I had AI read 4,400 Hacker News posts. These are the 10 tools people most want built.
Every week on Hacker News, a few people write things like "I'd pay for this", "I wish there was a tool that…" or "someone should build…". Each comment disappears into its thread. But when you collect them for three month
Every week on Hacker News, a few people write things like "I'd pay for this", "I wish there was a tool that…" or "someone should build…". Each comment disappears into its thread. But when you collect them for three months, patterns show up: the same need, described by different people in unrelated threads.
So I did that. A small script collects candidate posts from Hacker News every day: request phrases, every Ask HN post, plus a few Stack Exchange and GitHub requests. An AI model throws out the jokes, quotes, support questions and self-promotion, and I grouped what was left by hand.
The numbers: about 4,400 posts screened, 790 real requests kept, Aug 19 – Nov 17, 2026. Then I ranked needs by how many different people asked. Saying "I'd pay" counts double.
Here are the top 10.
1. Sandboxing and permission controls for AI coding agents
11 people asked
Developers want hard limits on what coding agents can touch: which files they may read or modify, which shell commands need approval, and where they must stop and wait.
“The second there is a visual interface from which to easily configure "these files are fine to read, these to modify" before the task is run, I will agree fully.” (HN comment)
Build angle: A small, harness-agnostic policy layer (a file- and command-allowlist with a visual editor and an approval queue) that wraps Claude Code, Codex and open-source agents alike.
2. Private AI with zero data retention
7 people asked, 4 said they'd pay
People and companies want frontier-quality AI without their prompts being stored or trained on, and several say they would pay a premium for it: 2-3x enterprise pricing, or about $20/month extra.
“we would even be willing to pay a premium of 2x or 3x the price to have the enterprise benefits of private data that isnt retained.” (HN comment)
Build angle: An OpenAI-compatible API on top of open models with contractual zero retention, published audits and simple per-seat pricing for small teams.
3. Power-user tools for Hacker News
11 people asked
Hacker News regulars keep asking for things HN itself will not build: dark mode, custom CSS, following users, reply notifications, collapsing repetitive comments, filtering low-quality posts and a usable Who's Hiring view.
“Would anyone be interested in a more user-friendly frontend to the whoishiring thread?” (Ask HN)
Build angle: One polished browser extension (or alternative client) that bundles these features and a filtered Who's Hiring board. The audience is exactly the people who share tools.
4. Secure, auto-updating home routers
8 people asked, 3 said they'd pay
Home users want a router that is secure without babysitting: OpenWrt-based, auto-updating and pre-configured with ad-blocking and separate networks for kids, guests and IoT devices.
“I would pay if I could get an openwrt router that was already set up with features like - ad-blocking - kid network vs grown-up network vs guest network” (HN comment)
Build angle: Sell pre-flashed OpenWrt hardware with a friendly setup app, automatic updates and an IoT allowlist as a subscription-free product.
5. Small, minimal and 'dumb' phones
7 people asked, 3 said they'd pay
Demand for small and deliberately limited phones keeps resurfacing: a modern iPhone-5-sized flagship (one commenter would pay $2,000), a flip phone that still runs MFA apps, and pocket devices without radios or cameras.
“Would happily pay a big premium for a modern version of it. Dreaming of a sub-mini phone trend. Sell me a $1500 Zoolander phone—not even joking!” (HN comment)
Build angle: Hardware is hard for small teams, but a curated store or configurator for minimal and 'semi-dumb' phones and watches (with honest MFA, banking-app and kid-safety compatibility notes) is very buildable.
6. Ad-free, user-first media and streaming apps
7 people asked, 6 said they'd pay
Strong willingness to pay for media without ads and enshittification: an ad- and sponsor-free YouTube client, a Plex alternative that puts users first, a managed media server, an open music API and one view of what's new across all streaming services.
“What is new this week? (across all the services I'm subscribed to) What service is show X on? (localized to my country) When will the next season of X be released? I'd pay for that (grudgingly, given the big …” (HN comment)
Build angle: A 'what's on and where' dashboard across the streaming services you already pay for, localized by country, with release alerts. Six of the seven people said they would pay.
7. Filters for AI-generated slop on feeds, forums and support chats
8 people asked, 1 said they'd pay
People want help filtering AI-generated content out of the places they read: YouTube, HN, GitHub feeds, forums and even helpdesk chat widgets spammed by AI agents.
“Would HN consider removing the many filler LLM comments on this site if I pay $9/mo to only see real people posts?” (Ask HN)
Build angle: A crowdsourced 'slop' label layer: a browser extension where users flag AI-generated content and others hide it, starting with YouTube and Hacker News.
8. Reviewing and verifying AI-generated code
8 people asked
Teams are drowning in AI-generated code and ask how anyone reviews, verifies and signs off on it at that volume.
“How do you read, sign off, verify, manage the volume of generated slop?” (Ask HN)
Build angle: A review tool that groups an agent's changes by intent, explains each diff in plain language and flags untested behavior changes before a human approves.
9. Orchestrating many AI agents in parallel
9 people asked
Running several coding agents at once creates chaos: decisions and questions get lost in transcripts, agents keep working after plans change, and output piles up unreviewed.
“Once I run several coding agents at a time, I get overwhelmed and disorganized. Every session makes decisions and raises questions, and I can't reread the transcripts.” (Ask HN)
Build angle: A dashboard that tracks every agent's task, open questions and decisions in one shared list, with a pause-all and reassign button.
10. Tracking model regressions and switching models safely
8 people asked
Builders want to know when a model update silently got worse, how to test a new model against real traffic before switching, and how to route prompts to the cheapest model that can handle them.
“I'd love to see OpenAI, Anthropic, and other LLM providers offer something similar: an official API-compatible test endpoint that doesn't invoke a model or consume billable tokens.” (Ask HN)
Build angle: A 'nerf tracker': continuous public and private evals on your own prompts, with alerts when a provider's quality drops and a replay tool for model migrations.
What stood out
- AI coding agents need a control layer. The biggest cluster isn't about new models. It is the tooling around them: sandboxes, secrets, reviewing AI-written code, keeping parallel agents in sync, and tracking regressions and cost.
- People pay to remove things. Ads, smart features, UI redesigns, recommendation algorithms, AI slop. These requests have the highest share of "I'd pay".
- Privacy is a feature people price in. Several commenters would pay 2-3x for AI that doesn't retain their data.
Caveats
HN skews heavily toward developers, the counts are small, and this is a map of signals, not market sizing. Read the original threads before building anything. The replies are often the best market research.
The daily feed this is built from is free at hn-idea-radar.pages.dev. I also put all 100 ranked needs (with quotes, source links and quick competitor checks) into a PDF report if you want the full list.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.