HN axiorank 17d ago

An AI gateway that signs a receipt for every LLM response

Security gateway for AI agents with routing, failover, guardrails, and cryptographic receipts for LLM responses.

HN oryx1729 17d ago

Show HN: Sanbox, batteries included sandboxes for AI agents

Sanbox: sandbox platform for running AI agents with isolation, persistence, templates, CLI integration, and self-hosting options.

HN clementrx 17d ago

I built an AI strength coach because I wanted my training backed by real studies

AI fitness coach application leveraging academic research for personalized training recommendations.

HN fahimulhaq 17d ago

Show HN: Fenzo – Build an interactive course on any topic

Fenzo: AI application that generates personalized interactive courses from uploaded notes or documents.

HN ds246 17d ago

Show HN: Spendict – a performance marketer's verdict for AI agents, over MCP

Spendict: performance tool for evaluating ad campaigns using AI agents over MCP protocol.

HN JosephSenior 17d ago

Show HN: Grinta – a local-first coding agent built for long autonomous runs

Local-first coding agent with LSP/DAP support. Provider-agnostic, long autonomous runs, no cloud lock-in. v1.0.0-rc1.

HN tha_infra_guy 17d ago

Show HN: Runeward: Sandboxing AI agents with policy gates

Runeward: sandboxing framework for AI agents with declarative profiles, deny-by-default egress, audit ledgers, policy gates, and cost guardrails via REST, MCP, CLI, and web dashboard.

HN thisbrt 17d ago

Show HN: DejaView – Terminal dashboard for all your Claude Code sessions

Open-source TUI dashboard for managing Claude Code sessions. Groups transcripts by project, enables quick resumption.

HN steamspaghetti 17d ago

Show HN: Godot running Gemma 4 inference in GDScript and Vulkan compute shaders

Godot 4.7 project running Gemma 4 LLM inference via GDScript and Vulkan compute shaders without external dependencies. Experimental, ~10x slower than llama.cpp.

HN verdverm 17d ago

Finetuning a Reasoning LLM with Supervised or Reinforcement Learning?

Discussion on fine-tuning small LLMs with reasoning traces and tool-calling using supervised or reinforcement learning approaches.

HN ryzvonusef 17d ago

Political Neutrality Benchmark of popular AI models

Benchmark measuring political neutrality across 18 AI models from multiple labs on six political dimensions with refusal rates.

HN Oxlamarr 17d ago

Show HN: Zero Trust Boundary for Agents

Security framework (Attestor) for AI agents enforcing trust boundaries between agent operations and system execution with access controls.

HN hannan2 17d ago

Show HN: Inkfold – workspace across multiple AI providers with shared memory

Multi-provider AI workspace maintaining shared conversation history and context across ChatGPT, Claude, Gemini and other LLMs.

HN mknight2690 17d ago

Agent Service – promptable AI agents with guardrails and downloadable packages

Tool for creating promptable AI agents with guardrails and downloadable packages. Limited detail provided.

HN vinhnx 17d ago

AI Should Build Its Own Research World Model

Approach for AI agents to build internal world models during exploration, tested on ARC-AGI puzzles achieving higher solve rates than baseline LLMs.

HN beardyw 17d ago

To Build More Believable Bots, Simulate the Neurochemistry

Robot emotion modeling via simulated neurochemistry. Uses LLM input to drive chemical state variables that project to emotional expressions.

HN mrauha 17d ago

LLMlet: P2P distributed LLM inference on browsers

P2P distributed LLM inference framework running llama.cpp in browser WebAssembly with peer computation.

HN moeidsaleem 17d ago

MSK – an AI agent that thinks like a CTO

iPhone app positioning AI agent as virtual CTO for architecture and strategy advice. Limited technical details.

HN cosmtrek 17d ago

Show HN: Mindwalk – Replay coding-agent sessions on a 3D map of your codebase

3D visualization tool replaying coding agent sessions on repository map, showing agent exploration and edit patterns during task execution.

HN eestrada 17d ago

Show HN: AsyncFutures – A Ruby gem for asynchronous/concurrent code execution

Ruby gem providing unified async API across Ractors, Threads, and Fibers for testing concurrent code performance.

HN piyushkhemka 17d ago

Built a tracker to estimate water wastage when talking to Claude

Claude Code status line tool tracking estimated water consumption from LLM inference, showing real-time environmental cost of API calls.

HN crisnahine 17d ago

Show HN: Chameleon – shows Claude a real file from your repo before it edits

Developer tool that shows Claude the actual codebase conventions before editing. Injects file archetypes via hooks to improve code generation quality.

LB thebeach.dev via azhenley 17d ago

An Agent in 100 Lines of Lisp

Personal retrospective on AI coursework from 2000 using Lisp, comparing historical AI education to modern deep learning approaches.

HN matt_d 17d ago

Reducing HBM Bottlenecks in JAX-Based LLM Training with Host Offloading

Addresses GPU memory bottlenecks in JAX-based LLM training using host offloading to optimize HBM utilization for large model scaling.

HN ninjahawk1 17d ago

Show HN: Confessor – replay what private info Claude Code accessed on your PC

Privacy monitoring tool showing exactly what files Claude Code agent accessed and commands executed on user's computer.

HN jhoho 17d ago

What xAI's Grok Build CLI Actually Sends to xAI

Technical teardown of xAI's Grok Build CLI analyzing what data is sent to xAI servers, with captured network artifacts and reproducible findings.

HN E-Reverance 17d ago

Learning-to-Optimize via Deep Unfolded Flows

FlowOptimizer: deep unfolded flow-based framework using velocity fields for learned iterative optimization.

HN shbhmrzd 17d ago

What happens between entering the prompt and seeing the first word appear

Technical deep-dive into LLM inference: how KV cache fills and tokens stream from prompt to first output word.

HN ykev 17d ago

Agentifying your software development lifecycle

Using AI agents throughout development lifecycle: code generation, testing, review, and maintenance rather than generation alone.

HN modinfo 17d ago

LM-Polygraph: Uncertainty Estimation for LLMs

LM-Polygraph: open-source uncertainty estimation toolkit for detecting LLM hallucinations and improving reliability in text generation.

LB mimo.xiaomi.com via sanxiyn 17d ago

Full-Pipeline Inference Optimization for MiMo-V2.5 Series

MiMo-V2.5 model family optimization combining Hybrid Sliding Window Attention, sparse MoE, and multimodal encoders for efficient long-context inference.

HN marzukia 17d ago

Fixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio

Bug fixes enabling Qwen3.5-122B to run efficiently on Mac Studio, resolving cache leak causing multi-minute token latency.

HN tomatoes2026 17d ago

Show HN: AgentTransfer – open-source file transfer for AI agents (one Go binary)

AgentTransfer: open-source file transfer system for AI agents with identity, folders, and inter-agent discovery (one Go binary).

HN tionis 17d ago

Mesh LLM: distributed AI computing on iroh

Mesh LLM: distributed inference framework using iroh for decentralized, privacy-preserving LLM deployment without vendor lock-in.

HN petilon 17d ago

AI and Job Postings: From Destruction to Creation?

Analysis of software job posting growth correlating with agentic AI tools like Claude Code becoming widely available.

HN Abhijeet_Buag 17d ago

ContextOps, an ESLint-like static analyzer for LLM context

ContextOps: deterministic static analyzer for LLM context that detects redundancy, token waste, and imbalance without model calls.

HN wzulfikar 17d ago

Show HN: Token Time – Screen Time, but for your AI agent tokens

Token Time: monitoring tool for tracking AI agent token usage analogous to screen time monitoring.

HN thomaslwang 17d ago

Show HN: OpenThomas – weather trader agent for prediction markets

OpenThomas: open-source AI agent for weather forecasting and trading on prediction markets.

HN gnabgib 17d ago

Secret Claude tracker surprises users after Anthropic's anti-surveillance stance

Anthropic removing stealth steganography codes from Claude Code designed to catch model theft by competitors.

HN alama24 17d ago

Show HN: BoundFlow – an open-source control plane for AI agents

BoundFlow: open-source control plane for LLM agents with cost caps, approval gates, and self-healing policies.

HN phpencil 17d ago

LLMs are adapting their environments to themselves

Analysis of how LLMs adapt their operational environments rather than undergo lifelong learning like humans.

HN RenderBear 17d ago

Harmonic Contour Integration: Compact, distributed edge detection algorithm

Harmonic Contour Integration: distributed edge detection algorithm for RGB images using learned harmonic metrics.

HN shanrizvi 17d ago

Bitemporal provenance in agent memory: What did we believe, when, and why

MnesticDB fork of CozoDB adds bitemporal provenance tracking for AI agent memory, enabling auditable change history and belief state management.

HN fenilsuchak 17d ago

Show HN: OpenBenchmarks – Helping agents discover and pick the right SaaS APIs

OpenBenchmarks: open-source reproducible benchmarks for SaaS APIs designed to help AI agents evaluate and select tools.

HN mrhustlex 17d ago

I built TradingSpy: local, privacy-first AI trading assistant(First Open Source)

TradingSpy: open-source local-first AI trading assistant with market analysis, strategy generation, and backtesting via Docker.

HN brainbootdev 17d ago

The AI Disagreement Index: 8 models agreed on the "best tool" 0 of 16 times

AI Disagreement Index tracks cross-model disagreement on tool recommendations across 8 models, measuring consistency rather than ranking.

HN rane 17d ago

Show HN: Use GPT-5.6 and other models in Claude Code

Tool enabling Claude Code IDE to use multiple LLM providers including GPT, Kimi, and Grok.

HN tejusarora 17d ago

Show HN: Sqlsure – deterministic semantic checks for AI-generated SQL

Sqlsure provides deterministic semantic validation for AI-generated SQL queries, catching logic errors in joins and aggregations before execution.

HN zapeterson16 17d ago

AI Arcade – which coding model can build the best arcade game

AI Arcade benchmarks coding models building arcade games with identical prompts and frameworks, comparing output quality and functionality.

HN antoinefornas 17d ago

I built a free tool to evaluate AI agent outputs (human labels and LLM judges)

Open source browser-based tool for evaluating AI agent outputs using human labels and LLM judges. Local, no backend required.