Optimized 💃

If you use the Nuxt 2 framework, be careful with plugins as they may cause significant memory leaks.

Key points:
1. Avoid using Vue.use() or Vue.component() inside the exported function, as it creates a new instance.
2. Choose the right plugins for SSR; not every plugin is needed during the SSR stage.
3. Remove custom events and timers on unmount.

These changes make the text a bit clearer and ensure it reads smoothly.

@islombe_dev

В Nature вышла интересная статья (pdf), в ней рассматрели «человеческий язык» как инструмент для передачи информации и пришли к выводу, что это клевый способ для коллаборативной работы нас как вида (и шитпоста в интернете), но язык не нужен мозгу для «мыслительного процесса»

Я никогда не задумывался, как выглядит «мысль» если исключить из нее «язык», поэтому вот пару интересных наблюдений из статьи о том как устроен процесс мышления:

1. Мысли, вероятно, опираются на абстрактные, неязыковые ментальные представления. Это могут быть визуальные, пространственные или другие специфические представления под конкретные задачи и концепции. Как человек, который в любой момент времени может вспомнить любой маршрут по которому я ходил где-то и был трезвый, могу сказать что все так, у меня «карта местности» просто существует в виде образа в голове, как оказалось, так не у всех

2. В голове есть ансамбль нейронок — в статье упоминается несколько «мозговых сетей», участвующих в различных задачах:

— Есть нейронка «множественных требований» для целенаправленного поведения и решения новых задач, она поощряет находчивость при решении проблем, активируется при выполнении когнитивно сложных задач, особенно тех, которые требуют гибкого мышления и адаптации к новым ситуациям. Она же отвечает за планирование и рабочую память;

— Есть нейронка «теории разума» или «эмпатии»: для социального рассуждения и понимания психических состояний – мыслей, чувств, убеждений и намерений других людей. Это та, которая у нас уже хуже чем в GPT4o;

— Есть «Дефолтная нейронка», активируется когда мозг в покое – связана с кучей когнитивных штук, включая: способность мысленно перемещаться во времени (вспоминая прошлое или представляя будущее), пространственное познание, размышление о себе, своих чертах и опыте. Интересно, что повреждение это "сетки" ведет к депрессии или даже шизофрении (в теории).

3. Символическое мышление может происходить без языка как инструмента. Это означает, что мозг может манипулировать абстрактными символами или концепциями, не опираясь на слова или грамматические структуры. Условно, когда вам что-то «пришло в голову из неоткуда», это этот эффект.

4. Рассуждение по специфичным задачам — разные типы мышления (математические, социальные, пространственные и т.д.), по-видимому, задействуют разные мозговые «нейронки», что предполагает, что процессы мышления могут быть в некоторой степени модульными и специфичными для конкретных задач; это как если бы у одной мощной LLM было много Lora натренированных на конкретных задачах датасета, а не единая модель для решения всех задач.

5. Параллельная обработка — все эти «нейронки» для решения специфичных задач (социальные, причинно-следственные, решение проблем и т.д.) могут работать параллельно, получая при этом информацию с разных «сенсоров» тела, а не определяться единой системой, такой как язык.

6. Мы сильно похожи на животных — в статье упоминается непрерывность человеческой эволюции, предполагая, что наши процессы мышления могут быть более похожими на процессы мышления других животных, чем считалось ранее, просто они оказались более сложными из-за наших расширенных когнитивных способностей. То есть, условная собака может оказаться намного ближе к нам по процессу мышления чем мы раньше думали.

7. Культура развивает мышление — хоть это и не прямой механизм мышления, в статье упоминается что передача культурных знаний от поколения в поколение, повышает когнитивные способности человека с течением времени, это уже влияние языка как инструмента.


Короче, все еще ничего не понятно — но на всякий случай продолжайте тренировать лоры ☕️

Hey everyone! It’s been a while since my last update. Life’s been a whirlwind, and I haven’t had much time to be a content creator 🤪

Recently, I made a presentation about streams in browsers, and I want to share some of those ideas with you here.
Multithreading(web workers) in web projects is a game-changer, yet I rarely see it utilized. It can significantly lighten the main thread’s load and make web applications more dynamic and responsive. So, where and when can it be used?

1. Complex Calculations 🖥
Heavy computations often block the main thread, causing the page to freeze or even crash. Offloading these tasks to web workers can prevent this. For example, if you’re processing large datasets to create a graph, you can delegate the calculations to a worker. The worker processes the data and returns only the (x,y) coordinates, which you can then easily plot.

2. Real-Time Analytics 📊
Perfect for traders! Imagine writing a custom plugin that monitors the market in real-time, providing notifications or triggering actions when certain conditions are met. Using web workers ensures that the main thread remains free for other tasks.

3. Image Processing 📷
Tasks like applying filters, cropping, and other image manipulations can be offloaded to web workers, making the main thread available for smoother interactions.

4. Inter-Tab Communication 💬
Need to synchronize actions across multiple tabs? For instance, logging into one tab could automatically trigger an “authorized” status event in other tabs. Web workers can help manage these tasks efficiently.

These are just a few examples of how web workers can make your web projects faster, smarter, and more efficient.

However, there are some drawbacks, like the lack of direct access to the Browser API and the need to use an event-driven pattern. But the benefits far outweigh the challenges!

I hope this sparks some ideas for your own projects!

Whats happening with Windows?

Thanks God internet runs on Linux 🐧

Random info about our Earth 🌍

All of Earth’s water, if gathered into a single sphere, would have a diameter of about 1,386 km! This sphere includes all oceans, ice caps, rivers, groundwater, and even the water in the atmosphere and living organisms. Despite the vast oceans, Earth’s water is just a thin layer compared to the planet’s size.

Learn more about the fascinating distribution of our planet’s water here: USGS - All Earth’s Water in a Single Sphere.

I gave an interview to our marketing team about my experience 😎

Check it out here!

A Genealogy of Technology and Power Since 1500 – collection of main inventions, from the printing press to the chatgpt.

Get hooked on high-quality content right here. 🤔

Why is the Developer Doing Everything? 🤔

Ever feel like you’re the team’s Swiss Army knife, doing everything from coding to setting up the coffee machine?

1. Developer = Jack of All Trades
We code, we test, we even do a bit of design! With a skill set like ours, it’s no wonder tasks keep landing on our plate. 😓

2. Small Team, Big Expectations
In startups, there’s no budget for specialists, so who’s going to handle it all? Yep, you guessed it—us. 💪

3. “Just Give It to the Dev” Syndrome
Without clear roles, everything from bug fixes to Friday night pizza orders ends up being a developer’s responsibility. 🍕

4. Communication Struggles
Sometimes it feels like “the dev will figure it out” is the team’s unofficial motto. 😅

5. Expectations Galore
When you’re a key player, it’s like there’s a neon sign over your head saying “Dump Tasks Here.” 🛑

#DeveloperLife #TechHumor

The GPT-4 model has now been marked as the 'Leagcy model' 🤯

A new 'o1' model has been launched, but it doesn't yet support web browsing, file handling, or image uploads.

Whom it’s for

...o1 can be used by developers in all fields to build and execute multi-step workflows. 👨‍💻

Sam Altman wrote a post about the future of AI — here are the key points:

1. In the next few years 🔜
- AI will become a personal assistant, handling specific tasks like coordinating medical care.
- Virtual tutors will provide personalized learning in any subject and any language.
- Healthcare will improve with AI advancements.
- We’ll be able to create any kind of software on demand.
- Changes in the job market will happen more slowly than people expect.

2. In the next 5 years 😑
- A breakthrough in creating superintelligence (AGI) could happen.
- AI will have a significant impact on the job market, but new forms of work will emerge, even if they don’t look like traditional jobs.

3. Mid-term (a few decades) ✍️
- Personal teams of AI experts will help solve almost any task.
- A rise in global prosperity and a fairer distribution of wealth will become possible.

4. Long-term ⌛️
- AI will accelerate scientific progress and help develop the next generation of AI systems.
- Solutions to global issues like climate change and space colonization are on the horizon.
- We could unlock all the laws of physics and gain access to limitless energy.
- Society will shift toward positive-sum games, with new forms of work that seem unimaginable today.

Big changes are coming, along with challenges, but the future looks exciting! Adapt or be replaced by AI 😐

You've thought you can hide from the Big Brother? Escaping from Matrix getting harder and harder

From Mr. Durov's post

Since 2018, Telegram has been able to disclose IP addresses/phone numbers of criminals to authorities, according to our Privacy Policy in most countries.

🐶BTW, Telegram encryption works only in secret chats, but since they have your IP and phone number, there’s more at play behind the scenes. While secret chats boast end-to-end encryption, standard conversations are stored on Telegram’s servers, accessible by the platform itself. This raises concerns about how secure your data truly is, especially when combined with metadata like your IP address and phone number.

Who has access to this information, and what are they doing with it?
Could these seemingly innocuous details be the key to uncovering deeper insights into your identity or tracking your movements in ways you’re unaware of?

However, you can relax until you are not interesting to the Gov. 🙂

Make a note of that 🤓

Task
You're given an array of integers a and two integers x and y. Count the number of elements in the array such that `x ≤ a[i] ≤ y, where i is the 0-based index of the element.

Seems easy, but:
code limit: less than 48 characters.

This is like an average business task: the task gets solved, but unnecessary stuff always seems to get in the way. 😂

Don't Overthink, Start Solving! 💡

When faced with a complex task, it's easy to overthink and get lost in endless solutions. But here's the key: just start solving!

Once you get the correct result, it's time to optimize.
Bit by bit, refine your code—make it more readable, maintainable, and sometimes even faster (but not always 😉).

Here is an example:

Task is

Implement a function that receives two IPv4 addresses, and returns the number of addresses between them (including the first one, excluding the last one).

All inputs will be valid IPv4 addresses in the form of strings. The last address will always be greater than the first one.

Solution optimizing evolution

First version, which works, but its a big unreadable code

// first version
export function ipsBetweenV1(start: string, end: string): number {
  const reversedStart = start
    .split(".")
    .reverse()
    .map((segment) => parseInt(segment, 10)) as number[]
  const reversedEnd = end
    .split(".")
    .reverse()
    .map((segment) => parseInt(segment, 10)) as number[]

  const amountInStart = reversedStart.reduce((acc, curr, index) => {
    acc += curr * Math.pow(256, index) + 1
    return acc
  })
  const amountInEnd = reversedEnd.reduce((acc, curr, index) => {
    acc += curr * Math.pow(256, index) + 1
    return acc
  })

  return amountInEnd - amountInStart
}

Nasty...

Implementing DRY and Divide-and-conquer:

// second version
export function ipsBetweenV2(start: string, end: string): number {
  const getReversedSegments = (ip: string) => {
    return ip
      .split(".")
      .reverse()
      .map((segment) => parseInt(segment, 10)) as number[]
  }
  const calcAmountInIps = (ips: string) => {
    return getReversedSegments(ips).reduce((acc, curr, index) => {
      acc += curr * Math.pow(256, index) + 1
      return acc
    }) as number
  }

  return calcAmountInIps(end) - calcAmountInIps(start)
}

Looks much better. But something is missing, and looks like we overdoing something...

Removing unneeded reverse(), and simplifying functions into one with readable structure:

export function ipsBetween(start: string, end: string): number {
  const ipToNumber = (ip: string): number =>
    ip.split(".").reduce((acc, curr, index) => acc * 256 + parseInt(curr, 10), 0);

  return ipToNumber(end) - ipToNumber(start);
}

Final solution looks 10 times better, much more readable and well structured 🗣

#CodingTips #ProblemSolving #DeveloperLife

The US has a special commission established in 2000, known as the United States–China Economic and Security Review Commission. The commission consists of 12 members and reports exclusively to Congress. It is not part of any agency or department. Every year, before December 1, the commission publishes a report that, among other things, makes recommendations to Congress.

And the first thing in the new recommendation:

- Congress establish and fund a Manhattan Project-like program
dedicated to racing to and acquiring an Artificial General Intelligence (AGI) capability.

😬 😬 😬


#news #ai