Flag risky clients the moment their first email lands in your inbox
Problem
Freelancers and small studios regularly get burned by scope creep, late payment, and ghosting. Many warning signs appear in the very first client email: phrasing, who is cc'd, whether they answer your questions, how fast "one quick extra thing" shows up. Nobody has time to check all of that when a lead comes in.
Solution
An email sidebar tool that connects to Gmail/Outlook and scores new-client first emails against a red-flag checklist, surfacing warnings like unanswered budget questions, unclear decision chains, or bargain-hunting signals. It also offers contract and payment-term templates.
Let writers actively prove a text was written by a human, verifiably
Problem
AI text is increasingly hard to distinguish from human writing. A common observation: labs will never solve origin verification, but people who want to prove they wrote something have an incentive to voluntarily collect evidence. Today there is no lightweight, third-party-verifiable way to do that.
Solution
A writing-time attestation tool: it records keystroke cadence, version snapshots, and timestamps locally, then produces a shareable verification page. Readers can confirm a text was human-typed up to a point in time without trusting any AI detector.
Measure and trim the context your AI coding agent actually consumes
Problem
When project files get long, agents quietly start ignoring the bottom half; developers build tools to measure where AI coding context goes. Yet most devs have no idea which rule files, pasted snippets, or redundant context are wasting tokens and attention every session.
Solution
A local editor/agent plugin that breaks down a session's context into system prompt, rules files, open files, chat history, and tool output, highlighting duplicates and low-value blocks. It suggests one-click trims and estimates token savings.
Surface the suspicious lines in AI-generated diffs so devs review only what matters
Problem
Developers using AI coding agents face a lose-lose: reading every line of a multi-file diff is exhausting, but skipping review risks silent breakage. Teams report small changes shipped past three review stages that still broke production, because reviewers left P2 comments while P0 issues slipped through.
Solution
A local CLI/IDE plugin that ingests an AI-generated diff and ranks changes by risk: removed auth checks, altered routing/config, quietly edited test assertions. It produces a short "must confirm by human" list and folds the rest, with per-repo whitelists for files AI may change freely.
An install-free browser FPS aim trainer with cross-game sensitivity conversion
Problem
A developer hooked on FPS games wanted a few minutes of aim practice before playing but refused to install a client for it, so they built a web-based range with Gridshot, Sixshot, and other scenarios, plus Valorant-style sensitivity conversion and crosshair code import. Players lack a lightweight, no-install, no-signup aim practice entry point, and they want their sensitivity to carry across games.
Solution
A browser-only FPS aim trainer with common practice scenarios, sensitivity conversion across Valorant, CS2, and Apex, one-click crosshair code import, and a weekly leaderboard requiring only a nickname. No download, no signup, just open and practice.
Split AI coding tool bills by client for freelancers and agencies
Problem
Someone built Tallyhook to price Claude Code and Codex logs per client, while another developer scripted seven projects to track mixed frontier and budget model usage. This shows per-project and per-client AI coding cost tracking is a real need. But existing tools cover only one provider or show aggregates, so freelancers can't split bills by client or paste numbers into an invoice.
Solution
A local-first cost allocation service that ingests usage from AI coding tool logs and API keys, tags it by preconfigured project or client, and generates exportable monthly billing details. Users can combine tokens, seat subscriptions, and self-rolled scripts under a single client line item ready for an invoice.
A desktop shell that brings split panes, notifications, and persistent sessions to Windows terminals
Problem
One developer envied Mac terminals for their split panes, persistent sessions, message notifications, and command completion, then built their own because Windows lacks a comprehensive option. Another user complained that tmux and TUI-based agent managers feel awkward to copy from and too heavy, showing abilities a terminal doesn't need. Windows users and lightweight agent runners lack a desktop-grade terminal shell that stays lean.
Solution
A lightweight terminal shell for Windows that keeps the system shell and layers on split panes, session persistence, fuzzy command completion, system notifications when long tasks finish, and output highlighting. It doesn't reimplement terminal emulation or tmux paste modes, reusing ConPTY and focusing on the desktop experience layer.
Auto-route tasks between frontier and budget models for indie devs
Problem
One developer running seven projects found a workflow of using frontier models for kickoff and reviews, and budget models for daily coding, cutting costs by 80%. But the method is hand-rolled: no tool records which model handled which task, what it cost, or how quality changed. Others in the same community debate whether Haiku 5.5 beats 6Luna and want free switching between providers. Everyone picks models by gut feel without cost-quality data.
Solution
A CLI/desktop tool that sits between AI coding agents and API providers, auto-routing by task type: architecture and review go to frontier models, autocomplete and test runs go to budget models. It logs each call's model, tokens, latency, and whether the output was accepted, generating a personal cost-quality dashboard plus head-to-head model comparisons.
A compliant hosted playground for browser-based retro game runtimes
Problem
Multiple developers in the community have independently brought Red Alert 2, Warcraft III, and Yuri's Revenge to the browser via v86 + WebGL. Each demo works in isolation, but every author has to rebuild CDN delivery, file handling, multiplayer matchmaking, and player-uploaded game files from scratch. Players either can't find an entry point or get stuck uploading local files and joining matches.
Solution
A portal that hosts these v86/WebGL runtime games. Authors upload their runtime and assets, then get an embeddable iframe and short link. The platform handles WASM delivery, asset caching, room creation with NAT traversal, and replay viewing as shared infrastructure. Players bring their own original game files and play without installs or signup.
Before upgrading dependencies or refactoring legacy code, see where changes are most likely to break something
Problem
Developers inheriting or modernizing legacy projects fear breaking one thing and toppling ten others because they cannot see which dependencies are coupled and which files are high risk. Some spend years studying system design yet still hesitate to touch real production code. The pain of understanding change risk in existing code is widespread.
Solution
Statically analyze a local codebase and produce a change-risk map: high-coupling modules, widely referenced internal utility functions, fragile dependency chains, and critical paths with no test coverage. Developers consult it before upgrading dependencies or refactoring to decide where to start.
One-click extraction of table data trapped inside iframes, virtual lists, and Canvas
Problem
Everyday work is full of data locked in iframes, paginated lists, virtual scrolling, or Canvas-rendered tables. Copy-paste fails and writing a scraper is overkill. A developer already shipped a Chrome extension covering iframes, pagination, and some Canvas cases, confirming this is a real, frequent pain.
Solution
A browser extension that auto-detects tables, iframes, and virtual lists on a page, with preview, column selection, and export to CSV, TSV, JSON, or XLSX, all processed locally with no upload. Add dedicated adapters for known heavy cases like Feishu sheets.
Find and fix drifting, unindexed, and misconfigured tables in Supabase projects
Problem
Many indie developers and small backend teams run on Supabase and iterate fast, so schemas quietly drift: missing indexes, mismatched field types, and RLS policies that were never configured. Problems only surface when queries slow down or permissions break. A Show HN tool dedicated to finding broken Supabase tables shows real demand.
Solution
Connect to a Supabase project and regularly scan schema and RLS policies for missing indexes, type conflicts, disabled RLS, and orphaned foreign keys, then suggest copy-pasteable SQL fixes. Read-only diagnosis and recommendations, no invasive migrations.
An explainable disk audit for bloated MacBooks that tells developers what is safe to delete
Problem
Long-time Mac users accumulate node_modules, Xcode caches, Docker images, simulators, and build artifacts. Apple's built-in storage manager is vague, and third-party cleaners often delete the wrong things or cost too much. Developers are already building and sharing macOS disk diagnostics, which shows this is a frequent, hard requirement.
Solution
Scan the disk and group usage by developer toolchain: Xcode, Docker, Node, Python, simulators, logs, etc. Each item shows size, last access time, and a clear safety verdict. Risky deletions go through Trash or a quarantine step with undo.
Make coding agents like Claude Code remember past sessions so they stop re-reading files and repeating the same mistakes
Problem
Developers complain that coding agents start every session from zero: ask for a change similar to last week's and the agent re-reads the same files, falls into the same test traps, and you pay for all of it again. Multiple Show HN projects are reinventing this wheel, which signals a real, recurring pain around reusable agent memory.
Solution
Locally record, for each finished session, the change type, target files, verification steps, and mistakes made, building a searchable repo-level memory. On the next agent run, inject the relevant memories into the prompt or expose them via MCP so the agent skips redundant exploration. Core value: local, per-repo, zero-config.
Warn users before they download software from a spoofed official site
Problem
A user searching for 'Soda Music' landed on a fake site, downloaded a poisoned installer, and didn't notice for three weeks. Search results and UI are highly convincing, making it hard for average users to tell. Developers have no easy way to warn users about impersonators.
Solution
A browser extension or lightweight API that checks software download pages using domain age, page fingerprinting, and community reports to warn if the site is likely fake. Developers can claim official download URLs for verification.
Make Whisper's Chinese transcription accurate enough to actually use
Problem
A V2EX user reports Whisper-large-v3-turbo works well in English but Chinese output has errors in almost every sentence, hurting meeting notes and podcast transcription. Switching models is costly, and many users need local deployment.
Solution
A local-first post-processing tool that uses lightweight Chinese correction models or rule engines to fix Whisper output, restore punctuation, and optimize segmentation. Supports batch audio processing and can be embedded as an API.
Auto-generate maintainable E2E tests for multi-service, multi-queue SaaS systems
Problem
Complex SaaS systems have business logic scattered across frontend and backend. A core flow can affect a dozen related services via transactions, MQ, or scheduled jobs. Traditional API integration tests can't keep up with AI-accelerated coding, and writing E2E tests manually is too costly.
Solution
An E2E test assistant that analyzes repos and service dependencies to auto-generate test cases covering core business flows. It supports DB snapshots, MQ interception, and scheduled job simulation, making E2E tests as maintainable as unit tests.
Decide when an AI agent in a group chat should stay silent, not reply to everything
Problem
Teams adding coding agents to Slack/GitHub find agents don't know when to speak. Every message triggers a full LLM call to decide if a reply is needed, adding noticeable latency, and the model often returns explanations instead of a clean boolean, forcing string parsing hacks. Eventually the channel gets noisy and the agent is removed.
Solution
A lightweight 'should-reply' middleware that uses a typed, probabilistic decision model instead of full LLM calls to quickly decide if a message needs agent attention. Supports rule config, confidence thresholds, and quiet periods to cut latency and cost.
Let indie developers distribute invite codes/keys without manual replies in threads
Problem
After shipping an AI app quickly, indie developers invite friends to try it. Threads fill up with 'XXX claimed' and 'XXX used', forcing the developer to manually skip claimed codes and track nothing. It's slow and error-prone, with keys leaking publicly.
Solution
A lightweight invite distribution page: upload a batch of keys, choose claim modes (direct claim / email delivery / manual binding), auto-deduplicate, track statuses, and show a live dashboard. Developers see who claimed what and can export data.
Track terminology drift and version chaos in AI-assisted data analysis teams
Problem
Data teams using AI for analysis face terminology drift, inconsistent vocabulary, and algorithm version chaos leading to varied results. One user accumulated 500+ issues and noticed AI error rates rising, fearing 'AI will say sorry and undo all previous work.' The rigor of Rmarkdown/notebooks is diluted in the AI era.
Solution
A lightweight workflow guard that logs inputs, outputs, function versions, and terminology definitions for each AI-assisted analysis, auto-detects drift and version conflicts, and generates traceable reports. Integrates with Jupyter/RStudio and flags uncertainty in AI-generated code.