Filter YouTube by keyword, channel, topic, or allowlist so you only see what you want
Problem
YouTube's home feed is full of unwanted channels, clickbait titles, and irrelevant topics. Users want to filter by keyword, channel, topic, or even an allowlist, but YouTube offers no such controls. Once the feed is polluted, regaining focus is hard.
Solution
A browser extension that injects filtering logic into YouTube: keyword blacklist, channel blacklist, topic filtering, and allowlist mode; rules can be grouped and imported/exported; matched items are hidden in real time on the home feed and recommendation sidebar, with a count of hidden items.
Review a whole AI-generated changeset in one view with a file tree and syntax-highlighted diffs
Problem
After using Claude Code or similar AI coders, a single change often touches dozens of files. Traditional diff viewers like Meld force you to open files one by one and cannot show all files from a Git diff sequentially in the same view. AI output is dense and voluminous, and without syntax highlighting and a changed-blocks-only mode, review becomes painfully slow.
Solution
A desktop diff viewer that lists every file in a Git diff sequentially in one view with a side file tree; toggle between changed blocks only and full files; tree-sitter-based syntax highlighting; works against local Git repos, comparing working tree, index, and arbitrary commits.
See how your AI agent represents you in social networks with other agents
Problem
Someone wants to build an AI-native social network where each user has one agent that browses and socializes on their behalf. But agent-to-agent interactions are opaque to humans: users can't tell what their agent said or whether it overstepped, making trust hard to build.
Solution
An agent observability tool that connects to existing AI social/agent platforms (or self-hosted setups) and shows real-time summaries of your agent's conversations, a relationship graph, and anomaly alerts. Users set behavior boundaries and can retract inappropriate messages.
Discover, install, and update GitHub-hosted open-source apps in one place
Problem
Windows users download and install open-source software from scattered GitHub repos with no unified management UI, manual updates, and no discovery. Someone explicitly wished for a Windows open-source app store with recommendations and rankings, but no such wheel exists on GitHub.
Solution
A Windows desktop store that aggregates release info from GitHub open-source projects, offering one-click install, auto-updates, category recommendations, and community rankings. Data comes from the GitHub API and winget/scoop manifests; the app doesn't host binaries.
Control and slim down MCP tool calls for Claude and Codex agents
Problem
As the MCP ecosystem explodes, AI coding agents mount more and more tools, causing context bloat, token waste, and risky tool calls. Someone built CtxWise as a schema tree-shaker and firewall, proving this is a fresh, real pain point, but the tooling is still immature.
Solution
A middleware that sits between the agent and MCP tools: strips unused schema fields to save tokens, enforces allow/deny rules and approval flows for sensitive tools (file deletion, network, shell), and logs every call.
Cross-device Pomodoro and todos that sync without any sign-up
Problem
Many people want a cross-device Pomodoro and todo tool, but mainstream apps force account creation or rely on cloud services that raise privacy concerns. Someone built Pomoflow with no-sign-up sync, showing real demand, yet such tools remain fragmented and hard to migrate between.
Solution
A local-first Pomodoro + todo app that syncs peer-to-peer across devices via end-to-end encrypted WebRTC or LAN discovery, with zero servers and zero accounts. Users pair by scanning a QR code or entering a short code.
5-minute browser games for developers waiting on their AI coding agent
Problem
Developers increasingly wait for AI coding agents (Claude, Codex, etc.) to finish tasks. Waiting leads to social media doom-scrolling that destroys focus. Someone built a CS 1.6 1v1 match for this exact waiting window, proving the pain is real and frequent, but a single game lacks retention.
Solution
A curated collection of lightweight browser games designed for 1-5 minute waits: instant play, no sign-up, local progress saving. Optionally detects when a local agent is running and surfaces a game, then reminds you to return when the agent finishes.
See which Douban Top 250 and festival titles are missing from your library
Problem
Cinephiles chasing Douban Top 250, IMDb Top 250, and major festival winners struggle to check what's already in their Emby library. The same film appears on several lists under different localized titles, so manual cross-checking is slow and has to be redone every time a list updates.
Solution
Maintain curated TMDB lists (Douban, IMDb, Oscars, Cannes, Berlin, Venice) and automatically match them against your Emby/Jellyfin library, outputting collected / missing / duplicate lists and letting you export the to-acquire set. Match by TMDB ID plus an alias table to resolve title conflicts.
Auto-file your Mac Downloads into folders you already have
Problem
Mac Downloads folders fill up with files like download (3).pdf, IMG_4821.jpg, and Untitled.docx that give no clue what they are. Users already have topic folders under Documents but never move anything because they'd have to open each file first, so the pile just grows.
Solution
Read the start of each file to understand its real content: PDF text, OCR for scans and photos, Word, Excel, CSV, and HTML. Then suggest the best-matching folder the user already created. Show a preview first, move only after confirmation, and allow one-click undo.
An internal doc tool developers will actually keep updated
Problem
Developers have long hated Confluence for its heavy UI, templates, weak search, and disconnection from code, so nobody writes in it. The recurring community question of whether a wiki developers actually like exists has no good answer, and team knowledge ends up split between stale Confluence pages and scattered READMEs.
Solution
Start read-only: auto-extract changes from the repo, PR descriptions, and commit messages to generate a structured weekly changelog and per-service overview pages, with Git as the source of truth. Developers only add a sentence to existing Markdown, and the tool handles aggregation, indexing, and search.
Make your phone's AI agents ask first and follow your rules
Problem
Personal agent apps on phones (Muse, Grok Bot, Dots, Instinct) assume intent and act the moment they get tools, connectors, and data access, without asking the user or respecting stated rules, values, and goals. Users explicitly want agents to align with their own constraints, but no product exposes that boundary today.
Solution
A local rule layer between the agent and its tools: users write in natural language plus structured conditions when it must ask, what is always forbidden, and daily budgets. The layer intercepts tool calls, surfaces a confirmation card on a match, and lets the user allow once or whitelist permanently.
A cash flow forecast that actually accounts for credit card billing cycles
Problem
Existing finance apps (Personal Capital, Monarch, Copilot) model cash flow as the net sum of monthly transactions, but credit card charges are counted immediately while the money only leaves checking at the next statement due date. Users end up seeing a healthy balance that quietly turns into an overdraft.
Solution
Explicitly model each card's statement cycle, due date, and the checking-account debit, pushing every card purchase's cash impact to the actual payment date. Layer in recurring income and subscriptions to produce a day-by-day 60–90 day balance forecast and alert when the projected balance dips below a safety floor.
Manage all your side-project SaaS emails, subscriptions, and todos in one panel.
Problem
For those running multiple micro-SaaS products, the pain isn't coding — it's cycling through different email accounts, checking separate Stripe dashboards, replying to support tickets, and handling domain renewals. These chores drain energy.
Solution
An aggregator dashboard that connects multiple products' email, payment, domain, and analytics accounts, shows each product's MRR, unread support emails, upcoming renewals, and todos, with quick reply and renewal reminders.
Track your brand's exposure in ChatGPT, Perplexity, and other AI assistants — locally.
Problem
Existing AI visibility tools like Profound are highly valued but expensive; small teams and indie developers need a local-first way to track how often they appear in AI answers without a large monthly bill. Open-source projects like Canonry show the demand exists but are still early.
Solution
A lightweight local tool that periodically asks preset keywords to major AI assistants, scrapes whether your brand or site is mentioned, combines Google Search Console and site traffic data, and generates visibility trend reports. All data stays local and exportable.
Automatically log and alert when AI agents touch systems they shouldn't.
Problem
Tencent admitted its eval agents browsed where they were never sent, and India's cybersecurity agency issued a critical advisory after agents reached live systems and read confidential records — with nobody detecting it until the providers self-disclosed. The model is the easy part; everything around it is hard.
Solution
A local-first agent behavior auditor that hooks or middleware records each AI agent's actual accesses to URLs, files, API endpoints, and data tables, builds a timeline of overreach, and sends real-time alerts when operations exceed preset boundaries. Core is recording and alerting, not cloud-vendor dependent.
Detect and prevent massive API billing overcharges caused by region-specific number formatting.
Problem
A LegalTech attorney from Costa Rica tried to load credits onto his Anthropic API account and was overcharged 100x due to an internationalization input bug — in Spanish-speaking regions, a comma is used as decimal separator, but the payment pipeline misread it as a thousands separator, inflating the amount. Such i18n payment errors are disastrous for small teams and indie developers.
Solution
A lightweight payment gateway middleware that checks localized number formats, currency units, and locale settings before submitting any payment amount, normalizes the value, and prompts a confirmation to prevent catastrophic overcharges. Could also be a browser extension or iOS shortcut.
See at a glance if any of your parallel AI coding agents have quietly gone off-task.
Problem
When running multiple Claude Code sessions, the hardest part isn't writing code — it's managing terminals, tabbing between them, and checking whether an agent quietly wandered off and started refactoring something you never asked for.
Solution
A lightweight dashboard that collects each parallel AI coding session's goal, current actions, and file changes, flags off-task agents with color, and lets you jump back to the right terminal in one click. Core features are drift detection and status overview.
Scan if your AI coding assistant leaks sensitive data
Problem
When using AI coding assistants (e.g., Copilot, Cursor), they may send code snippets, keys, or other sensitive info to remote servers without user awareness. Recent discussions on AI assistant data leaks are growing, but simple self-check tools are lacking.
Solution
A local tool that monitors AI coding assistant network requests and file access, identifies potentially leaked sensitive data (API keys, passwords, PII), and generates reports. Supports real-time alerts and historical queries to help developers understand privacy risks.
One-click fix for macOS .local domain resolution conflicts
Problem
macOS mDNS resolution logic conflicts with .local domains, causing developers to fail resolving internal test environment domains (e.g., xxx.cloud.local) in tools like Postman and database clients. Even manual /etc/resolver configs often don't work, wasting hours.
Solution
A Mac utility that detects .local domain resolution conflicts, generates correct /etc/resolver configuration files, or provides a GUI to quickly add DNS servers for specific domains. Supports one-click resolution testing and gives fix suggestions.
Predict and free up disk space eaten by Siri and other system processes before macOS upgrade
Problem
macOS upgrades often consume large disk space due to system processes like Siri, causing failures or post-upgrade issues for users with limited free space. Users can't predict usage beforehand and are forced to clean up reactively, or avoid upgrading.
Solution
A Mac tool that scans and simulates disk space that system processes (Siri, Spotlight) may take before an upgrade, provides a prediction report, and offers one-click cleanup or disable suggestions to ensure smooth upgrade. Also analyzes post-upgrade space usage and frees it.