costs.

HN grenzfrei 19d ago

Show HN: CiteReady – Can AI search (ChatGPT, Perplexity) cite your site?

SEO audit tool checking whether AI search engines (ChatGPT, Perplexity) can crawl and cite websites.

HN hardmaru 19d ago

The AI Picbreeder Experiment

Explores whether AI agents can be creative and interact autonomously, treating agents as model organisms for cultural production rather than tools.

HN vinhnx 19d ago

The Spec Ceiling: Why AI Coding Speed Moves the Bottleneck to Product Discovery

Essay on how AI coding speed shifts bottlenecks from implementation to product specification. Development insights.

HN brryant 19d ago

AI Web Design (Opus vs. Sol)

Practical guide comparing Claude Opus and Sol for web design. Technical tips on prompting and model performance.

HN parweb 19d ago

Show HN: My AI agent has 1 hour to earn €10 – it shipped this page itself

Autonomous AI agent (Claude) completed income-generating task in 1 hour using Stripe and web deployment. Practical agent demo.

HN zknill 19d ago

AI Transport v0.5.0: durable execution with Steps

AI Transport v0.5.0 release adds durable execution framework support for agents with resumable streams. Technical tool update.

HN coder_xyz 19d ago

Show HN: Hallint – lint AI-generated code for security issues

Hallint: static analysis linter for AI-generated code security issues. Catches common LLM-introduced bugs like hardcoded secrets and SQL injection.

HN amichail 19d ago

Predict AI's impact on creativity by studying what Python did to programming

Essay comparing Python's impact on programming to predict AI's impact on creative fields.

HN harpreetseehra 19d ago

MCP Server on Edge Compute 167 Lines of Python

Implementation of Model Context Protocol server on edge compute in 167 lines of Python.

HN faDOOM 19d ago

Negative squaring – pre-tilted 3-bit quantization beat naive 4-bit

Research on 3-bit quantization technique achieving better performance than naive 4-bit approaches for model compression.

HN mhmt_dmr 19d ago

Show HN: MCP Gateway – Turn existing APIs and databases into MCP servers

Open-source gateway tool converting existing APIs and databases into MCP servers for AI agent integration without custom development.

HN mrunix 19d ago

Hardening stb libraries using AI agents

Using AI agents to improve security hardening of stb libraries through automated vulnerability detection.

HN alex_brajkovic 19d ago

LLMs are bad at novelty, but that is our chance to Singularity

Philosophical exploration of LLM limitations in novelty generation and creative capability.

HN Eapz_06 19d ago

Production Data and AI Agents

Discussion about safe database access patterns for AI agents in production environments.

HN lamne 19d ago

20M data entries, a backtest engine, ML models, agents. Still haven't traded

AI agent system converting natural language to tradeable rules with backtesting engine against 20M data entries and historical scenarios.

HN ermantrout 19d ago

Unified Memory, Explained: Why Mini PCs Can Run 70B Models a Big GPU Can't

Explains unified memory architecture enabling mini PCs to run 70B parameter models vs larger GPUs.

HN noah1995 19d ago

Show HN: A possible open-source desktop alternative to OpenAI Codex

Open-source desktop tool mimicking OpenAI Codex functionality, supporting code generation and writing tasks across Windows/Linux.

HN teleforce 19d ago

Exploiting Sparsity for Long Context Inference: Million Token on Commodity GPUs

Research on exploiting sparsity patterns for million-token LLM inference on consumer GPUs.

HN shumway 19d ago

CorvinOS – self-hosted OS for AI agents, compliance built into the runtime

CorvinOS is a self-hosted operating system runtime for AI agents with compliance auditing, supporting multi-messenger deployment, browser automation, RAG, and voice control.

HN theanonymousone 19d ago

DeepSWE Benchmark Results for GPT 5.6

DeepSWE is a long-horizon software engineering benchmark for evaluating frontier coding agents on complex, original tasks, addressing saturation in existing benchmarks.

HN ashu_trv 19d ago

Record and Replay, teach AI agents desktop workflows by showing them once

MCP server that records desktop workflows once and compiles them into reusable agent skills for automation; enables agents to perform tasks by example.

HN quantumgarbage 19d ago

The triage is the product: running AI agents against Ethereum's protocol code

Case study of deploying AI agents to analyze and triage Ethereum protocol code for security and optimization purposes.

HN mpfect 19d ago

Crawl4AI: Open-source web crawler and scraper for LLMs

Open-source web crawler and scraper designed to extract data for LLM applications.

HN HalaDefense 19d ago

Verity – Chrome extension that auto fact-checks AI responses with cited sources

Verity is a Chrome extension that automatically fact-checks AI responses from Claude, ChatGPT, Gemini and Grok using web sources to detect hallucinations.

HN s03nk3 19d ago

ADDing agent-context annotations for existing API specs

ADD standard enabling IoT devices to self-describe to AI systems via JSON documents for safe agent integration without custom code.

HN gshutler 19d ago

Why LLMs get dates and times wrong (and how to fix it)

Technical analysis of LLM temporal reasoning failures and solutions using temporal infrastructure in scheduling agent systems.

HN Present_Flow 19d ago

Show HN: Makoto the Claude Code plugin that keeps the AI honest

Makoto plugin monitoring Claude Code agent tool calls for truthfulness, blocking fabricated actions via integrity verification.

HN crowd51 19d ago

China may restrict foreign access to Chinese open-source AI models

China may restrict foreign access to Chinese open-source AI models.

HN Krishnaswaroop 19d ago

Why Industry 5.0 is more than AI-powered analytics

Panel discussion on Industry 5.0 and AI-led decision-making in Indian manufacturing sector.

HN sachamorard 19d ago

Promising inference software, written in Zig

ZML/LLMD: Self-contained LLM inference server supporting LLaMa/Gemma/Qwen/Mistral across 5 architectures with continuous batching and tool calling.

HN vkaku 19d ago

TinyToT – Tree of Thoughts Inference Server

TinyToT: Ollama-compatible inference server using knowledge retrieval and reasoning chains, 97% benchmark score, runs on CPU in <1ms.

HN balivandi 19d ago

Ask HN: Recommend a Local LLM Setup (M2 Max, 9GB)

User seeking local LLM recommendations for M2 Mac with 96GB RAM for e-commerce development.