14d ago
Show HN: Oodle.ai – $10 per million agent traces
Oodle.ai: agent trace observability platform with columnar storage, stores non-deterministic LLM agent traces at $10 per million.
14d ago
Context Bombs: stopping AI attackers in their tracks
Context bombs: hidden strings in canaries that trigger AI agent safety guardrails to prevent autonomous cyberattacks.
14d ago
Show HN: Hiver – Chrome DevTools for Agents
Hiver: Chrome DevTools for debugging and optimizing AI agents, providing visibility into model behavior, memory, and token usage.
14d ago
How do you go from junior to staff engineer when AI writes the code?
Article exploring career progression for engineers as AI tools automate code writing, questioning traditional mentorship models.
14d ago
Show HN: Cruxible – Terraform-like ontology config to governed state for agents
Terraform-like configuration system for managing agent state and memory with governed ontology structure.
14d ago
Hookami Anywhere – A Read-Only YouTube Research MCP for ChatGPT and Claude
MCP server for YouTube research integration with ChatGPT and Claude.
14d ago
Oya – Keep tool outputs away from the LLM to cut tokens and stop injection
Stub entry with minimal content about tool output isolation.
14d ago
Show HN: Liveshortly:Live,Collab,Pair Prompt with Your AI Agents
Collaboration tool for real-time pair prompting with AI agents. Turn sessions into blog posts. Beta launch seeking testers.
14d ago
Show HN: Aireceipts, itemized cost receipts for AI coding agents
Cost tracking tool for AI coding agents showing real-time billing, model usage, and itemized receipts per session.
14d ago
Show HN: Halley – Turn production LLM traffic into $0 CI regression tests
Tool converting production LLM traffic into CI regression tests at zero cost.
14d ago
Show HN: MCP-Billing – self-hosted auth and usage billing for MCP servers
Self-hosted Next.js boilerplate for MCP servers with OAuth 2.1, API key management, Stripe usage billing, and Redis rate limiting.
14d ago
DSLs Enable Reliable Use of LLMs
Explores using Domain-Specific Languages as abstractions to guide LLMs toward reliable code generation, with Tickloom distributed systems example.
14d ago
High-integrity HTML extraction for AI agents (with native MCP)
Open source HTML parser with CSS selectors for AI agent data extraction. MIT/Apache-2.0, zero dependencies, supports MCP protocol.
14d ago
Show HN: I gave an LLM a clock and it usually won't look at it
MCP time server tool that provides LLMs with accurate clock access, solving issues where AI assistants lose track of time in long conversations.
14d ago
Show HN: Loopers – Atomic pre-call budget checks for LLMs using Redis Lua
Redis Lua-based atomic pre-call budget checking system for LLM requests to prevent token overspend.
14d ago
Show HN: RavenGate – LLM gateway that redacts PII across SSE chunk boundaries
LLM gateway with analytics, PII redaction across SSE boundaries, and per-provider support (OpenAI, Anthropic, Gemini, Groq, Mistral, DeepSeek, xAI, OpenRouter, Azure).
14d ago
Show HN: Run SAM only when your tracker is uncertain – #1 on SportsMOT
Computer vision optimization that runs SAM2 segmentation only when tracker uncertainty is high, achieving #1 on SportsMOT benchmark.
14d ago
Show HN: Kmux – Parallel terminal workspace optimized for AI coding agents
Terminal workspace multiplexer optimized for running multiple AI coding agents in parallel with git integration.
14d ago
Show HN: VisualEther – Wireshark PCAPs to sequence diagrams; MCP for AI analysis
VisualEther converts Wireshark network captures to sequence diagrams with MCP protocol support for AI analysis. Open source developer tool.
14d ago
Show HN: Two Next.js templates with zero dependencies and one config file
Nova is an observability and orchestration dashboard for AI products. Tracks prompts, evaluates quality, and runs controlled experiments on prompt variants.
14d ago
Don't bring an AI detector to a deepfake fight: provenance over detection
Proposes cryptographically-signed media provenance as alternative to AI-generated content detection, arguing detection is unwinnable arms race.
14d ago
Ask HN: Any objective research on which languages are best for AI agents?
Discussion asking for objective research and benchmarks comparing programming languages for AI agent development across frontier LLMs.
14d ago
Show HN: Town – Discord in a pixel town where the NPCs have skills
Town is a Discord-like pixel town interface where users interact with AI agents as NPCs with specialized skills. Built as a shareable tool leveraging Claude for multi-agent reasoning patterns.
14d ago
Guardian Angels: LLM Personalization for Productivity and Security
Framework for personalized LLMs that emulate user values and preferences using imitation learning and Decision Transformers for productivity and security.
14d ago
Mnemo AI – Local agentic assistant for any LLM that learns from its failures
Mnemo AI is a local agentic assistant built on LangGraph supporting multiple LLM providers, MCP integration, and RAG. Learns from failures with conversation management.
14d ago
Switching an LLM's tier changes its "best tool" answer about half the time
Analysis showing that different LLM tiers (Opus vs Haiku, Pro vs Flash) produce inconsistent answers on identical tool recommendation queries.
14d ago
Show HN: I RL-trained an agent that trains models with RL (for –$1.3k)
Open-source project training an RL agent that trains other models via RL. Includes trained weights, code, task families, and cost breakdown. Meta-learning approach to AI training.
14d ago
Lenz – A fact-checking API for AI-generated content
Lenz is a fact-checking API for AI-generated content using multi-model debate and citation verification. Supports integration with ChatGPT, Perplexity, Gemini.
14d ago
The US-China AI arms race has taken an unexpected turn
Analysis of US-China AI competition following DeepSeek R1 and other open-source Chinese LLM releases, market impact, and government responses.
14d ago
Lessons learned by a non-developer learning to deploy apps in production
Lessons learned deploying production AI apps: discusses complexity scaling, context window limitations, code quality degradation when using AI to build features iteratively.
14d ago
Adapting offensive security for the AI agent age
Discussion of offensive security adaptations needed for AI agent systems and autonomous attack capabilities.
14d ago
The database wars and where LLMs are (maybe) headed
Speculative essay comparing LLM adoption trajectory to database industry: potentially becoming essential infrastructure rather than revolutionary technology.
14d ago
Show HN: I built a deterministic check for fabricated quotes in LLM output
Verbatimeter: Open-source tool for detecting fabricated quotes in LLM outputs using deterministic verification of groundedness. Includes decorator and CLI for RAG agents.
14d ago
Codex starts encrypting sub-agent prompts
Codex multi-agent framework regression where sub-agent prompts encrypt after code merge, affecting spawn_agent and message handling.
14d ago
Merged at the Speed of AI
Case study of Bun JavaScript runtime replacing 1M lines of Zig with Rust using Claude AI coding. Documents AI-assisted large-scale code migration.
14d ago
Show HN: BYO AI free notetaking with optional screen reading for OpenClaw/hermes
On-device macOS meeting transcription app using Parakeet, Gemma 4, OpenClaw, and Hermes Agent. Free, source-available, supports local model connections.
14d ago
Enhancing GNU-Pth for m:n threading using Claude and Codex
Using Claude and Codex models to enhance GNU-Pth threading library with m:n threading support. Minimal detail provided.
14d ago
Thoughts on starting new projects with LLM agents
Technical narrative on using LLM agents for Python project restructuring and building a new project (watgo) from scratch with agent assistance.
14d ago
Spitting chips: A dive into the data and token industry, & who carries GPU risk
Analysis of GPU supply/demand dynamics for AI/ML infrastructure. Projects supply-demand convergence around 2028 amid token demand growth.
14d ago
The Unfair Judge: A Mechanistic Interpretability Account of LLM-as-Judge
arXiv research on mechanistic interpretability of LLM evaluators. Analyzes how LLMs function as judges.
14d ago
Mensfeld/code-on-incus: Give each AI agent its own isolated machine
Open source tool providing isolated containers for AI coding agents with full system access, credential protection, and active defense.
14d ago
Show HN: ZenStack – access control at the ORM layer, built for coding agents
ORM-layer access control system designed for AI coding agents with RBAC/ABAC support.
14d ago
Nested-ReFT: Efficient Reinforcement Learning for Large Language Model Fine-Tuning via Off-Policy Rollouts
Efficient RL fine-tuning for LLMs using off-policy rollouts to reduce computational cost in verifiable reward-based reinforcement learning for reasoning tasks.
14d ago
Beyond Slater's Condition in Online CMDPs with Stochastic and Adversarial Constraints
Algorithm for online constrained MDPs achieving improved regret bounds under stochastic and adversarial constraints beyond Slater's condition.
14d ago
Graph Optimization Foundation Model: Tokenizing Graph via A Language-Model Paradigm
Foundation model for graph optimization using language model paradigm to handle OR problems on graph structures while managing combinatorial constraints.
14d ago
Unveiling the Mechanisms of Multi-Hop Reasoning in Transformers via Identity Bridge
Identifies missing bridge entity supervision as cause of multi-hop reasoning failure in transformers; proposes identity bridge supervision improving out-of-distribution composition.
14d ago
Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization
Distribution Matching Policy Optimization: RL algorithm designed for diffusion LLMs enabling reasoning tasks with higher inference throughput than autoregressive models.
14d ago
CANDI: Hybrid Discrete-Continuous Diffusion Models
CANDI: hybrid diffusion model for discrete and continuous data, analyzing token corruption mechanisms through identity and rank degradation.
14d ago
Controllably Efficient Language Models
Framework enabling transformers to trade off inference efficiency and quality dynamically, controlling sparse/linear attention and convolutions per layer.
14d ago
Enabling Agents to Communicate Entirely in Latent Space
Interlat: enables LLM-based agents to communicate directly in latent space, bypassing discrete tokens for richer information exchange in collaborative problem-solving.