Automatically detect mobile UI issues that leak conversions
Problem
Many sites see poor mobile conversion, but developers don't know which UI element (tiny buttons, broken layout) causes drop-off. Manual checks are slow. A recent HN Show HN noted mobile conversion data can lie and showcased a responsive comparison tool, showing the pain is widespread but lacking automated diagnosis.
Solution
A SaaS tool that renders a given URL across multiple mobile viewports, detects tap-target size, overlap, and overflow issues, and generates a severity-ranked report with fix suggestions. Can integrate into CI to prevent regressions.
Semantic search over local notes, PDFs, and images on CPU, no cloud upload
Problem
Knowledge workers accumulate local notes, PDFs, and images, but keyword search fails to find semantically relevant content. Cloud semantic search raises privacy concerns. A recent HN Show HN demonstrated CPU-based semantic search, proving lightweight local cross-modal search is feasible but not yet user-friendly.
Solution
A local desktop app that indexes text, PDFs, and images (via OCR/CLIP) in chosen folders, offers natural language queries, and returns the most relevant snippets and files. All computation runs on local CPU without GPU or network.
An encrypted, Markdown-native writing experience for terminal dwellers
Problem
Many developers live in the terminal but must switch to GUI apps for blogging or note-taking, while also worrying about cloud privacy. A recent HN Show HN introduced an encrypted terminal Markdown app, showing users want to write securely in the terminal, yet such tools remain few and fragmented.
Solution
A TUI application with Markdown syntax highlighting, local file encryption (e.g., age or GPG), fast search, and export. All data stays local, with optional Git sync, enabling the full draft-to-publish flow inside the terminal.
Let developers write and edit code by voice, reducing keyboard dependence
Problem
Long coding sessions cause hand fatigue and RSI, but existing voice input tools handle code poorly, especially when away from a keyboard. HN posts about rewriting a site in Swift and using a phone as a Windows mouse/keyboard show sustained interest in alternative input, yet code-optimized voice tooling remains scarce.
Solution
A local desktop app that listens to the microphone, converts natural language or symbol dictation into code snippets, and inserts them into the active editor. It supports language-specific keywords, indentation, and symbol mapping, plus custom commands like 'define a function'.
Give AI agents persistent, forkable workspaces for easy rollback and parallel exploration
Problem
Developers tuning AI agents often lose context and struggle to try multiple approaches in parallel. Existing tools force rebuilding the entire session or don't allow branching from a key checkpoint. A recent HN Show HN highlighted persistent, forkable agent workspaces, proving real demand that generic tools don't meet.
Solution
A CLI or lightweight service that snapshots an agent's complete workspace (files, env vars, conversation state) locally, allowing one-click forks into independent branches. Developers can try different prompts or tool calls from the same state and compare results side by side.
Extract characters, places and events from your chapters and turn them into an explorable graph
Problem
As a long story grows, authors struggle to remember who did what, when and where. Existing writing tools and general AI assistants don't automatically pull story elements into a connected graph.
Solution
A writing aid that imports chapters from Google Docs or pasted text, automatically identifies characters, places, events and their relationships, generates a visual graph, and orders them by chapter timeline so authors can spot contradictions and gaps.
Paste text and see which phrases make it read like AI
Problem
Readers can increasingly spot hollow AI-generated prose, and writers, students and content creators get called out or rejected without knowing which specific phrases trigger that 'AI slop' feel.
Solution
A lightweight detector with 100+ common AI writing patterns (boilerplate, over-parallelism, vague qualifiers) that highlights suspicious sentences in pasted text and suggests more natural rewrites. Runs entirely in the browser.
Scan hospital reports and track home meds so families avoid duplicate prescriptions and bad reactions
Problem
When a doctor prescribes medicine for a child, the family often already has it at home and ends up paying twice. Parents also want to record food and drug reactions and track their own health metrics over the years, but the data is scattered across paper reports and phone photos.
Solution
A mobile app that uses OCR to digitize hospital reports and medicine boxes, lets users search existing home medicine by keyword, records visits, medication and adverse reactions for each family member, and charts health metrics over time.
Merge log files from different time zones into one unified timeline, entirely in your browser
Problem
Debugging often means juggling logs from multiple services and time zones with inconsistent timestamp formats, making manual conversion and sorting painful. Existing log tools usually require uploading data or deploying heavy services, raising privacy and cost concerns.
Solution
A local-first web tool that auto-detects common timestamp formats, normalizes them to a user-specified time zone, and merges multiple logs into a single chronologically ordered view. Files never leave the browser.
Ad-free, one-time-purchase classic card and board games for families and seniors
Problem
Classic games have been ruined by ads and subscriptions, and older family members just want to play a quiet game of solitaire without pop-ups and upsells. They also won't tinker with emulators or command-line tools to revive old software.
Solution
A collection of ad-free, no-IAP classic games sold as a one-time purchase, shipping on Windows, macOS, iOS, Android and web. Each game includes adjustable difficulty, local AI opponents, large fonts and high-contrast mode for older users.
Safely clone an aging business PC—including the boot partition—to new hardware for non-technical shop owners
Problem
After taking over a hotel, a relative found the management software on the old PC had no installer and only runs on that machine. Boot takes over 30 minutes, memory reporting is broken, and the disk can't be touched for fear of losing the software forever. They need a safe way to clone the entire disk, including the boot partition, but have no actionable guidance and no guardrails.
Solution
A migration wizard for non-technical small business owners: it detects boot mode, partition table, and disk health on the target machine, recommends a cloning plan, generates ready-to-run ddrescue/Clonezilla commands or boot media, verifies the clone, and walks through the disk swap step by step—with a mandatory back-up-first checklist.
Strip AI watermarks and metadata from images, videos, and docs locally in the browser
Problem
More and more generated or edited files carry AI watermarks and metadata. Users want them gone before uploading, submitting, or sharing, but they don't want to send sensitive files to a third-party server, and no simple local tool covers multiple formats.
Solution
A browser-only tool that processes files locally without upload: it detects and removes AI watermarks, C2PA metadata, and generation model markers from images, videos, and documents, with before/after comparison of filename, size, and preview, plus batch drag-and-drop.
Stop runaway OpenAI and AI API bills before they happen with email and phone alerts
Problem
When experimenting with AI APIs or running agents, a runaway loop, missing rate limit, or out-of-control agent can rack up a huge bill before you notice. Cloud budget alerts are coarse, delayed, and don't cover the OpenAI balances individual developers actually use.
Solution
A lightweight monitoring service: users connect their OpenAI (and similar) API keys, it polls usage and balance, applies hourly/daily/monthly thresholds, and alerts via email, Telegram, or webhook the moment spend spikes or approaches budget. It can also pause usage or throttle through a proxy layer.
Find and kill orphan dev servers and headless browsers that AI coding agents forgot, saving Mac RAM
Problem
AI agents often start dev servers and headless browsers and then forget to close them, draining Mac RAM every couple of hours. Users resort to randomly quitting apps and hoping, with no idea who is actually eating memory or where the orphan processes are.
Solution
A menu bar app that scans and lists orphan processes still using memory after their parent exited (dev servers, headless browsers), shows reclaimable RAM, and frees it in one click. It warns a few minutes before memory pressure hits and supports a whitelist so you don't kill something important.
Instant cross-reference lookup and AI-completed proofs for dense math textbooks
Problem
When reading books like The Elements of Statistical Learning, the text constantly says things like 'see (3.41)' or 'by Theorem 2.3', forcing readers to scroll back through the whole book. Whenever a proof skips steps, readers screenshot it, ask ChatGPT, then manually paste notes back into the PDF—a broken workflow.
Solution
A desktop math reader that detects formulas, theorems, and lemmas in PDF/EPUB and turns them into clickable anchors. Clicking shows the referenced content in a popover. For skipped proof steps, it sends the context and formula to an AI to return a step-by-step derivation, saved as an inline annotation. All annotations live locally and export to Markdown.
A non-custodial USDT/USDC checkout page and webhook for indie devs and small merchants—funds go straight to their wallet.
Problem
Accepting crypto payments forces merchants either to use platforms that custody funds and require private keys (risking exit scams and theft) or to build their own chain listeners and webhooks (costly and error-prone). Indie developers and small merchants need a simple, non-custodial, ready-to-use solution.
Solution
A non-custodial checkout tool: merchants register and configure their own receiving address; the platform generates a payment page, API, and webhook. Funds go directly to the merchant wallet; the platform never touches funds and never needs private keys or seed phrases. Supports multiple chains and common stablecoin pairs, with optional wallet-connect quick pay.
Compare official, aggregator, and reseller channels by model—price, hidden costs, and reputation—so developers pick cheap and reliable endpoints.
Problem
LLM API channels are exploding: official, aggregators (OpenRouter, Together, Groq, etc.), and resellers. Prices for the same model vary wildly, with hidden costs like cache pricing, free tiers, and minimum top-ups, plus risks of resellers wrapping models or shutting down. Developers must manually compare multiple spreadsheets and community reviews, with information badly fragmented.
Solution
A continuously updated price and reputation board that normalizes per-1M-tokens pricing across channels for major models (GPT, Claude, DeepSeek, GLM, Qwen, Kimi, Gemini, Grok, etc.), flags hidden costs like cache rates, free quotas, and minimum deposits, aggregates third-party benchmarks and community sentiment, and marks reseller risk and inclusion criteria.
Visualize Chinese public holidays and real factory shutdown windows so overseas buyers can predict vague lead times like 'after Chinese New Year'.
Problem
When ordering from Chinese factories, delivery dates are often vague promises like 'after Chinese New Year.' Overseas buyers and indie sellers don't know China's official holiday schedule, nor that factories actually shut down longer than official holidays, making it impossible to plan inventory and logistics accurately—leading to stockouts or overstock.
Solution
A supplier lead-time tool built around Chinese holidays: shows whether a factory is on holiday right now based on China time; for each holiday displays both the State Council official days off and the longer 'factory reality' window; supports adjustments by supplier city or industry; exports ICS calendars or integrates with Slack/email for advance N-day alerts.
Remember window positions by monitor hardware—screen and Space—and restore everything after plugging/unplugging an external display.
Problem
Commuting daily between home and office, every time an external monitor is plugged or unplugged all windows get squeezed back to the built-in display and scattered across different Mission Control Spaces, taking minutes to reorganize. Existing tools like Rectangle and Magnet only snap windows to halves; Stay and Display Maid restore window positions per display config but cannot put windows back to their original Space—so for multi-desktop workflows, position is right but desktop is still wrong.
Solution
A macOS window layout manager that identifies physical monitors via EDID (vendor/model/serial) and saves a layout per screen setup. On restore it not only moves windows to the correct display but also back to the original Space and window stacking, with one-click or automatic restoration after plugging/unplugging.
Turn the MacBook notch into a status light for coding agents, so devs can see which agent needs approval or has finished without switching windows.
Problem
Running Claude Code, Codex, Devin and other coding agents in parallel means each task takes many minutes. Once the developer switches to other work, there is no way to know when an agent needs approval, has crashed, or finished—forcing constant tab-switching back to the terminal and breaking flow.
Solution
A lightweight macOS notch-resident status island that auto-detects running agent processes (Claude Code, Codex, Grok Build, Devin, Qwen Code, etc.) and shows each agent's state via color and short text in the notch: running, awaiting approval, errored, done. Notifications appear only when human action is needed, cutting unnecessary interruptions.