A desktop-app-like sidebar for Claude Code CLI with full session context
Problem
The Claude desktop app is great, but daily work means using both CLI and app; the CLI lacks a sidebar showing directory, branch, model, context usage, sub-agent status, and linked PR/Issue info.
Solution
A CLI plugin or local sidebar that aggregates the current session's directory, Git branch, model, context and quota usage, sub-agent timing and tokens, and live status of PRs/Issues mentioned in the session.
Unify and compare branching conversations across multiple AI chat windows
Problem
Users keep ChatGPT, Claude, and Ollama windows open and rabbit-hole into interesting answers, but the mental chain of thought gets lost and comparing the same branch across models is impossible.
Solution
A cross-platform chat history aggregator that captures conversation trees from each window, lets users merge branches manually, tag them, and view side-by-side multi-model answers to the same prompt.
A free, no-login webmaster toolkit with common utilities in one place
Problem
Webmasters and developers often need scattered web tools like encoding/decoding, URL parsing, IP lookup, or regex testing. Existing sites force logins, are ad-ridden, or upload data to servers, hurting privacy and efficiency.
Solution
A suite of pure front-end, locally running webmaster tools where all computation happens in the browser with no data upload and no login. Covers common tasks like UTM building, encoding conversion, HTTP header analysis, and DNS lookup, with a clean UI and dark mode.
Zoom into tiny screenshot text on V2EX without leaving the thread
Problem
Browsing V2EX often brings screenshots embedded in posts or replies with text too small to read. To see details, users must open the original image and switch back, which is especially inconvenient for long images and interrupts reading flow. Similar plugins exist, but the clear need and positive feedback show sustained demand.
Solution
A lightweight Chrome extension: click any image in a V2EX thread to open a large viewer with scroll zoom, drag-to-pan, one-click fit-to-window, and Esc to close. When an image links to the original, the viewer opens directly instead of navigating, preserving reading continuity.
A simple dashboard to track your product's mentions in ChatGPT/Perplexity answers
Problem
AI search is replacing traditional search, but founders don't know if their product is recommended by ChatGPT or Perplexity. Manual testing doesn't scale, and existing AI search trackers are bloated enterprise dashboards charging $200/mo—small teams can't afford or need them.
Solution
Solves one core question: is my product/brand mentioned and how well ranked in major AI search answers? Users set a few seed questions, the tool periodically queries AI and parses answers, giving a simple visibility score and trend without complex charts.
Link Git blame with coding agent history to explain why each line exists
Problem
Code is increasingly generated by AI agents. Git blame shows the commit and author, but the original request and reasoning behind each line are lost in closed sessions. When developers need to understand or modify AI-generated code, they lack context and can only guess.
Solution
Locally saves coding agent sessions and uses git blame to automatically associate files, lines, or commits with the corresponding agent session, showing the original request and rationale. It also finds user edits after the session, forming a traceable chain of code intent.
WebBluetooth/WebNFC debugging tool for hardware devs, no spare Android needed
Problem
When building NFC/BLE hardware products, iOS browsers lack WebBluetooth and WebNFC support. Developers must rely on native apps or a spare Android phone to test new features, which is slow and breaks flow. Existing tools are scattered, with no ready-to-use debug entry point.
Solution
An iOS browser (or standalone app) that directly exposes standard WebBluetooth/WebNFC JavaScript interfaces, enabling any existing web app to run on it. Developers can quickly validate hardware interaction logic on iPhone, inspect broadcast packets and NFC read/write results in real time, and save common debug scripts.
Put ad revenue and cloud bills on one chart so indie devs know if they are actually profitable
Problem
One developer shared their mini-program breakdown: only interstitial ads enabled because other ad ecpms are pitiful, a 3D model feature blew up CDN traffic costs, and the final tally was a 1659 yuan loss. Another posted revenue screenshots with ecpm peaking at 670. Their numbers differ by orders of magnitude, but neither had a tool to tell them which cost to cut.
Solution
Connect ad platforms and cloud billing to automatically aggregate revenue and costs, plot daily net margin curves, and flag cost ratios; alert when CDN, storage or ad revenue anomalies appear and simulate 'turn off this mode / switch to cheaper tier' scenarios.
See whether ChatGPT and Claude mention you when buyers ask, and who they cite
Problem
One team systematically checked and found ChatGPT and Claude never mentioned them across 29 simulated buyer questions, then dug into how the models pick sources. As buyers increasingly ask AI directly, being absent from generated answers is being silently invisible.
Solution
Enter the questions your buyers actually ask plus your brand name, and the tool periodically runs the major models to report mention rate, competitor share of voice, and the sources being cited; it then suggests where to publish content — forums, docs, review sites — to get picked up.
Capture the decisions that live only in people's heads, before an agent's helpful refactor breaks prod
Problem
An engineer with eight years on legacy frontends notes AI handles anything with a standard answer but cannot handle 'why can't this field be deleted' or 'this ugly code worked around a bug once' — that knowledge lives in people's memory, not the code. When an agent optimizes it away, prod breaks, and the knowledge leaves with the old-timers.
Solution
A lightweight decision memory: one annotation line in code links to a 'why it is this way' note, and the tool aggregates them into a visual landmine map. Coding agents get a read-only query API to check history before editing and auto-append a change rationale afterwards.
Add a lock to ⌘-Tab so blind switching while coding stops getting hijacked
Problem
Heavy agent users keep one eye on the coding agent and the other on docs in the browser; the highest-frequency gesture is a blind ⌘-Tab tap. But chat apps like WeChat keep jumping into the switch chain and breaking focus. These pains are small, unglamorous, and happen daily.
Solution
A macOS menu-bar utility that pins frequently paired windows into a controlled switching ring; ping-pong mode bounces only between two targets and temporarily excludes other apps from the ⌘-Tab sequence, with one-key pause, hotkey summon, and auto-profile switching based on the focused app.
Turn one group photo into a pack of transparent stickers, batch-exported
Problem
Most photo-to-sticker tools only cut out a single subject, so photos with groups of people or a table full of objects require repeated uploads and manual steps. One hobbyist tool that does support splitting is still fighting accuracy, first-run speed and batch UX — the need is real but the supply is rough.
Solution
Upload one photo and automatically detect people, pets and common objects, cutting each into a transparent-background sticker; support folder-level batch uploads, uniform outline width and color adjustment, and one-click ZIP or platform-specific export, with processing done locally in the browser for privacy.
Help people who struggle with business emails make drafts professional without sounding AI-generated
Problem
Many people run their outreach emails through AI before sending, but worry the tone sounds too AI-like and gets ignored. They want emails that are professional, clear, and human — not obvious machine-generated boilerplate.
Solution
Provide a lightweight tool where users paste a draft, choose a style like 'more professional', 'more conversational', or 'shorter', and get a rewrite that preserves personal voice and key information; include a checklist to avoid common AI phrasing and explain the changes.
Help local small businesses detect competitors maliciously clicking their Google Ads
Problem
Local small businesses find their Google Ads are being maliciously clicked by competitors who use cheap click bots through proxy services to simulate different users, wasting ad budget and reducing real customer calls. Most merchants don't know how to detect or respond.
Solution
Provide a lightweight monitoring tool that analyzes ad clicks by IP, device fingerprint, timing, and behavior patterns to identify suspicious clicks and generate reports; also offer blocking recommendations and appeal templates to help merchants reduce losses.
Help concert fans easily log every show, save favorites, and auto-generate poster walls
Problem
Existing concert-logging tools either look ugly or have tedious data entry; after each show users still have to find information and fill in details themselves, making logging a high-friction chore. And the Damai API isn't open to individual developers, so there's no convenient concert database available.
Solution
Build a database covering 100k+ shows so users can search by artist, find the event, and save it with one tap; support logging spend, ratings, and setlists, and auto-generate a personal concert poster wall.
Let AI coding agents operate a real iPhone from Linux without juggling multiple tools
Problem
Developers who want AI to operate a real iPhone have to stitch together go-ios, pymobiledevice3, WDA, and other tools — each requiring separate installation and signing, tunnels dropping constantly, and custom scripts to babysit the process. The workflow is fragmented and unreliable.
Solution
Provide a unified Rust-based tool exposing screenshot, tap, swipe, text input, app launch, and other actions via MCP protocol, letting AI agents read UI elements and act directly with fewer round trips, eliminating multiple standalone tools and signing steps.
A fast, native desktop Markdown editor for people working with very large files
Problem
Many users complain that mainstream Markdown editors (Typora, Obsidian, MarkText) are mostly Electron wrappers with slow startup, high memory usage, and lag when opening large files — a 25MiB document takes seconds to load and 100MB files are nearly impossible to edit. Users who want speed and a native feel have no real lightweight alternative.
Solution
Build a pure native Markdown editor in C++/Qt without bundling Chromium, offering fast cold start and low memory usage; implement lazy-loading viewport layout so huge files open and edit smoothly; support WYSIWYG, Mermaid diagrams, and math formula rendering.
Collect friends' and family's gift needs before the sale so you only buy the right things
Problem
Every year before Singles' Day people scramble asking what to buy; gifting family and friends is guesswork about sizes and needs. Forums fill with "what are you buying this year" threads, and comments scatter across dozens of replies with no way to aggregate.
Solution
A shared gift list where each person anonymously or openly enters needs, sizes, budget and priority; the app cross-checks platform discounts before the sale and reminds you to order. An anonymous collection mode avoids the awkward direct ask.
Ship temporary mitigations for Linux CVEs before the vendor patch lands
Problem
For a 9.8 Ubuntu CVE, the vendor fix can take days or weeks, while AI-discovered disclosures flood in faster than teams can respond. Small teams without a security function wait passively and piece mitigations together from advisories and forums.
Solution
Subscribe to CVE feeds, auto-generate temporary mitigation plans for high-severity Linux CVEs affecting your hosts (config changes, iptables rules, service degradation, rollback scripts), and run them with one click plus verification. Prompt rollback once the vendor patch ships.
Automatically catch the Chrome extension update that maxes out your CPU
Problem
Browsers freeze and run hot, so users disable extensions one by one; a single V2EX Polish update left a Service Worker in an infinite loop eating a full core, and only a thread of users confirmed it. Nothing tells you which extension started misbehaving after yesterday's update.
Solution
A side panel samples each extension's background CPU and memory, aligns samples with extension update times, and flags ones that started hogging after an update, with one-click disable or rollback. Alerts can be pushed to phone.