24d ago
Show HN: Fence – Jiminy Cricket for AI coding agents
Open source tool that prevents AI coding agents from executing dangerous commands by understanding intent rather than using denylists.
24d ago
I found 3 self-contradictions in the Agentic Commerce Protocol spec
Validation tool for Agentic Commerce Protocol spec compliance, catches violations before OpenAI conformance certification.
24d ago
Why AI Agents Forget by Design
Analysis of why AI agents lack persistent memory by design, examining stateless architecture chosen by LLM providers and its implications.
24d ago
Conversing with antiquity: Agentic AI partner for expanding historical research
Research on agentic AI system for historical inscriptions analysis, reconstructing damaged ancient texts with uncertainty estimation.
24d ago
Kennel 1.0.0 – Native Desktop App for Managing AI CLI Agents
Native desktop app for running and managing multiple AI CLI agents simultaneously, preserving session state across restarts.
24d ago
Tools vs. Subagents: Building Effective AI Agents Without Over-Engineering
Technical article on architecture patterns for AI agents, deciding when to implement functionality as tools versus subagents to avoid over-engineering.
24d ago
Show HN: Watch how an AI agent thinks
Educational visualization tool with open source code showing AI agent reasoning and decision-making processes step-by-step.
24d ago
Show HN: I wrote a 1-bit WebGPU runtime to run a 1.7B LLM in the browser
WebGPU runtime implementation enabling 1.7B parameter LLM inference in browser using 1-bit quantization.
24d ago
Show HN: Humanize – Best static AI text humanizer, based on 50 research papers
Open source tool for humanizing AI-generated text based on 50 research papers, analyzing sentence rhythm, punctuation, repetition patterns.
24d ago
I have an interesting idea to reduce the output tokens almost by fifty percent
Technique to reduce LLM output tokens by ~50% in agent sessions by eliminating redundant code block echoing in file edits.
24d ago
Beating a text-to-SQL benchmark: can you get better than plain Claude?
SLayer, an open-source semantic layer for AI agents, evaluated on text-to-SQL benchmarks to improve product quality across diverse use cases.
24d ago
Show HN: Baerly-storage, a document DB that runs per request, no DB server
Serverless document database running entirely in S3/R2 buckets with HTTP request execution layer, designed for LLM API compatibility.
24d ago
Show HN: Tracking GenAI cost and endpoint fragility so app teams don't have to
Monitoring and optimization tool for GenAI costs and endpoint performance across multiple LLM providers, with pricing comparison and usage tracking.
24d ago
Stop Waiting for a Bigger Context Window
Analysis of multi-agent orchestration as alternative to larger context windows for handling long sequences, showcasing SwarmOS platform for specialized agents.
24d ago
Reducing Doom Loops with Final Token Preference Optimization
Technical approach using final token preference optimization to reduce inference-time doom loops where models repeat spans, improving over repetition penalties.
24d ago
Codex makes fewer bugs, but more people use Claude
Cubic code review analysis comparing bug rates between Codex and Claude models across thousands of commits.
24d ago
Show HN: Wayflow – an embeddable AI workflow builder (open source)
Wayflow, open-source embeddable AI workflow builder with node-based editor and integrated runtime for agentic products.
24d ago
Show HN: A notepad that turns meeting-full days into shareable memory
Pitara meeting notes tool that organizes discussions into shareable memory for integration with AI tools.
24d ago
Weighing smoke: why AI visibility dashboards are mostly useless
Critique of AI visibility dashboards in search optimization, questioning their cost-effectiveness.
24d ago
Show HN: Prizmi – a proactive AI assistant for everyday life
Prizmi proactive AI assistant for cross-device tasks with privacy-first design, controlled via messaging apps.
24d ago
Ask HN: Are LLMs slowly making companies dysfunctional?
Discussion on how over-reliance on LLMs for quick solutions may cause complexity creep and cognitive debt in organizations.
24d ago
Show HN: I built a free website that makes LLM prompting easier in 40 languages
Free website tool that simplifies LLM prompting across 40 languages.
24d ago
Show HN: Rowboat – Open-source, local-first alternative to Claude Desktop
Rowboat, open-source local-first alternative to Claude Desktop with extensible work surfaces for custom applications.
24d ago
Letting Untrusted AI Agents Use Credentials They Can Never Read
Declaw credential vault design for safely giving untrusted AI agents access to secrets without exposing them, even under prompt injection.
24d ago
GPU-accelerated context memory for on-device AI agents on Apple Silicon
GPU-accelerated context memory implementation for on-device AI agents on Apple Silicon processors.
24d ago
Show HN: Atrophy – measure whether AI is eroding your unaided coding skill
Atrophy command-line tool measures coding skill degradation by regular unassisted exercises with automatic grading and skill ratings.
24d ago
Show HN: SOCBench – an open benchmark for AI on SoC tasks
SOCBench: open benchmark measuring AI performance on security operations tasks, addresses 36% false positive problem with LLMs.
24d ago
Show HN: TraceGen – realistic OpenTelemetry traces, incl. AI-agent, one binary
Developer tool for generating realistic OpenTelemetry traces including AI agent scenarios. Open source single binary.
24d ago
Show HN: Context Warp Drive – Deterministic context folding for AI agents
Tool implementing deterministic context folding technique for AI agents. Lacks detail but addresses agent optimization.
24d ago
Automating AI Away
Developer shares experience using Claude for code review on Beagle SCM, notes LLM limitations with precision and context.
24d ago
PrivAiTe, a local LLM proxy that redacts PII in tool-call arguments
Local proxy that redacts PII in LLM API calls and tool arguments before data leaves machine. Drop-in middleware with zero telemetry.
24d ago
Optical Scale Up Fabrics Are Limited by Manufacturing, Not Architecture
Technical specification for optical interconnect networks for AI infrastructure co-designed by major tech companies.
24d ago
Theory of constraints, AI, and code review
Theory of constraints applied to AI-assisted code review. Title only, insufficient content.
24d ago
Show HN: Bloomiro, an MCP tool for improving SEO and AI search visibility
Bloomiro: MCP tool enabling AI agents to analyze SEO and search visibility issues. Practical agent application.
24d ago
Mozilla's Otari: The Open Source LLM Control Plane
Mozilla's Otari: open-source control plane for managing LLM deployments and inference.
24d ago
Build an AI Audiobook Narrator with Telnyx AI Inference, TTS, and Cloud Storage
Tutorial on building an AI audiobook narrator using Telnyx's AI inference and text-to-speech services.
24d ago
Open source OCR with vision language models: High throughput and low cost
Open source OCR pipeline using vision language models for document parsing. Processes 20 pages/sec on single GPU at $0.04 per 1000 pages.
24d ago
Show HN: New Search MCP Using Cloudflare AI Search
Open source AGPL MCP tool using Cloudflare AI Search to index repos, enable agents to search documentation and answer project questions.
24d ago
My AI Agents Were Working, but I Had No Visibility
Multi-agent system for Applied: researchers gather AI adoption data, translator makes Spanish content, matcher connects users. System expanded with model research agents.
24d ago
Show HN: Rewire Text – system-wide text transforms with BYOK and local-models
Rewire Text: cross-platform tool for system-wide text transforms using local models or APIs (OpenAI, Anthropic, Gemini, etc).
24d ago
Show HN: The Banana Test: coding agents grow a 3D banana plant'
Benchmark comparing 5 AI coding agents (Claude, Codex) on 3D plant simulation task using three.js and biology simulation.
24d ago
Show HN: Building My Own Public HTML Library for Async Learning
html-hub.com: public library for organizing HTML artifacts from AI agents, supports transcripts and citations.
24d ago
The AI UGC ad math: $2 a video, and the bill nobody prices in
Analysis of AI-generated UGC ad economics: $1-11 per video vs $185 human-made, excludes conversion and platform labeling costs.
24d ago
Show HN: Bastion – self-hosted VMs for background coding agents
Bastion: open-source orchestrator for running background coding agents in self-hosted VMs with reproducible environments.
24d ago
Show HN: DocFlow – AI documentation updates for GitHub pull requests
DocFlow: GitHub App that auto-generates README, API docs, and changelog drafts when PRs merge.
24d ago
CorvinOS – an agentic OS where GDPR/EU AI Act compliance can't be switched off
CorvinOS: self-hosted agentic OS with GDPR/EU AI Act compliance built-in, connects multiple LLMs to messaging platforms.
24d ago
Claude Code Antigravity CLI Agents Skill
Claude Code skill enabling orchestration of Google Antigravity CLI agents for autonomous subtask delegation during development.
24d ago
A visual calculator to estimate and compare LLM API pricing
Visual calculator for comparing LLM API pricing across providers like OpenAI, Claude, Gemini.
24d ago
Show HN: SigMap – deterministic repo maps for AI coding agents
SigMap generates deterministic repository maps designed for AI coding agents to understand codebases.
24d ago
Agents don't need smarter models, they need a map
Research on computer-use agents showing need for spatial understanding maps rather than smarter models to interpret UI significance.