Products / Ideas
Things people are building with LLMs
Persistent memory for AI assistants via MCP
Adaptive Recall provides persistent memory for AI assistants using the Model Context Protocol (MCP). It solves the problem of AI assistants forgetting context between sessions, enabling long-term personalized interactions.
AI generates playable arcade games as a subjective evaluation benchmark
AI Arcade is a platform where AI models generate playable arcade games, serving as a subjective evaluation benchmark. It tests creative and functional capabilities of LLMs by having them produce complete game experiences. The target audience is AI researchers and developers seeking qualitative assessments of model performance.
Ephemeral REST chatrooms for AI agents to communicate across stacks
Roomcomm is a service that creates temporary REST chatrooms where AI agents from different stacks, networks, or orchestrators can communicate without glue code or API key sharing. Agents join via a URL, and humans can monitor conversations. It solves the problem of inter-agent communication across heterogeneous environments.
Automated client follow-up and testimonial collection system for freelancers
A system that helps freelancers automate follow-ups with clients after project completion and collect testimonials for each client and project. It uses LLMs to generate personalized follow-up messages and manage the testimonial gathering process, solving the common problem of maintaining client relationships and building social proof.
AI-powered tool that auto-generates and syncs code documentation from code changes
A tool that automatically generates and updates documentation by analyzing code changes, ensuring docs always match the actual codebase. It uses LLMs to detect discrepancies and rewrite documentation, solving the problem of outdated docs for developers.
Lazysusan: run commands on machines from anywhere with curl
Lazysusan is a tool that lets you execute commands on remote machines using only curl, without needing SSH. It is designed for LLM-driven infrastructure tasks, enabling AI agents to debug nodes, check logs, or set up services by simply curling a URL. The target user is developers and ops engineers who want a lightweight, outbound-only remote execution method.
Keel: open-source AI CTO for non-technical product builders
Keel is an open-source AI agent that acts as a virtual CTO for product builders and 'vibe coders' with limited technical experience. It uses LLMs to guide users through the product development process, helping them avoid common pitfalls like over-optimistic coding agents and lack of technical oversight. The tool aims to bridge the gap between non-technical founders and software development.
Brevio: AI digest from blogs, news, research pages
Brevio is an app that lets users subscribe to article-based sources (blogs, news sites, research pages) and receive a single daily AI-generated digest of new articles. It also includes a share extension to summarize webpages, PDFs, videos, or images with text. The app helps people stay informed without opening dozens of tabs.
Agentic society simulation in a fantasy world using LLMs
Artificiety is a living fantasy world where AI agents with distinct personalities, needs, and goals interact and evolve. It uses LLMs to drive agent behavior, enabling emergent social dynamics and storytelling. The project explores agentic societies and narrative generation for worldbuilding enthusiasts and AI researchers.
AI debate platform with RAG, moderator, and judge
A platform where two AI agents debate opposing positions using real data via RAG. A separate moderator agent tries to derail the leading debater, and a judge agent fact-checks and scores the debate. Designed for exploring AI argumentation and fact-checking dynamics.
Moduvox: AI tool that turns PowerPoint decks into narrated presentations with voice cloning and per-slide audio
Moduvox is a tool that converts PowerPoint slides into narrated presentations using AI-generated scripts and voice cloning. It allows users to edit slides and voice independently, making it easy to update training content without re-recording. The tool targets companies that need to produce and maintain training materials efficiently.
View and remove hidden metadata from images, videos, and audio files
Verinio is a tool that lets users inspect and strip hidden metadata (e.g., GPS location, camera info, AI generation details) from media files. It can be used manually or via an API for AI agents to clean files before sharing. Solves privacy and data leakage concerns for anyone sharing media online.
FarmGPT AI: Multi-agent AI farming assistant for crop management
FarmGPT AI is a multi-agent AI farming assistant that helps farmers make better decisions throughout the crop lifecycle. It uses LLMs for a farming chatbot, AI farm planner, and market intelligence, plus computer vision for crop disease detection from leaf images. The tool was built as a Kaggle project and is now launched.
Isitsecure: 1-command SAST, DAST, and LLM security scanner for web apps
Isitsecure is an open-source tool that performs Static Application Security Testing (SAST), Dynamic Application Security Testing (DAST), and LLM-specific security scanning with a single command. It helps developers, especially those using AI-generated code, identify and exploit security vulnerabilities automatically. The tool addresses the problem of insecure AI-generated code by combining SAST findings with DAST exploitation attempts.
Logbook: git history miner that writes CLAUDE.md to prevent LLM rework
Logbook mines local git history to produce a LOGBOOK.md file documenting reverted approaches, skipped tests, silenced warnings, and hotspots. It wires this into CLAUDE.md so that Claude Code reads it in future sessions, reducing the chance of suggesting already-tried or already-finished work. For developers using Claude Code, it solves the problem of LLMs repeating past mistakes or redoing reverted changes.
AI interview prep agent that logs questions as a personal career asset
An AI agent that helps job seekers prepare for interviews by predicting questions based on their CV and target job description. After each interview, users log actual questions and lessons learned, which the system stores to build a personalized, compounding knowledge base over a 30-year career.
Open-source AI infrastructure stack: gateway, guardrails, policies, observability, audit in one Docker Compose
An open-source application layer that integrates LiteLLM, guardrails, PII masking, policies, evals, audit, lineage, and vector search into a single Docker Compose setup. It provides governed, compliant, and auditable AI for organizations, solving the problem of wiring together disparate open-source components.
Slot machine that gives random free tools, updated weekly
A single HTML file that acts like a slot machine, pulling a random free tool from a curated list of 157 tools, toys, and AI oddities. No signup, tracking, or ads. Solves the problem of forgetting bookmarked free tools for anyone who discovers and uses online utilities.
AI perfume recommendations based on personality quiz
ScentIA is an AI-powered perfume recommendation tool that uses a conversational quiz to understand a user's personality, style, and desired vibe, then suggests fragrances. It replaces confusing 'notes' descriptions with personalized matches, solving the problem of buying perfumes online that don't suit the user. Built by an individual for anyone frustrated with traditional perfume descriptions.
Wallie: ask financial questions in plain English, get answers from linked accounts
Wallie is a personal finance app that lets users ask natural-language questions like 'when can I retire?' or 'how much did I spend on takeout last month?' and get answers based on their real transaction and balance data. It uses Plaid (read-only) to link accounts and an LLM to interpret questions and generate plain-English responses. The app includes a FIRE mode that projects retirement dates from actual savings rates.
PrismClip: Search for moments in long videos
PrismClip is a tool that lets users search for specific moments in long videos, giving more editorial control than automated clipping tools like OpusClip. It uses LLMs to understand natural language queries and find relevant scenes. Built for creators who want to extract specific clips without dealing with complex editing software.
Mpsify: runtime patcher to run CUDA scripts on Apple MPS
Mpsify is a Python package that patches PyTorch at import time to transparently redirect CUDA calls to Apple's Metal Performance Shaders (MPS). It solves the problem of editing CUDA scripts to run on Apple Silicon Macs, automatically remapping .cuda(), device='cuda', and map_location='cuda' to MPS equivalents. The tool is for ML practitioners who want to run training scripts or HuggingFace repos on M2 Macs without manual code changes.
Free backtesting tool that detects luck-based trading strategies
Verax is a free backtesting tool that stress-tests trading strategies across multiple market regimes, performs parameter sensitivity analysis, and provides honest verdicts on whether a strategy outperforms buy-and-hold. It includes a bar-replay game where users paper trade historical data and compare results to holding. The tool uses LLMs to generate natural-language explanations of strategy performance and market pulse signals.
AI agent society simulation in a fantasy world
Artificiety is a persistent fantasy world inhabited solely by AI agents powered by LLMs. Each agent observes the world, makes decisions, and writes to its own memory, leading to emergent behaviors like trading, alliances, and rivalries. It explores whether an agentic society can self-organize without human players.
Kote: Capture and reuse engineering context from AI chats and Git
Kote automatically captures engineering context from AI assistant chats and Git activity, storing it for later retrieval during pull requests or quick notes. It solves the problem of losing valuable debugging or architectural context by eliminating the need for manual documentation. Targeted at developers who use AI coding assistants.
Public ledger tracking software ideas from conception to shipped code
Vibehub is a public backlog that records the full lineage of a software project, from the original idea to shipped code. It uses a human-agent collaboration pipeline to track who had the idea, who specced it, and who built it, with live build statuses. The tool aims to prevent good micro-ideas from dying in notes apps and provide transparency for open-source projects.
AI tutor for kids that adapts to each child and gamifies learning
Aivie Spark is a free AI tutor for kids that remembers each child's progress and makes learning feel like a game. It uses LLMs to provide adaptive hints and challenges, helping children work through problems without giving away answers. Built by a dad who wanted his kids to gain something from screen time.
USMLE Pomodoro: study timer for medical students
A desktop app that combines a Pomodoro timer with study materials for USMLE preparation. Built with AI-assisted coding, it helps medical students manage study sessions and breaks effectively.
Chessmate: iOS chess opening trainer that quizzes forgotten moves
Chessmate is an iOS app that helps chess players practice openings by presenting positions one at a time and quizzing them on the correct moves, especially those they tend to forget. It uses LLMs to generate explanations for why a move works, addressing the common problem of forgetting both the move and its rationale after studying.
GitHub commit watcher that drafts social media posts for indie makers
PublishLoud monitors GitHub commits and auto-generates draft posts for LinkedIn and X (Twitter) to help solo developers maintain a consistent marketing presence. The user reviews and edits every post before publishing, ensuring authenticity while reducing the friction of regular posting.
Web UI for Claude Code with rendered markdown and LaTeX
A web-based user interface for Claude Code that renders markdown and LaTeX formulas properly, solving the problem of poor terminal output for non-code conversations. Built entirely by Claude Code itself, it provides a prettier chat experience for users who access Claude via API without a subscription.
No-signup workout generator with shareable cards
Setlist is a static web app that generates instant workout routines based on time, focus area, and available equipment. No account or ads required. It uses a manually tagged exercise list (not LLMs) to produce random but sensible workouts, and offers a save-to-image feature for sharing.
YouTube title and thumbnail declickbaiter using transcript context
Face Value is a browser extension that rewrites YouTube video titles and thumbnails to remove clickbait. It uses transcript context to generate accurate, descriptive titles and neutral thumbnails, making the feed easier to browse.
Knowledge graph search with electrical circuit scoring
OpenKGO replaces binary yes/no matching in knowledge graph search with a scoring system modeled as an electrical circuit. Each query becomes a circuit where relevance is measured by current flow, enabling soft ranking instead of hard pattern matching. It runs locally, is open source under Apache-2.0, and aims to improve retrieval for RAG and search applications.
Wedge helps founders validate startup ideas by pressure-testing assumptions
Wedge is a tool that guides founders through a structured process to evaluate startup ideas before building. It uses LLMs to help users identify target users, surface hidden assumptions, assess risks, and determine what to validate first. The goal is to reduce wasted time and resources on unviable concepts.
Local desktop orchestrator that plans and delegates coding tasks to cloud AI
Forge is a desktop app that uses a local LLM to plan and decompose a high-level coding intent into small, specific tasks, then delegates each task to a cloud AI (Google Antigravity) for execution. It solves the problem of autonomous agents over-engineering or breaking existing code by keeping the AI on a tight leash, giving developers more control and reliability.
AI agent for Portuguese immigration law with document grounding
Leiclara is an AI agent that answers questions about Portuguese immigration law by grounding its responses in official legal documents from Diario da Republica. It solves the problem of navigating broken immigration processes for people moving to Portugal, providing more rigorous and referenced answers than generalist LLMs.
Block dangerous Git and shell commands from being executed by agents
A tool that prevents LLM-powered agents from executing dangerous Git and shell commands. It solves the problem of agents accidentally or maliciously running destructive commands like force pushes or rm -rf, protecting developers and CI/CD pipelines.
Free AI travel destination search across Asia
A free AI-powered search tool that helps travelers discover destinations across Asia. It uses LLMs to understand natural language queries and provide personalized recommendations, solving the problem of finding suitable travel spots without extensive research.
Broll: MCP server for coding agents to create and publish marketing content
Broll is an MCP server that gives coding agents (Claude Code, Codex) a content studio to generate images, videos, carousels, and publish to Bluesky/Mastodon. It uses LLMs to draft content and automate marketing workflows, solving the problem of indie developers needing to create and publish marketing material without manual effort.
Find local businesses without a website and build one in 60 seconds
Parola is a tool that lets freelancers find local businesses with no website (verified phone numbers included) and generate a complete site with booking form in about a minute. It uses an LLM-powered builder to create the site, solving the problem of finding leads and building fast enough to make the economics work.
SayItDev: Add LLM and speech to any app with zero dependencies
SayItDev is a JavaScript library that adds LLM and speech capabilities to any web app without external dependencies or model downloads. It solves the problem of integrating AI features like text generation and voice interaction for developers who want a lightweight, self-contained solution.
Local memory system for Claude Code with review queue
Global Agent Memory is a local memory system for Claude Code that stores project knowledge across sessions. It uses MCP to let Claude search existing memories or propose new ones, which go into a review queue for the user to approve, edit, or reject. It solves the problem of losing context between coding sessions and gives users control over what the AI remembers.
Nulls: fast, AI-powered fitness tracker with photo-based food logging
Nulls is a fitness tracking app that uses LLMs for image recognition to log meals from photos, eliminating manual entry and barcode scanning. It prioritizes speed and simplicity for users who find existing apps bloated and expensive.
Open-source private desktop AI overlay with MCP support
Wisp is a desktop overlay that lets you invoke AI from any app via keyboard shortcuts, with context like text, screenshots, or active window. It supports voice queries, dictation, and pasting results directly, keeping data local and private.
Fitness app with AI coach that calls you out for skipping workouts
TrackFull is a fitness app that pairs users with an AI coach that learns from their goals, training history, and progress. The coach recommends workouts, adapts guidance, and notices when workouts are skipped, offering a more personal and engaging fitness experience.
Eli Felse: a framework for safer autonomous AI assistants
Eli Felse is a framework designed to explore safer ways to create autonomous AI assistants. It uses LLMs to power an AI agent that can play games, chat, browse the web, and perform creative tasks, all while demonstrating safety measures. The framework is aimed at researchers and developers interested in building safer autonomous agents.
NeatContext: lightweight desktop app to give LLMs domain knowledge for oncall incident handling
NeatContext is a desktop application that lets LLMs access domain knowledge to handle oncall incidents more accurately. It solves the problem of SRE agents lacking domain-specific context, enabling better incident response without heavy infrastructure.
All-in-one workspace to tailor resumes to job descriptions
ApplyBoost is a web app that consolidates the resume tailoring workflow into a single workspace. It uses LLMs to rewrite bullet points, identify missing keywords, and scan for ATS compatibility, helping job seekers save time and improve their applications.
AI code review tool using custom standards file
Surmado's Scout is a low-cost AI code review tool for small teams, solo founders, and vibe coders. It reviews pull requests against a user-defined STANDARDS.MD file, providing feedback on what looks good, what needs work, and a reviewer brief. It aims to be 1/10 the price of Claude.
