DEV Community: Milind Gupta

Why Unlure makes you wait three days to turn blocking off

Milind Gupta — Fri, 23 Oct 2026 11:28:54 +0000

Disclosure: this article was drafted with AI assistance. I built Unlure.

Most blockers can be switched off in a few seconds. I built Unlure (https://unlure.in), a porn blocker for Windows and Android, around the opposite choice: turning protection off, unblocking a site or switching off a filter each takes three days. Protection stays on while you wait.

This post explains the idea behind that wait, using two public sources: a 2023 study of a phone app that adds a short pause, and a rule that regulates online gambling in Britain. Neither tested Unlure, and neither says three days is the right length. They show why a pause before acting is a recognised design idea, and Unlure is my application of it.

A short pause changes what people do

In 2023, researchers published a field study in PNAS of one sec, a smartphone app that makes chosen apps wait 10 seconds before opening and offers a button to close instead. The researchers followed 280 participants over six weeks. On average, in 36% of attempts to open a chosen app, people closed it again after the pause. Over the six weeks they also attempted to open those apps 37% less than in the first week. Together, the authors report that one sec decreased actual openings by 57% after six weeks.

A preregistered online experiment with 500 participants then tested the app's parts. The option to dismiss had the strongest effect. The authors report that the time delay also played a role.

The limits matter. The six-week field study compared people with their own first week and had no control group. Participants chose to install the app, and the work covered phone apps, not blocking adult content. One of the authors, Frederik Riedel, created one sec, and the paper discloses this.

Tighten now, loosen later

The UK Gambling Commission sets technical standards for online gambling operators. Requirement 12D covers the financial limits customers set on their own accounts. A customer's request to lower a limit must be applied immediately, unless a system or technical failure prevents it. A request to raise a limit can only happen after a cooling-off period of at least 24 hours, and only once the customer takes a positive action at the end of that period to confirm.

The rule is about gambling limits, not blockers. I found it useful as a public example of the same asymmetry: making yourself safer should be instant, and making yourself less safe should take longer.

How Unlure applies the idea

Unlure uses that asymmetry. Adding a site to your block list or switching a filter back on happens at once. Turning protection off, unblocking a site, switching off a filter or shortening the wait takes three days. During the wait you can cancel the request at any point and nothing changes. When the wait ends, a code appears in your dashboard. Type it in and protection switches off. If you do nothing with the code for a day, the request lapses and protection stays on.

Three days is my design choice for Unlure. The studies above did not test it, and nothing here shows it works better than another length for any particular person.

What Unlure is

A blocker for adult sites in every browser and app on Windows 10/11 and Android. iPhone, iPad and Mac are not supported yet.
Designed to work on the address lookups a device already makes, rather than routing pages through Unlure.
Designed not to record the sites you visit, screenshots, what you type, your location or which apps you use. The dashboard shows your settings, your blocked sites and counts of lookups blocked and answered.
Not a medical product. It blocks; it does not diagnose or treat anything.

Limits worth knowing

On Windows, someone with an administrator account can still remove it by hand. For stronger protection, use a standard account day to day and keep the administrator password with someone you trust.
Unlure is new, so Windows may show a warning when you open the installer.

Pricing

Unlure is sold as time-limited packs of 1, 3 or 6 months, from Rs 199, paid once per pack by UPI or card through Dodo Payments. Nothing renews by itself. When a pack ends, protection stays on for three more days and then switches off. If Unlure does not work on your device, you can ask for a full refund within 7 days of paying.

Details and the case studies: https://unlure.in and https://www.unlure.in/case-studies

Sources

Grüning, Riedel and Lorenz-Spreen (2023), Directing smartphone use through the self-nudge app one sec, PNAS: https://www.pnas.org/doi/10.1073/pnas.2213114120
UK Gambling Commission, Remote gambling and software technical standards, RTS 12 Financial limits (requirement 12D): https://www.gamblingcommission.gov.uk/standards/remote-gambling-and-software-technical-standards/rts-12-financial-limits

Frontend machine coding rounds: the 5 patterns I would practice first (with code)

Milind Gupta — Fri, 02 Oct 2026 13:02:52 +0000

Machine coding rounds at Indian product companies are a different skill from DSA rounds. You get 45 to 90 minutes, a blank editor, and a spec like "build a star rating" or "implement debounce". Nobody wants the cleverest answer. They want working code, clean structure, and edge cases you thought of before they asked.

After looking at a lot of these problems, I noticed the same few patterns come up again and again. If you only have a weekend, practice these five first. For each one I'll show a solution that works, what the interviewer is really checking, and the mistake I see most.

1. Debounce

Delay a function until the caller has stopped calling it for wait ms.

function debounce(fn, wait) {
  let timer;
  return function (...args) {
    clearTimeout(timer);
    timer = setTimeout(() => fn.apply(this, args), wait);
  };
}

What they check: the timer lives in a closure, not a global; this and arguments are passed through; a new call resets the clock.

Common mistake: using an arrow function as the returned wrapper and then losing this, or forgetting clearTimeout so every call still fires.

Follow-ups to be ready for: a cancel() method, and a leading option that fires on the first call instead of the last.

2. Throttle

Run at most once every limit ms, no matter how often it's called.

function throttle(fn, limit) {
  let last = 0;
  return function (...args) {
    const now = Date.now();
    if (now - last >= limit) {
      last = now;
      fn.apply(this, args);
    }
  };
}

This is the simple "leading edge" version. Interviewers often follow up with "what if the last call gets dropped?" That is the trailing-edge version, which keeps one pending call and fires it when the window ends. Say out loud which one you are building before you write it.

Debounce vs throttle in one line: debounce waits for silence (search box), throttle caps the rate (scroll handler).

3. Flatten an array

Collapse [1, [2, [3, [4]], 5]] into [1, 2, 3, 4, 5].

function flatten(value) {
  const out = [];
  for (const item of value) {
    if (Array.isArray(item)) out.push(...flatten(item));
    else out.push(item);
  }
  return out;
}

What they check: you detect arrays with Array.isArray, you don't mutate the input, and you handle arbitrary depth and empty arrays.

Follow-up: "do it without recursion". An explicit stack works:

function flattenIterative(value) {
  const stack = [...value];
  const out = [];
  while (stack.length) {
    const item = stack.pop();
    if (Array.isArray(item)) stack.push(...item);
    else out.push(item);
  }
  return out.reverse();
}

Also know that arr.flat(Infinity) exists. Mentioning it, then writing your own anyway, is the right move.

4. Deep clone

JSON.parse(JSON.stringify(x)) is the answer everyone gives first, and it is the one interviewers push back on: it drops undefined, functions and symbols, turns Date into a string, and throws on circular references.

function deepClone(value, seen = new WeakMap()) {
  if (value === null || typeof value !== "object") return value;
  if (seen.has(value)) return seen.get(value);

  if (value instanceof Date) return new Date(value);
  if (value instanceof RegExp) return new RegExp(value.source, value.flags);

  const copy = Array.isArray(value) ? [] : {};
  seen.set(value, copy);

  for (const key of Reflect.ownKeys(value)) {
    copy[key] = deepClone(value[key], seen);
  }
  return copy;
}

What they check: the WeakMap is what handles circular references. Dates and regexes are the cases people forget. If you have time, mention Map, Set and structuredClone, and what each one does not cover.

5. Image carousel (React)

The first UI problem most people get. It looks trivial, and it is where state mistakes show up.

import { useState } from "react";

export default function Carousel({ images }) {
  const [index, setIndex] = useState(0);
  const last = images.length - 1;

  const prev = () => setIndex((i) => (i === 0 ? last : i - 1));
  const next = () => setIndex((i) => (i === last ? 0 : i + 1));

  return (
    <div>
      <img src={images[index]} alt={`Slide ${index + 1} of ${images.length}`} />
      <button onClick={prev} aria-label="Previous slide">Prev</button>
      <button onClick={next} aria-label="Next slide">Next</button>
      <p role="status">{index + 1} / {images.length}</p>
    </div>
  );
}

What they check: one piece of state (the index), wrap-around at both ends, functional state updates, and accessible labels. Add dots and autoplay only after the basics work and you have said you will.

The mistake that costs the round: spending 30 minutes on CSS transitions before the logic is done. Get it working, then make it pretty.

How I'd actually practice

Pick one problem, set a timer for 20 minutes, and write the solution without hints.
Then compare with a clean solution and note what you missed. Edge cases are usually the gap, not the main logic.
Talk while you code. In the real round, narrating your approach is half the score.
Repeat the same problem a week later without looking. That is when it sticks.

Where to practice these

I built Coding Activist partly because I wanted a place to run these in the browser with tests that grade the result instantly. To be clear about what is free and what isn't:

Free to solve: Image Carousel, Debounce, Throttle, Deep Clone and Flatten Array. That is the five above, 5 of 94 problems.
Paid: the rest of the machine coding set (React components, hooks, DOM widgets, async utilities) is part of the paid course. If you only want the free ones, the five above are a solid start and you do not need to pay for anything to follow this article.

If you would rather not use my site at all, everything in this post works in any editor, and the solutions above are yours to use.

What was the machine coding problem you got asked that you did not see coming? I'd like to hear it in the comments.