15d ago
Show HN: I implemented a neural network in SQL
Implementation of neural network in SQL using Xarray-SQL array database library. Demonstrates mapping n-dimensional arrays to 2D tabular representation.
15d ago
Show HN: PlanWright – A control plane for AI coding agents
PlanWright is an MCP-driven control plane for AI coding agents, enabling planning, implementation, and triage with full documentation.
15d ago
Claudio? Why we need and open source Claude
Mindshub is an open-source AI coworker alternative to Claude, emphasizing inspectable, modifiable, and forkable AI systems.
15d ago
Building intuition about LLM parameter counts
Educational content about understanding LLM parameter counts and model scaling. Machine learning research.
15d ago
GPT-5.6 Luna Showed a Better ROI on Cybersecurity Benchmark
OpenAI released GPT-5.6 models that cleared high cybersecurity capability threshold, automating vulnerability discovery and attack scenarios.
15d ago
J-Space Oddity: Do VLMs Dream of Text Tokens?
Anthropic interpretability research exploring whether vision-language models think in text tokens, testing Jacobian-based methods on VLMs like Qwen2.5-VL.
15d ago
What we measured after moving 100B tokens/week to open-weight models
Production case study migrating 100B tokens/week from GPT-5.3 to open-weight MiniMax M3, achieving 55% cost reduction for QA agents.
15d ago
Ask HN: What makes someone good at using Claude Code?
Promptster: AI fluency platform measuring engineering teams' proficiency with Claude Code and other AI coding tools.
15d ago
Attention is all they want – blocking content in websites with local LLMs
Using local LLMs to block unwanted website content. Browser-based LLM application. Limited details.
15d ago
Amid AI IP Theft Concerns, Microsoft CEO Floats New AI Patent Concept
Microsoft CEO proposes new patent framework to address IP theft concerns from AI models trained on corporate data.
15d ago
What would it take for AI to discover penicillin?
Discussion of agentic science and autonomous labs discovering new compounds, exploring AI's role in scientific discovery.
15d ago
Show HN: kassette – Durable agent workflows backed by object storage
Durable execution framework for AI agents using object storage and job queues. Solves agent failure recovery.
15d ago
AI agents write Ruby but can't navigate it: a 5-model, 13-codebase benchmark
Benchmark comparing 5 LLM models across 13 Ruby codebases, measuring code generation vs. navigation capabilities.
15d ago
Show HN: BillAI Bass, an AI-Powered Big Mouth Billy Bass Using Strands Agents
Demo of AI-powered Big Mouth Billy Bass toy using Strands agent framework for autonomous control.
15d ago
AgentMint: Research on how AI shopping agents pick products
Research on decision-making processes in AI shopping agents, analyzing product selection behavior across models.
15d ago
Formally verifying AI-generated GPU kernels
Gimlet uses formal verification to validate AI-generated GPU kernels, improving trust in agent-optimized code.
15d ago
Show HN: Finterm.ai Bloomberg terminal for Claude Code
CLI tool providing Claude/GPT coding agents direct access to financial data: stock prices, options, SEC filings via structured API.
15d ago
How to build AI agents business teams want
Guide on designing and deploying AI agents for enterprise business team workflows and adoption.
15d ago
Show HN: ContextVault – the shared memory layer for your AI and your team
Shared memory/context management system for AI tools and teams. Infrastructure for LLM applications.
15d ago
Show HN: The Safe, Production-Ready Backend for AI-Generated Apps
Show HN post about a backend for AI-generated apps. Minimal details provided.
15d ago
Show HN: We are giving away > over the next 12 months to hack AI agents
Platform crowdsourcing adversarial attacks and jailbreaks against AI agents, with rewards for vulnerability submissions.
15d ago
I Cut an AI Agent's Token Use by 94%
Technical case study: optimization techniques reducing token consumption in AI agent systems by 94%.
15d ago
Show HN: AutoFlow – Turn screen recordings into user guides
AutoFlow automates user guide generation from screen recordings using chrome extension and ML transcription. LLM-powered product documentation tool.
15d ago
Show HN: Clay Seal Identity – Agents need accountability
Security framework for AI agent accountability using identity verification and access controls for privileged operations.
15d ago
Show HN: QuantumReckon – find your real cloud and AI spend, including tokens
Show HN tool for tracking cloud and AI token spending. Helps monitor LLM usage costs.
15d ago
Show HN: Docs.dev – A docs site you own, where AI drafts and you edit in place
Docs.dev platform for AI-assisted documentation where Claude Code agents draft content and users edit in place, with keyboard.dev integration.
15d ago
I built a prompt injection defense middleware for LLMs (Python/FastAPI)
Python/FastAPI middleware for defending LLMs against prompt injection attacks. Security developer tool.
15d ago
Learning from Experience Instead of Curated Datasets
Machine learning research on learning from experience vs. curated datasets. Training methodology comparison.
15d ago
Show HN: A prompt to make LLMs write property-based tests
Prompt engineering technique to generate property-based tests using LLMs. Developer tool for testing.
15d ago
Show HN: Anonymous LLM proxy. Pay in crypto, no account needed
Cryptocurrency-based anonymous proxy for LLM API access. LLM infrastructure. Minimal details.
15d ago
Show HN: Claude Organizer, folders and search for Claude.ai, all local
Browser extension adding folders, search, bookmarks to Claude.ai conversations. Local-only, no data collection.
15d ago
Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor
Benchmarking of Apple's SpeechAnalyzer against Whisper on 5,559 utterances. SpeechAnalyzer most accurate, 3x faster than Whisper Small. Transparent results published.
15d ago
Show HN: Containerized AI development with cross-compatibility
Docker-based template for AI development with multi-provider CLI support and isolation. Works on Linux, macOS, WSL2.
15d ago
Would you use AI to discover investment opportunities?
Article about using LLMs with real-time market data for stock research. Generic LLM application overview.
15d ago
A/B Test results: Trying different homepage heroes got us 2.6x sign-ups
A/B testing results for forms.app's AI generator hero section achieving 2.6x signup increase. Includes LLM-powered form generation feature.
15d ago
Flow – a compact data pipeline language that cuts LLM token usage by 33%
Domain-specific language for data pipelines that reduces LLM token usage by 33%. LLM optimization.
15d ago
CortexWP AI: WordPress Agentic system for debugging and development
WordPress agentic system for debugging and development that can inspect actual site configuration and errors.
15d ago
The AI Safety Paradox
Discussion of AI safety concerns as LLM-based agents become capable of unsupervised work over extended periods.
15d ago
Show HN: A developer tool guide for AI coding agents (900 tools and counting)
Comprehensive guide to 900+ developer tools for AI coding agents, accessible via CLI and MCP integration.
15d ago
Show HN: Yes-Brainer – A council of LLMs that debate in the browser, BYOK
Show HN: Yes-Brainer tool enabling multiple LLMs to debate in browser with bring-your-own-key. Multi-agent LLM application.
15d ago
Show HN: Harpist – convert any website into a refined and documented API
Harpist tool converts undocumented websites into refined APIs using code analysis and HAR file recording.
15d ago
Sysdig documents the first ransomware attack run end to end by an AI agent
Report on JadePuffer, first ransomware attack executed end-to-end by autonomous AI agent.
15d ago
AI agents 136.5 times less efficient than conventional AI
Title-only submission comparing efficiency of AI agents versus conventional AI.
15d ago
Languages in 3 days: field notes from localizing an AI translation product
Case study on localizing an AI translation product across languages in rapid timeframes. Real-world implementation notes.
15d ago
Show HN: Replaces Supabase and FastAPI
Jerrycan: AI-native Rust backend framework where agents design, generate, verify backends without manual coding. MCP server integration.
15d ago
LingBot-Video: An open-source MoE video foundation model for embodied AI
Open-source mixture-of-experts video foundation model for embodied AI. Machine learning research in multimodal domain.
15d ago
Show HN: Learn OCaml, exercises checked by WASM in browser
Interactive OCaml course with WASM compiler in browser for self-checking exercises. Built with Claude Code. Open source.
15d ago
Evaluating the GPT-5.6 Family
Benchmark evaluation of GPT-5.6 family models (Sol, Terra, Luna) vs Anthropic alternatives on agentic tasks. Cost-performance analysis provided.
15d ago
Show HN: Build Rules for Claude Code, Cursor, and Codex
Kastra intercepts and evaluates AI agent tool calls against policies before execution. Control layer for Claude Code, Cursor, Codex agents.
15d ago
Show HN: Crowdmind – open-source tool to test ideas against AI personas
Open-source tool for testing ideas against multiple AI personas. Limited technical details provided.