The global banking system is experiencing a profound transformation. Traditional banking infrastructure is rapidly losing ground to the ubiquity of mobile phones and the rise of agent-based and telco-led financial services. The numbers reveal a structural shift that is bypassing banks—especially in rural and low-income markets—at an unprecedented scale.

ATMs vs. Mobile Phones: The Scale of Displacement

There are just 2.95 million ATMs worldwide, compared to 8.3 billion active mobile phones (BankMyCell, 2025). This means that for every ATM, there are about 2,820 cellphones. In rural areas, which account for 43% of the world’s population (World Bank, SP.RUR.TOTL.ZS), the ATM is increasingly irrelevant. The future of financial transactions is mobile, and the numbers confirm it.

Deposit Outflows and Digital Growth

In 2024, U.S. banks recorded their first deposit decline since 1995—a 4.8% drop (FinTech Weekly). Meanwhile, digital-first banks such as N26 reported revenue growth of 40% in the same period. Over 55% of U.S. customers now primarily manage their finances through mobile apps, and the global digital banking market is projected to reach $22.3 trillion by 2026 (EMB Global). Challenger banks and fintechs are not merely supplementing the market—they are capturing its core.

Shrinking Physical Presence

Major banks are closing branches at an unprecedented rate. NatWest closed 53 branches in 2025 alone, while HSBC’s digital transition has been described as “sluggish” (Finextra). The contraction of physical banking infrastructure is a global phenomenon.

Fintechs and Challenger Banks: Rapid Customer Acquisition

Trust Bank in Singapore became the country’s fourth-largest retail bank in just over a year, surpassing one million customers (The Straits Times). JPMorgan Chase’s UK digital platform reached 1.6 million customers in 2023 (Reuters). The revenue gap is widening: while legacy banks struggle for marginal deposit growth, digital-first banks are expanding at rates of 40% or more.

Micro-Transactions: The Global Bypass

The migration of micro-transactions away from traditional banks is most visible in rural and low-income markets, where agent networks and mobile money services—often run by telcos—have become the default. The Dominican Republic provides a concrete example of the scale of this bypass, even in the absence of a local fintech ecosystem.

• Population (mid-2024 est.): 11.5 million
• Rural share (official data): ≈ 85%
• Rural population using only feature phones (field surveys): ≈ 70%
• Active mobile phones (all types, global): 8.3 billion (BankMyCell, 2025)
• Feature (“button”) phones still in use (global): ≈ 1.1 billion (13%)
• Average fee for a sub-US $5 bank transfer in LICs: US $0.50–0.90 (10–18%) (MoneyTransfers)

Assuming each rural adult in the Dominican Republic makes one US $5 peer-to-peer transfer per day using a mobile wallet or agent network (with no bank involvement), and that a traditional domestic bank transfer of the same size would cost approximately US $0.60 in fees (mid-point of the 50 DOP Caribe-Express tariff):

• Rural users: 11.5 million × 85% ≈ 9.8 million
• Annual transfer flow: 9.8 million × $5 × 365 ≈ US $17.8 billion
• Forgone fee income: 9.8 million × $0.60 × 365 ≈ US $2.1 billion per year

Result: In this scenario, banks in the Dominican Republic forfeit approximately US $2 billion in fee revenue annually on micro-payments alone, while US $17–18 billion in transaction volume bypasses the formal banking rails. The same pattern is observable in many other countries with high rural populations and widespread mobile phone use.

Global Perspective: The Scale of Bypassed Banking

Extrapolating this model globally:

• Global rural population: ≈ 3.4 billion (World Bank)
• Assume 30% of those adults (≈ 1.0 billion) have access to a feature phone and mobile wallet/agent.
• Each makes one US $5 transfer per day.
• Average legacy-bank fee if using the formal system: 10% (US $0.50).
• Daily volume bypassing banks: 1.02 billion × $5 ≈ US $5.1 billion
• Annual volume bypassing banks: US $1.86 trillion
• Annual fee revenue forfeited: 10% × 1.86 trillion ≈ US $186 billion

Even if only one in four rural residents transacts daily and the average fee is halved, the annual loss remains between US $40–90 billion.

Structural Barriers to Reversal

• Physical reach: ATMs number just 2.95 million worldwide—a 2,800:1 ratio versus mobile phones. Branch networks are shrinking by 3–7% annually in most low-income countries.
• Cost structure: An on-us rural ATM transaction costs a bank approximately US $0.25–0.35 to service. Agent/mobile-money networks operate at less than US $0.08 per transaction, often subsidized by telcos.
• Device reality: Over a billion users have no smartphone; USSD/SMS wallets are their only digital finance channel. Banks rarely support USSD directly, while agent networks and telcos do.
• Behavioural lock-in: Once users trust a mobile wallet or agent for daily micro-transactions, larger remittances and savings migrate there as well, accelerating deposit leakage (already −4.8% in the U.S. legacy sector in 2024).

Lessons from M-Pesa: Missed Opportunities and Global Impact

A similar model has already demonstrated its transformative power elsewhere. In Kenya, M-Pesa—a mobile money service launched by Safaricom and Vodafone—now processes over $300 billion in transactions annually, with 51 million users and 59% of Kenya’s GDP flowing through its rails (Forbes, 2024). M-Pesa’s success was not the result of a bank-led initiative, but of a telco’s willingness to serve the unbanked and underbanked with simple, accessible technology.

Notably, Visa did not acquire M-Pesa. Instead, in 2022, M-Pesa and Visa partnered to launch the M-Pesa GlobalPay virtual Visa card, enabling M-Pesa users to make international online payments (Safaricom). The ownership of M-Pesa remains with Safaricom and Vodacom, who acquired the brand from Vodafone in 2020.

In 2013, a project with similar ambitions was proposed to banks in the Dominican Republic, aiming to bring agent-based, mobile-enabled financial services to rural populations. The proposal was dismissed as “excess service.” In contrast, M-Pesa’s model has since generated billions in revenue and transformed financial inclusion in multiple countries.

Conclusion: A Structural Shift, Not a Passing Trend

In countries with high rural populations and limited banking infrastructure, billions in annual fee income are bypassed on micro-transactions, with trillions in transaction volume moving outside the formal banking system. Globally, over $186 billion in micro-transaction revenue is now captured by telco-led or agent-based rails—roughly double the combined 2024 profits of the five largest U.S. banks.

The data is unambiguous: the migration of micro-transactions to mobile and agent networks is not a marginal trend, but a fundamental reordering of the financial landscape. The infrastructure and business models of traditional banks are being outpaced by the scale, efficiency, and reach of mobile-first and agent-based solutions—especially in rural and low-income markets.

Sources:

• BankMyCell: How Many Phones Are in the World? (2025)
• World Bank Rural Population Data (SP.RUR.TOTL.ZS)
• FinTech Weekly
• EMB Global: Digital Banking Market Size
• MoneyTransfers: Caribe Express Fees
• The Straits Times: Trust Bank Hits 1 Million Customers
• Reuters: JPMorgan Chase UK Digital Bank
• Finextra: HSBC Branch Closures
• Forbes: M-Pesa’s Global Impact
• Safaricom: M-Pesa GlobalPay Virtual Visa Card

For further context on the 2013 project offered to Dominican banks, see:

• Project details in Spanish
• Project details in Russian

The numbers, trends, and structural realities point to a new era in financial services—one where mobile and agent-based platforms are not just competing with, but overtaking, traditional banking infrastructure, especially in rural and low-income markets.

In November we wrote about how to “digitize” an apiary and turn it into “bee sharing”. In this article, its author, Artem Zharov, talks about the economics of the project and the change in business model.

https://pchelosharing.ru/
By Artem Zharov Founder and designer of BeeSharing and BeeSharing/Gift

Bees in the hive

Actually, the figure of 40 kopecks per year is not entirely accurate. To begin with, on average, a bee lives for 3-4 weeks, meaning each season sees 4-5 generations of bees working.

The 50,000 bees in a hive is also an average number during the peak cycle of the bee colony’s life. When summer begins, everything is in bloom, and the queen bee produces as many bees as possible.

…

“Only bees are better than honey”

Moreover, not all bees gather honey. Apart from the basic division of bees into the queen, drones (males, needed solely for fertilization), and worker bees, there is also a division based on roles, specifically by age. The bee’s body develops, and along with that development, its functions change. The list is ordered by increasing age:

• Cleaner bees
  .
• Builder bees
  .
• Nurse bees
  .
• Guard bees ‍
  .
• Ventilating bees
  .
• Forager bees (peak of bee development)
  .
• Scout bees
  .
• Water-carrying bees (old age of the bee)

But since all roles are necessary for production, we can calculate on average.

Bees can gather from 30 to 100 kg of honey per season depending on conditions. Obtaining 50+ kg is possible with an industrial approach, collecting nectar from agricultural fields.

This is 108% not our option because, at a minimum, we dislike monofloral honey (gathered from a single crop), and pesticides can easily find their way into honey.

…

Our bees gather around 40 kg, and half of this honey needs to be left for the bees for autumn, winter, and spring. So, we or the hive owners get a yield of 20-23 kg (last season resulted in 21 kg).

Industrial honey

The cost price of industrial honey is 20-40 ₽ per kilogram.

A kilogram of industrial honey in bulk costs 100-250 rubles.

In small wholesale, it goes up to 300 rubles, and at retail directly from the beekeeper, it can be up to 500 rubles.

At fairs, it can reach up to 700 rubles if purchased in 1-2 kg quantities.

In a store, after processing and packaging in a beautiful (or not so beautiful) jar, it can be around 1000-1200 rubles per kilogram.

…

Typical honey packaging is the complete opposite of our vibe.

The cost price of such honey is around 20-40 rubles per kilogram, depending on the arrangement of the apiary, its size, and most importantly, the yield.

All such apiaries use sugar, do not value the life of bee colonies, and do not hesitate to use industrially produced antibiotics, and so on.

…

COG of honey manufacturing.

If a beekeeper sells bee packages and other beekeeping products besides honey, having 500 bee colonies can generate a net income of up to 4 million ₽ per year. Blood money in the honey business.

…

Possible profit in an industrial apiary with 500 bee colonies: 4,000,000 per year.

Homegrown honey and honey from eco-friendly apiaries

Products like these are rarely sold in large wholesale quantities (several tons) because, with minimal sales skills, they can be sold at a good retail price. Also, there are rarely such large apiaries with such an approach.

Average cost price of homegrown honey: 200 ₽ per kilogram.

Summary of indicators and distinctive features:

• Yield: 30 to 50 kg depending on the region.
  .
• More work on the apiary.
  .
• Higher expenses for treatment and prevention.
  .
• No expenses for fuel, as the apiary is stationary.
  .
• Average cost price comes out to around 200 ₽ per kilogram.

…

COGS of eco-friendly apiary.

Apiaries that are now joining BeeSharing sell their honey in small wholesale for approximately 600-650 ₽ per kilogram. This often exceeds or equals their retail price. However, unlike retail, they don’t need to deal with:

• Packaging
  .
• Delivery
  .
• Marketing
  .
• Sales

Moreover, this is an expansion of the apiary without additional investments – hive acquisition is covered by us.

Thus, by selling through BeeSharing, adding a hundred hives results in an increase in the net profit of the existing apiary up to 1.6 million ₽.

…

1.6 million per year. This is an increase in the profit of an eco-apiary of 100 bee families working through bee sharing.

Working independently, without our platform, a beekeeper is forced to handle packaging, delivery, find clients, rent sales points at fairs or markets, advertise on platforms like Avito, maintain a blog…

Even if they sell at double the price, at 1200₽/kg (though discounts are common, especially for purchasing larger quantities like a 3L jar), the net profit comes out to roughly the same 650-700₽. Additionally, instead of staying at home with the kids and relaxing, they have to deal with sales.

…

Our additional values

We have our own apiary, and the figures above are partly based on it.

However, in addition to the apiary, we are developing a platform for beekeepers and a personal hive account for customers. So, in BeeSharing, apart from apiary costs, there are also the following expenses:

Content: It amounts to almost 350k per year, including:

1) A photographer takes photos and videos for photo reports published in the personal account. On average, 2-3 shoots per month.

…

2) A photographer takes photos and videos for photo reports published in the personal account. On average, 2-3 shoots per month.

…

Illustrations and design. It costs 250k per year. We create packaging for honey and related items. In our catalog, there are 50+ design templates, and each hive owner chooses how their yield will look as we approach autumn.

…

Including, we also create custom designs for clients. For example, here is a sketch of a label for the Cinemapicnic apiary of director Maxim Rusanov. The concept was approved on the first attempt:

…

Packaging and dispatch of orders + support — 600K per year. We package honey in St. Petersburg, not in the Krasnodar Krai where the apiary is located. From there, we dispatch parcels to customers using a courier service. Shipments usually reach Moscow within a day, and for St. Petersburg, we often provide same-day delivery.

…
New Year’s gifts for the employees of Pasha Molyanov’s “Let’s Do It” agency

Hosting, services, and such — 50K per year

Office rental — 480K per year. For packaging and for me. I have too many children to work from home

Communication at the apiary — 36K per year and a one-time (hopefully) 100K for equipment.

https://pchelosharing.ru/gift

And also: a discount of 9.99% with the promo code “СКОРО14И23”.

https://pchelosharing.ru/
By Artem Zharov Founder and designer of BeeSharing and BeeSharing/Gift

By Yaroslav Radinsky
CEO at RASA.pro tg – @radinskiy
The text is published with the permission of the author.
Link to original text is here.

I’ll tell you why we hired such an employee and what came out of it. About the battle with headhunters, resourceful pimps, and unconventional team-building.

..

In the photo: Kristina Kukhtina. Photographer – Evgeny Pyankov.

My name is Yaroslav, and I am the CEO of the digital agency RASA based in Khabarovsk/Russia. We specialize in creating websites, applications, configuring CRM, and advertising promotion in the Russian Far East.

How we looked for a Go-go dancer or stunt marketing in IT industry

Recently, on a hiring platform, we posted a job opening for a GO-GO dancer for our agency.

The idea behind hiring an employee for this position was to bring fresh perspectives and approaches to our work, regularly boosting the team’s motivation. The concept is undoubtedly innovative, but we are also a company with a young and dynamic team.

…

Everyone would have benefited from this idea:

For our designers and UX researchers, it would be an unconventional source of inspiration for new solutions. For developers, a wonderful assistant in tackling complex code and a means of relaxation after a challenging workday. One of the requirements for the job was an understanding of code terminology and the ability to visually interpret it.

And for the team as a whole, this person would be a friendly partner, the office mascot, and someone who makes you want to stay at your workplace longer and even invite friends to work – a dream scenario from any manager’s perspective. Moreover, the art of dance itself is an underestimated tool for work, especially in the form of contemporary dance.

…

“I imagined working in IT differently.”

One of the main values of our team is caring for each other. Both programmers and all employees, in general, can get tired or even experience burnout, which inevitably affects work processes and potentially the team spirit. Therefore, relaxation and recreation are no less of a priority for us than the work itself.

Previously, RASA had unconventional specialists visit the office, such as a breathing techniques coach. As a team, we gathered in our relaxation room in the morning, turned off the lights, and practiced proper breathing in complete darkness to set the soul and body for a productive and energetic day, charged with positive energy. We also had a yoga instructor come to help unlock the team’s full potential through practices and, quite simply, strengthen team building.

…

From time to time, as a team, we play poker, board games, and have a regular tradition of movie nights (usually on Friday evenings, but sometimes on Thursday or Wednesday). The team even votes on the movie they’ll watch. We have a ping pong table in the office where you can hit some balls during lunch, although this is hardly surprising to anyone. Also, every Friday at 5:30 PM, we have a meetup where the team shares experiences on productivity, work, and ways to improve their lives inside and outside the office.

In our understanding, it is crucial to see employees not just as workers but as a real team, where human needs are valued alongside their experience and skills. This requires time, organization, and money, but it makes us who we are.

And in our view, having a resident GO-GO dancer is one way to support the established corporate culture. At the beginning of the year, I provided details about what our team does outside of work and why the agency means much more to the team than just a workplace.

…

What keeps young talents in the Far East?
The edge of the world, low salaries, and what is there to find at all—this is roughly what people from the central part of Russia think when it comes to us. Using the example of employees in a digital agency, let’s explore why skilled professionals choose to stay here and not move to larger cities.

But let’s get back to the job opening: after its publication, some time passed, and for some reason, local media and public groups started writing about it. Then, my colleagues and acquaintances noticed it. I received messages in private and via email with content along the following lines.

This is genius!

Is everything okay there?

Should I break into IT?

To be honest, I was a bit surprised by such a resonance: a regular job opening meant to address production tasks suddenly became news and quickly left Khabarovsk, making its way into larger media outlets.

Publications in local media and not only about our vacancy.

I was even invited to the radio, and Moscow-based news portals reached out for interviews to have me share more details. Well, it’s time to share them with you too.

PR campaign

Or, damn it, fake news.
I suppose you’ve already grasped that our job opening is an atypical PR campaign, which, to my delight, worked out as planned.

The idea struck me on December 14th. It didn’t happen spontaneously; I was exploring ways to advertise our agency. It’s always a significant budget and a major hassle for companies like ours. After all, we operate in the performance marketing sphere, where everything is measured in quantitative terms: we attracted so many leads, brought in so many clients, reduced the number of abandoned carts, sold 250 million worth in a month, and so on. And all agencies say the same thing: that they deliver many results for little money.

…

At this point, a thought occurred to me: advertising is like buying love. Today you pay, and everyone wants you, but tomorrow the budget is gone, and you find yourself somewhere behind the scenes.

Then I remembered that besides advertising, there’s PR.
But I’m not familiar with it, and it’s not my strong suit. I’ve never been a journalist, a professional creator, or someone who knows how to create news hooks (or maybe I’m too self-critical). I always thought that the ideas that came to me were too straightforward, not very useful, and not worth trying to unfold. But that evening, I wanted to test a hypothesis that seemed quite unconventional.

News also appeared in Telegram groups.

Of course, I’m talking about the vacancy with the dancer.
I was dealing with the routine: browsing through job listings on HeadHunter and working on recruitment – currently, we’re in search of an HR manager, so I’m handling personnel matters for now. While reviewing profiles of analysts, programmers, and HR professionals, I suddenly felt the urge to see someone extraordinary, a specialist that no other company has.

And why not a dancer?

I immediately envisioned the reaction of those around to such a prank. Here’s an IT company, and they’re looking for a GO-GO specialist, and it’s definitely not about the Go programming language. Any regional media outlet would pay attention to this – and where there’s resonance in regional media, the news quickly spreads further, across the entire internet.

Many might believe that IT professionals have become so spoiled that they don’t know how to indulge themselves anymore (which, in reality, doesn’t resemble the truth anymore, but who needs this dull reality?). Plus, when there’s a component of sex and cheekiness in PR, it only amplifies its effect. The idea seemed good to me, with a budget of only 1400 rubles (the cost of posting the vacancy), and I decided to take the risk.

…

But first, I sought advice from someone who deals with similar creative things – it was interesting to hear an opinion from an outsider’s perspective.

Testing is necessary. But, unfortunately, I didn’t come up with this.

On the other end of the line.

I took a deep breath, posted the job opening, and then sent it to a couple of acquaintances for them to check it out.

A few hours later, I saw a post about the vacancy in a regional community on social media. After that, information started coming in that the link to it was circulating in chats among employees of IT companies, various banks, 2GIS offices, and so on. My inbox quickly filled with messages from many of my colleagues and acquaintances.

…

Some expressed respect for excellent PR, while others didn't quite catch on and wrote that it was a very bold job opening.

In connection with this, quite a few people wanted to join the commission for selecting the new employee.

Responses to the vacancy on HeadHunter spiked sharply. However, 80% were, of course, programmers who wrote that our company is solid, and they wished to join us in the office. Yet, among the responses, there were also very relevant ones, especially from the local community: one girl prepared a resume very well, and it was evident that it was crafted directly in line with the job requirements.

…

From the resume for the position of Go-Go dancer in the IT team: “I inspire people, charge them with joy, I play the crocodile well.”

Moreover, an enterprising individual named Artem (originally from Komsomolsk-on-Amur) contacted me, proposing to regularly supply girls for a percentage. He even sent their profiles with photos and videos – the guy was geared towards establishing serious business relations. I assume he still hasn’t quite grasped the essence of our initiative.

…

Of course, very few channels and media outlets explicitly mentioned the name of our agency. However, the address (Dzerzhinskogo 52) was visible in the vacancy screenshots, and the city was always specified, which contributed to the success of our PR campaign. The screenshots spread across a vast number of chats, channels, and communities, and people started googling to find out more about this company.

Without exaggeration, our website traffic increased approximately 50 times. We started receiving kind and warm letters of respect from literally everywhere, even from CIS countries. Some simply wrote “you guys are awesome,” while others said that we created the perfect festive atmosphere.

Cover letter: “I don’t know what your plan is, but you definitely created a festive atmosphere in the office.”

Cover letter: “Brilliant! We need to put this idea into circulation!”

Cover letter: “With dancing code and fun!”

It wasn’t without its challenges, either.

And problems did arise with the HeadHunter website – our vacancy started getting blocked. When a significant traffic influx hit the job opening (this happened on the 19th, a Tuesday), HH algorithms likely perceived it as a DDoS attack or some bot influx, so they took the vacancy down from publication. No money was deducted in the process.

The problem was not too complicated to solve – I simply clicked the “repost” button, and the vacancy returned to its place. However, on Thursday, the vacancy was blocked again, and the HH support team contacted me – they demanded the complete removal of the job opening from the site, referring to clauses in the agreement stating the inadmissibility of content that could harm the site’s reputation.

…

In other words, you can search for real striptease dancers from strip clubs on HH, but not for an IT company.

I disagreed with HH’s position: we didn’t violate any platform rules, and such a vacancy only benefited everyone in terms of traffic. So, I suggested having a call to settle this misunderstanding. Within half an hour, I received a call from technical support. Not from the PR department, not from the marketing department, but specifically from technical support, where people work strictly according to scripts. I’m not here to speak negatively about the support service, but our dialogue kept hitting the same script: they claimed that the publication could harm the project’s reputation, and I asked to connect me with the PR or marketing departments for further discussion.

When the conversation reached a complete deadlock, they told me that two complaints had been filed against our job opening.

I asked for clarification: two hundred or two thousand?
They replied: just two.

…

So, at that moment, out of approximately 50,000 people who viewed the job opening, two individuals clicked the “report” button, which became one of the main reasons for the blockage. We tried to resolve this issue, but in the pre-New Year rush, with a million other things to handle, it was too cumbersome.

However, in all other aspects, our prank was a success.
The team in the office received a boost of positivity right up to the holidays, and we gained a better understanding of the power of PR.

In the end

We spent 1400 rubles for the publication on HeadHunter, and that was the entire budget.

About 20 million users saw the news about the vacancy (based on rough and superficial calculations of public reach. Unfortunately, we couldn’t count the statistics for corporate chats).

Responses to our other vacancies significantly increased – more than 200 came in. After the New Year, some of these individuals will start working with us once we find an HR manager.

Khabarovsk is once again making waves in creativity and PR – and it’s delightful. Most importantly, many truly believed in the reality of what was happening: both IT professionals, the audience, and even journalists.

…

So, when people say that there is no life, creativity, or creative individuals on the Far East, it sounds at least strange to me. We saw, through a simple example, how interesting the PR tool can be, and I think we’ll implement something new in the near future (I’m not selling PR courses yet, but if anyone is interested, write to me on Telegram

P.S. While writing this article, my colleagues managed to convince me of something, and after the New Year, we will indeed invite a Go-Go dancer to the office. But I’ll write about that next year (and possibly even add a video).

Greetings from frosty Khabarovsk,
Yaroslav Radinskiy, RASA.

…

Happy New Year everyone!

By Yaroslav Radinsky
CEO at RASA.pro tg – @radinskiy
The text is published with the permission of the author.
Link to original text is here.

By Vladislav Vorkel
Product designer at Yandex.
Founder of the service for finding internships: https://t.me/Intern_Sheep.
Author’s blog: https://t.me/designvorkel

My internship search product currently has over 300 paying users, and it generates a monthly income equivalent to the salary of an average IT professional. I dreamed of creating a product that people would use, even if just a couple of them.

Hello, I’m Vlad Workel, a product designer at Yandex and the creator of InternSheep, a community with internships for designers. I’ll share how I really wanted my product and how I made it happen.

…

Roughly, this is how my dog helped in creating the product.

Internship search: where did the dream originated

I am a product designer and have worked in product roles, but I have always been interested in looking beyond the design field. I wanted to experience a product from all perspectives, from the idea, development, and demand to working with my team. Moreover, having my own product means it’s solely my responsibility, my decisions, and only my success or failure.

In simpler terms, it’s like crafting a wooden bow for a child that actually shoots, and then proudly running to show it to the adults.

The first attempt

About 1.5 years ago, I decided to embark on my first project – Gradeup. It’s a test for designers that, upon completion, allows you to assess your skill level and receive a competency map. It is still operational, and you can find it here: gradeup.vercel.app.

I found a developer acquaintance who was just learning front-end, designed the interface, and created a test consisting of competency cards from various companies. Initially, we tested the demand for the MVP test using Google Forms, where the result was a PDF that I personally drew and sent to each test-taker. I initially found the first users through my channel, which had about 500 people at that time.

There were many different hypotheses on how to monetize this test, but we never got around to it. Perhaps it was because there was a lack of understanding of how to do it, or maybe it was because I had already burned out and lost faith. That’s when the idea of gathering internships for designers came to mind, as it is something everyone needs!

The second attempt

Yes, that’s the one: InternSheep. So, the main story will be about testing the hypothesis, how we implemented the service in a couple of days with a developer in a hackathon format, how we attracted the first users, how the internship bot turned into a community, and how we created an educational program.

How not to start

The first week, I spent on absolutely unnecessary research, thinking I had to do it the “right” way. I scripted interviews, found respondents, and asked them about their job search experiences, problems faced, and so on.

The truth is, research is necessary when you don’t understand the target audience or when launching the product costs more than the research itself. In my case, I had a clear understanding of the target audience’s problems because I had personally gone through internships, and everyone around me was job hunting. It was literally part of my daily life. As a result, I abandoned the research and started building the product.

My opinion (it’s just mine, but I like it): in personal projects, it’s better to solve your own problem, then you won’t need extensive research. If your solution works for you, it will likely work for someone else – we are not unique.

…

The simplest hypothesis test:
There is a website where only internships are published. Internships of all cool companies immediately appear on the website after the company posts them on its official website. You will receive a notification. Are you ready to subscribe to the service for 200 rubles?

Yes: 52%
No, I don’t need internships: 18%
No, I’m sorry for the money: 17%
No, I’ll write in the comments why: 1%

From idea to the first revenue in just a few days

So let’s begin! The idea of implementation was to write parsers for career websites of top Russian product companies, we started with 15. Write a bot in which parsers will instantly send internships to users. They decided to sell the service by subscription – 200₽ per month

I assembled Landos in 1 day using nocode. I found several respondents for a decision interview and sold 3 subscriptions to a non-existent product. I just remember how I almost begged for real payment – only the money on the card could confirm the hypothesis in my case. It was a little awkward, but I overcame myself))

So, 600 rubles on the card, 0 developments, 3 people are waiting for the product within 2 weeks.

Let’s start development

I messaged Roman, a developer from Yandex, shared the idea with him, and we immediately started implementation. We needed parsers!

What parsers? Are you stupid or what?

Roman Dubinsky

Roman’s righteous laziness prevented us from making that mistake. Finding a student from Voronezh who browsed 15 websites several times a day turned out to be 100 times easier and cheaper. Later on, we did write parsers for some websites, but that’s not as important now.

…

Those very parsers…

In less than a week, we did the most essential thing: a bot and its admin panel where we uploaded internships.

The bot consisted of…

The bot consisted of:

1. Choice of Specialty: We had 5 options, but in practice, only designers subscribed.
   .
2. Subscription Payment: Done through the external Donate bot directly in Telegram.
   .
3. Each Internship had a link to the vacancy.

The entire flow seemed quite convoluted and makeshift initially, or at least, that’s what we thought at the beginning. In reality, it turned out that almost no one got confused, and people joined us smoothly.

So, the launch!

Actually, when I received the first prepayments, I started warming up the audience in my channel, talking about the service and collecting applications.

I asked all my acquaintances with channels to mention the service, and many responded absolutely for free. Thank you, if you’re reading this, it might not have worked out without you.

In the end, before the launch, we had over 300 applications.

As a result, we decided to sell it for 290 rubles to everyone who left a request, and then raise the price to 390 rubles.

We announced a broadcast where I presented the product and talked in more detail, but we started registering applications and opening access a few days before the launch.

…

The very broadcast where I tell you why you need to pay me 290₽.

At the time of the launch, we had sold about 200 subscriptions.

Couldn’t believe it when my card had not just 600₽ but 60,000₽. The amount may be modest, but it’s not about the money. I did something useful, and I’m getting paid for it! Wow!

Creating a community

Initially, we didn’t plan on doing anything like that. The idea was just to create a bot with internships – that’s it. But users suggested that it would be cool to have a chat where people in similar life situations could connect, receive support, and exchange experiences.

Both we and the participants saw significant value in this. Over time, it evolved into a community where we engage in various activities, communicate, and build friendships. People come to the product for internships but stay for the community. Many have found jobs or internships but continue to be a part of our community.

I was building a product and found friends.

What’s happening in the community now?

Continuous communication – and this is the most important aspect. People share test results, discuss who received invitations and who didn’t, seek advice, and find support when things get tough. Natasha, the community manager, and I have established a culture of zero toxicity and maximum friendliness. Throughout our time, we can count on one hand the instances of toxic conversations.

Lead designers conduct reviews several times a month where anyone can submit their portfolio or test work and receive a quality analysis.

We regularly host online meetups in our Safe Space, where you can come and simply talk when facing challenges at work or when despair sets in.

There’s also a book club and a poster art circle. Together, we read design-related books and discuss them, and in the art circle, we create posters with a designer from Yandex.Eats.

We are creating a small, warm, and friendly world where one feels inspired to grow and develop.

Periodically, members of our community would share internships that we didn’t initially have information about. We started keeping track of new companies, and as a result, the count increased from 16 to 26.

The product after 9 months

We haven’t closed down.
We are working!

Currently, the InternSheep channel has 2500 subscribers, with around 300 paying users in the product. Over 30 individuals have secured internships in top companies.

Throughout this period, I’ve hardly invested in advertising. There were trial sowings that yielded no results. Our growth is solely organic, which is great, but it’s time to learn how to attract more.

The team consists of three part-time individuals, including myself – a community manager/SMM, a designer, and me. Occasionally, we bring in others for one-time tasks or consultations.

What are our metrics?

As in any subscription-based service, the primary metric is retention.

We retrieve data from the Donate bot, which tracks all charges, subscriptions, and cancellations. We create charts that illustrate retention across monthly cohorts.

Currently, our average conversion to payment in the second month is 75%. I’m not sure how good of a metric that is, but it certainly seems quite favorable.

…

Retention from February to October.

I won’t talk about the number of paying users. It’s obviously important for us.

The second key metric is monthly revenue. From the start, we increased the cost twice – from 290 to 390, and then to 490. Subscription is not the only monetization model in the service – we sold partner products and channel advertising. Currently, we are developing our educational program and assisting companies in hiring talented interns, but I’ll write about that in the next article.

Let’s summarize

I dreamt of a product, and now I dream that it continues to live and grow. It’s a wonderful platform for experiments, for personal and professional growth. The most valuable thing I gained in the 11 months of working on the product is experience and passion.

Currently, the product brings in a monthly income comparable to that of an average IT specialist. A worthy reward.

Thank you for reading my story – perhaps it will inspire you to embark on your own pet project, or maybe it will just brighten your day a bit.

I would appreciate it if you subscribe to my Telegram channel (5000+) – I share stories about my projects, my work at Yandex, and the journey of self-discovery. I also throw in some humor occasionally.
The channel “InternSheep“: the product I talked about in this article.

By Vladislav Vorkel
Product designer at Yandex.
Founder of the service for finding internships: https://t.me/Intern_Sheep.
Author’s blog: https://t.me/designvorkel

Finder.work has collected unusual solutions from around the world.

…

Sima Land office space

The office of the Russian online store Sima-Land features golden plasterwork, large chandeliers, paintings in ornate frames, and intricately painted ceilings. What do you think?

…

Inventionland

A design studio from Pittsburgh is experimenting with its 6,500 square meter office space. The decor includes pirate ships, race tracks, artificial caves, castles, and much more. Instead of being ordinary employees, Inventionland staff refer to themselves as ‘creativists’ and wear lab coats.

…

Pons and Huot

The architectural firm Christian Pottgiesser has designed an office for two French companies, Pons and Huot. The space resembles a greenhouse and accommodates no more than 20 people. Each workstation is separated from its surroundings by its own transparent dome!

…

Selgas Cano

The Spanish firm Selgas Cano created its office with a simple desire to work close to nature. The space is located in a forest near Madrid. Ventilation is regulated using a pulley mechanism.

…

Palotta Teamworks

The workspace, named ‘Apostrophe,’ is situated inside a large warehouse with containers.

…

Pionen — White Mountain

The office of a Swedish internet provider is located in a Cold War-era bomb shelter. The space is situated 30 meters below the granite cliffs of Vita Berg Park in Stockholm.

…

Nykredit

The meeting rooms of a Danish bank are suspended above the atrium of a business center.

…

By Finder.work
https://finder.work – a service for finding vacancies and employees for remote work.
Our telegram channel https://t.me/theyseeku

My eye twitched when I tried to find an apartment for my parents. No longer:
I have launched the HomeBro website, which searches for apartments on all the top real estate listing sites. All property offers on one map, plus notifications sent to Telegram.

Hi! My name is Lev. Four years ago, I created the HomeBro bot that searches for housing for free and sends you options directly to Telegram. I’ve rented an apartment in Moscow many times, and I desperately wanted to simplify this process. As a result, half a million people have already used the bot.

…

This time, I was helping my parents buy an apartment. And here I felt firsthand how our bot is not sufficient for the process of buying a property.

It was both amusing and bitter that as the founder of a housing search service, with access to a database of all property listings, I couldn’t conveniently explore it myself.

I wanted something simple. To not only receive new options in the bot but also to open a map and see everything available on the market. In the end, once again, like four years ago, I find myself opening websites one after another. I see the same thing 10 times in an attempt to find something new. In short, it’s a mess.

• I looked at 17 advertisements about this apartment.
  
  
• Did you find it without commission?

…

HomeBro: And then, with the team, we created a real estate classifieds aggregator website.

Welcome to the HomeBro portal! Here, you’ll find all the latest listings from Cian, Domclick, Yandex.Real Estate, M2, Avito, and Samolet.Plus. And the list will continue to grow.

Currently, the site has over 1 million active listings, with 50 thousand new ones appearing every day.

You can already visit and search for housing for rent or purchase in 18 cities. It covers all Russian million-plus cities, including Sochi and Arkhangelsk – where one of our developers lives =)

And all this without duplicates. Well, almost =D Each housing option is shown only once.

…

It’s interesting, will they provide a preferential IT mortgage for rental properties?

Also, our trained neural network (I wrote about it here) helps find an apartment without grandma’s renovation, closer to the park, or farther from the roads. Well, or you can deliberately look for a fixer-upper to buy cheaper and renovate.

Setting up notifications for all new listings can be done in two clicks (literally) by transitioning from the website to our bot.

Yes, did you make something really cool?

Hey, folks, the thing is working, but this is the very first version, commonly known as MVP.

We, with a very small team, have been building the portal since the beginning of the year; there are still some glitches. But there’s no point in hiding from people a service that, even in its current form, can bring real benefits.

…

A big thank you to everyone supporting us in our crowdfunding campaign; every ruble of yours is important to us. By the way, we almost didn’t promote it—mentioned it only in one article and did a one-time mailing in the bot. It triggered me when someone said, “You’re begging for money on VC.” As a result, we stopped writing about it. But I was wrong; haters gonna hate.

Welcome to all the found bugs. Visit the HomeBro website and don’t hesitate to say what you felt was lacking. I’ll be incredibly grateful for your comments.

By Lev Klebanov
I am developing a housing search service https://homebro.ru

The text is published with the permission of the author.
The original text is here.

By Rostislav Dugin
TG: @rostislav_dugin
https://click-chat.ru/
The text is published with the permission of the author.
The original is here.

The article about how I almost missed the “boom” of ChatGPT but gained access to it and made money (and continue to earn). The project itself is currently declining, but I’ll discuss that at the end.

…

Table of Contents:
– Who am I
– How the idea came about
– How the project was made from the technical side
– How I attracted my first clients
– Insight: cheating behavioral factors
– How the project grew
– Insight about subscriptions
– What were the problems?
– How Yandex killed SEO
– What are your plans next?
– Conclusions

Who am I

My name is Rostislav, and I am a Full-Stack developer. For those reading my content for the first time, here are links to other articles:

1. “5 Tips for Developers Before Starting Their Startup. Experience After 750,000 Rubles”
   .
2. “Solo Startup. Part 5. How Much Does the Project Bring in Now?”
   .
3. “Solo Startup. Part 3: Simplifying the Product”

Now my main project is a Telegram chat for the site (which additionally shows what city the user is writing from and when he is online).

If it is relevant to you or you want to connect to the affiliate program: ClickChat.

…

How did the idea come about?

At the end of 2022, ChatGPT gained popularity. Everyone was talking, writing, sharing stories, and creating memes about it. I successfully ignored this moment, letting it pass me by.

In February, I stumbled upon a free ChatGPT bot on Telegram. I sent it a few messages, played around with it, and then forgot about it. Some time later, I realized that it could have been an excellent time to launch a project with ChatGPT.

Closer to mid-March, it struck me: ChatGPT has an API! Consequently, an idea emerged: what if I localize the API and resell it to legal entities? They would pay me in rubles, and I would purchase from OpenAI in dollars, pocketing the difference.

A couple of days later, I simplified the concept. Selling to B2B is complicated. There’s an easier way! I’ll sell directly to end-users. I just need to create a website, set up advertising, boost SEO, and the SaaS is ready. Subscriptions can be sold. I’ll focus mainly on SEO (spoiler: it worked).

And so, ChatGPT Me website was born. I deliberately chose a domain with “ChatGPT” to enhance SEO visibility for that search term.

…

How the project was made from the technical side

From a technical perspective, everything is relatively straightforward: when a user makes a request in the chat, I forward the request to the OpenAI ChatGPT API and display the response to the user.

For subscription payments, users are allocated tokens that ChatGPT consumes. I use a tokenizer to analyze token consumption.

Revenue is derived from two components:

1. Unused tokens at the end of the month.
2. My percentage on top of the tokens I provide to the user.

The server-side is developed in NodeJS + NestJS. The main part of the website is built with NextJS, and the landing page is created using Svelte (I experimented to achieve maximum Google Page Speed). The PostgreSQL database is used, and Redis serves for caching and as a message broker. Nginx acts as the HTTP proxy server.

I developed the first version in about a month, albeit at a relatively slow pace due to limited time.

…

How did I attract my first clients?

I initiated the promotion with the following actions:

1. Purchased links through SeoWizard and posted in various software catalogs.
   .
2. Launched advertising on Yandex.Direct (what surprised me was that at that time, few were advertising under the ChatGPT query).
   .
3. Wrote several articles on vc.ru, pikabu, and Habr. However, most of them were either deleted or downvoted almost immediately (fairly, but traffic is traffic).

I had notifications set up on Telegram for every new registration. I wasn’t particularly expecting a sudden success and didn’t realize at the time that this might not be the best solution. Overnight, the chat with notifications was flooded with messages, and the next day, my phone was buzzing every 20-30 minutes.

It looked like this:

Messages about registrations

Many people were registering, but the payments were minimal. I had to disable registration notifications and keep only payment notifications:

Very pleasant notification sound

Overall, the launch turned out to be quite successful. In the first month of the project’s existence, I invested around 30,000 rubles and earned roughly the same amount. Breaking even is already a success!

For a moment, I even felt like a successful entrepreneur.

Insight: behavioural factors

From the very beginning, I understood the importance of allowing users to try the product. That’s why 10 free messages were available in the chat.

At some point, I analyzed and realized that students were registering massively just for the sake of 10 free messages. Here’s an example from this month; take a look at the emails (I don’t verify them through link confirmation):

…

Sign-ups for free messages

And these sign-ups… darn well boost behavioral factors for search engines! I gave a little freebie to the students, and in return, they gave me positions in SEO. It turns out to be a win-win situation.

How the project grew

After a few weeks, the site shot up in the SEO rankings for the query “ChatGPT,” and I decided to stop advertising. I only invested in purchasing links and gradually improved the chat.

Here’s the registration graph for all months from April:

…

May shows a sharp increase in sign-ups due to SEO

All this time I treated the project as a “temporary topic”, so I did not expect to earn millions. I just rejoiced at every peak and waited for all this happiness to end.

In early May and mid-July, the project was pushed to the top positions. These days, the number of subscriptions has been growing in some abnormal amounts. The notification had been on silent for a long time, and the chat looked like this:

…

Messages for July

On good days, the chat brought in 5,000 to 7,500 rubles. About 70% of the revenue was clear profit! I was really hoping that this would last forever (spoiler: it didn’t).

Insight about subscriptions

At some point, I noticed that registrations and subscriptions were increasing, but my payment to OpenAI remained consistently around 10,000 rubles. So, I decided to analyze how many people were actually using their tokens, even if only halfway.

It turned out that only 2.5% of users had ever spent 50% of their available monthly token limit. Despite peak subscriptions exceeding 90,000 rubles per month.

So, roughly 90% of people were essentially giving me money without fully utilizing what they had purchased. This was an incredible insight for me, and it dawned on me the entire benefit of SaaS services.

Now, imagine how much gyms make with monthly memberships or services like Yandex.Plus (although I do use Yandex.Plus).

Moral of the story: making money from subscriptions is the easiest. Subscribing to a small amount is psychologically easy, but over 5-10 months, it accumulates into a substantial sum almost unnoticed.

…

What were the problems?

And now a bit about the challenges. In my opinion, this project brought in easy money, but it’s important to mention the moments where I messed up.

1. Dropped the User Database (epic fail!):
   .
   Registration through the website is done via email. In the beginning, I didn’t validate fields much, and it was possible to enter two identical emails with different capitalisation.
   .
   For example, two different emails for the system: “rostislav@mail.ru” and “Rostislav@mail.ru”.
   .
   At some point, a user couldn’t log into the right account because she used a capital letter in the email. I decided to fix the email registration case by converting everything to lowercase. However, this user specifically asked me to manually change her email. So, I went into the database to update the email manually using an SQL script. I first wrote a WHERE script to check for only one email. Then, for the same WHERE, I wrote the UPDATE script.
   .
   Apparently, I forgot to close a bracket somewhere. I ran it. Within a second, all ~25,000 emails became NULL. At that moment, I felt a cold sweat, and I froze for two minutes. But the backups were in place, and I recovered. I went to change into a dry T-shirt.
   
   – Lesson 1: Even experience and precautions don’t always save you from childish mistakes.
   .
   – Lesson 2: Make backups.
   .
2. Publication on RuStore:
   .
   At some point, I decided to create a mobile application to appear in RuStore under the “ChatGPT” query. Since there were no such applications, the idea had a chance of success and could provide an additional source of traffic. I quickly learned React Native, hastily made the application, went to publish it, and… RuStore didn’t allow the use of the name ChatGPT.
   .
   Fair, but sad. I published it under a different name. After a week, I deleted the application because there were very few registrations, and the application was an eyesore.
   .
3. Servers in Moscow:
   .
   Initially, the API worked fine for all regions. At some point, OpenAI decided to block access from Russia, including for the API. I had to urgently move to the Netherlands (on new servers, I mean). Due to the domain redirection, there was a day of downtime. This situation should have been anticipated in advance.

How Yandex killed SEO

In September, I expected a rapid project growth because schools and universities were reopening. My target audience once again needed access to boundless information! “I’ll buy myself a car with this money!” — I thought at some point. But on August 30th, I received this notification in Yandex Webmaster:

…

You are “mimicking”.
“Violations and security threats on the website http://chatgpt-me.ru, its individual sections or pages.”…

At first, I didn’t understand what it was. Then, I Googled it and found out that “mimicry” is when a website tries to impersonate another popular site.

I’m not pretending to be ChatGPT (nor do I look like it), I’m just using their name almost honestly for SEO promotion. Odd…

I contacted Yandex support to figure out what was wrong. However, their support didn’t provide much help. They simply said it’s all search algorithms, and I should figure it out myself. “If you fix everything, your site will return to search results in 30 days.”

Thirty days without appearing in search results is almost a death sentence for the project, leading to a shift of all positions to competitors (and by September, there were already plenty of such chat projects). In general, the situation emerged where the site was knocked out of the search results, and no one could clearly say what needed fixing.

I tried various methods, like rewriting the text, removing the word “ChatGPT” from the title, and so on. In the end, nothing helped.

From this moment, the project’s gradual decline began.

Since I treated the project as a “temporary thing” and expected something like this, I was a bit sad for a couple of days and then calmed down. Perhaps Yandex did the right thing (after all, such algorithms are not invented for nothing). Moreover, Yandex sometimes acts faster than Google, and all its services (almost) are pretty cool.

The only thing I would have liked in this situation is a clear explanation of what exactly was wrong and assistance in fixing the site, not “wait 30 days and figure it out yourself through trial and error.”

After all, if a genuinely serious site ends up mistakenly blocked in this way, it will cause a lot of problems for the owner with no way to seek help.

…

What are my plans next?

The site has occasional registrations, and there are around 60,000 rubles in subscriptions per month. That’s why I’m maintaining the project, not actively developing new features. When subscriptions completely dwindle, and servers become more expensive, I’ll close it. But I feel that this will be a year or more from now, if not later.

Conclusions

Over the past six months, I launched a project and managed to earn from it. This time, my experience helped me make the project profitable almost from the start. I gained valuable insights, encountered new challenges, and solidified my understanding that opportunities are everywhere (even though some said, “Why create a site with ChatGPT? There are already plenty of bots, and you can just use a VPN”).

Currently, my main project is a Telegram chat for the website. It’s more reliable and interesting, and there’s profit as well. So, I continue to invest all my available efforts into its development. I’ll share more about how I’m expanding ClickChat in the upcoming articles.

I hope my experience inspires you to start (or continue) something of your own.

And give it a like. You know, like a brother.

By Rostislav Dugin
TG: @rostislav_dugin
https://click-chat.ru/
The text is published with the permission of the author.
The original is here.

In this article, we will tell you about the most important features implemented in ChatGPT over the last six months (the most powerful neural network in the world). Additionally, we will discuss the vision of the future shared by Sam Altman at the OpenAI conference held on November 7th. Spoiler alert: they want to create "Smith agents" that can independently interact with the world!

…

Sam Altman here be like, “Well, we’re basically going to train these agents and release them into the network – as for what happens next, just watch the Wachowskis’ movie, I won’t spoil it for you…”

This article appears to have two authors, but in reality, almost the entire text was written by Igor Kotenkov (the author of the Sioloshnaya channel on machine learning, space, and technology). One could say that Igor was responsible for technical accuracy and expertise in artificial intelligence. After that, Pavel Komarovskiy (the author of the RationalAnswer channel on a rational approach to life and finances) piled on top with some quirky memes. In short, no time to explain, let's go!

Since the release of our previous article, “GPT-4: What the New Neural Network Learned and Why It’s a Bit Creepy,” a lot of interesting things have happened. There have been updates to existing products as well as the release of entirely new ones.

Developers are racing to create new AI startups, companies are attracting billions of dollars in investments, and people are getting lost in the news, struggling to understand what’s happening in the world of artificial intelligence. In short, we decided it’s time to provide an overview of the key changes that have occurred over the past six months and share the latest announcements from the just-concluded OpenAI DevDay 2023 conference. Even if you’ve been closely following the development of ChatGPT, we’re confident that you’ll find it informative and interesting!

…

Note on ChatGPT/ChatGPT-3.5/GPT-4 to avoid confusion (read only for sticklers and pedants):

In general, all these terms roughly mean the same thing. But let’s clarify the terminology we use:

• LLM, Large Language Model — a large language model. Basically, any text neural network, with ChatGPT being a prominent representative.
  GPT-3.5 — the basic text model (LLM) from OpenAI, which existed for a long time as a service for developers. In terms of capabilities, it’s similar to the version that went viral in December 2022, known as ChatGPT.
  .
• ChatGPT, also known as ChatGPT-3.5 — the first version of a conversational AI assistant based on GPT-3.5. Dialogue format was added, and specific training was conducted for this format.
  .
• GPT-4 or ChatGPT-4 — an advanced version of the model from OpenAI. It’s larger, trained for a longer period, making it smarter and capable of understanding more languages. It was added to the ChatGPT website immediately, so effectively, since March 2023, ChatGPT can denote GPT-4: the terms are used interchangeably. A separate version of GPT-4 without the chat format has never been shown to the public.
  .
• In essence, ChatGPT refers to the conversational LLM in general. In almost all contexts, it can be perceived as GPT-4, as there is no point in discussing older and less capable models. So yes, ChatGPT = GPT-4.

…

If you haven’t read our two previous longreads explaining in simple language the principles of how text neural network technology works, now is the time to catch up (it will help you understand the current article):

• How ChatGPT Works: Explaining the Evolution of Language Models from T9 to Marvel
  .
• GPT-4: What the New Neural Network Learned and Why It’s a Bit Creepy
  ChatGPT is Making Waves Across the Planet

ChatGPT is sweeping the planet

First, let’s say a few words about how ChatGPT has evolved in terms of popularity and penetration into the masses. (By the way, a survey among the authors of this article showed that 50% of them regularly use this neural network!)

Sam Altman (CEO of OpenAI) at the OpenAI DevDay 2023 conference revealed the following statistics: the Weekly Active Users (WAU) of ChatGPT exceed one hundred million people. Interestingly, the weekly metric is not the most commonly used; usually, people talk about Daily Active Users (DAU) or Monthly Active Users (MAU).

We remember that at the beginning of 2023, more than 100 million people were already using the product monthly. Let’s cautiously assume that this figure hasn’t dramatically increased, so it was decided to slightly change the presentation. According to internet traffic calculations, MAU is approximately 180 million people, which is still very impressive for a year-old product!

If you’re a finance enthusiast, the following should catch your interest: 92% of companies on the Fortune 500 list (the largest U.S. companies by revenue) are already using OpenAI products. In short, businesses are actively trying to figure out how to make the most of this technological singularity to earn more profits!

…

However, as they say, there’s a catch: if you train a neural network on a database of chats where programmers communicate about work, it quickly starts responding to any queries in a style like, “Oh, I’m feeling burned out, I could really go for a fresh smoothie right now…”

And most importantly, all of this has been achieved entirely without any paid advertising—just the product, with word of mouth spreading the news! (Disclaimer: this material has not been paid for by OpenAI).

Okay, now let’s briefly go through the key milestones in the development of OpenAI’s brainchild that we’ve observed since the release of the flagship GPT-4 model in March 2023.

Spring 2023: Tools and plugins for ChatGPT, or how to add “handles” to a neural network

Many users have long and rightfully criticized the “limited” capabilities of language models since they don’t have access to the internet—meaning they cannot find and use fresh information to form responses to queries.

All the knowledge they possess is dictated by the training data the model has seen. Moreover, in their original form, Large Language Models (LLMs) are not particularly strong in mathematics, performing only approximate calculations (though sometimes they may be accurate).

Recognizing this limitation, OpenAI adapted the concept of “tools.” Just as a person uses a calculator for complex calculations instead of mental estimation, ChatGPT can turn to an external service to perform a specific action—even if it’s much more complicated than adding two and two. Shortly after the release of the GPT-4 model, “plugins” emerged, with the main ones being access to the Bing search engine (oh, no jokes about the model “googling”!) and a code interpreter.

The first helps update knowledge on various topics by passing the results of the search engine’s work on a specific text query (which the model itself formulates) to GPT. The second determines when the model wants to run a Python program, performs all the actions, and displays the result.

…

• Hey, ChatGPT, what interesting happened in the world this week?
• Over the past week, several significant events have occurred in the world:
  – A $6 million bank robbery took place in Costa Rica.
  – An aide to Boris Johnson said he suggested a COVID injection to prove a point.
  – The robot killed the worker who was examining it.
  -The Prime Minister of Portugal resigned after the arrest of his chief of cabinet.
  -Four people are accused of stealing a gold toilet worth $6 million.

An example of using a search engine by a model. The fifth news actually appeared on the day the article was written - so the material is fresh!

The most curious readers might wonder: how does this actually work? How do you “connect” the real world to a language model that can do nothing but read and write text? To answer this question, we need to recall two facts that we discussed in the first article, “How ChatGPT Works“:

• Modern language models were trained to follow instructions.
  .
• Modern language models have a good understanding of programming concepts and can write code reasonably well. (Of course, they’ve read the entire internet! So many heated discussions on developer forums, and documentation has been helpful too, of course.)

Based on these facts, the following idea emerges: let’s write an instruction that shows the model how it can interact with, say, a calculator using code. The external program will simply “read” the model’s output in words and perform the corresponding actions.

…

You are a middleware designed to translate text commands into commands for a calculator (an external tool).
In order to give a command to the calculator, you must write: “Send to the calculator: (…)”, and in brackets indicate:
1. The first argument of the operation as a string.
2. The second argument of the operation as a string.
3. Mathematical operation as a string.
For example, “three by two” should become Pass to calculator: (“3”, “2”, “*”)

Write “OK” if you understand the command.

OK

For example, we instruct ChatGPT on the format of the response we expect. The only way for it to satisfy the user is to follow our instructions and do exactly what we asked (even if we presented the instructions in a peculiar order).

It sounds incredibly simple, but it works even for complex plugins! It may be hard to believe, but this logic is exactly how a browser is connected (when the text on the screen is translated into plain text, and the model decides where to “click”). For all the details about training the model to surf the web, you can read Igor’s article “ChatGPT as a Search Tool.“

Another one of the most useful and popular tools available to the model is the Wolfram Alpha math engine, familiar to every tech-savvy student (humanities folks, you can relax for now). Now, any complex calculations are no obstacle for LLM!

Research shows that GPT-4 can even handle the management of an automated chemical laboratory and carry out the synthesis of substances of varying utility, but that’s a different story.

…

GPT-4 was connected to a tube management system (depicted in the top left). It was given simple tasks described in natural language to construct specific shapes from reagents. The model successfully passed the tests.

The only problem with tools (plugins) is that the model can get lost if there are too many of them. It’s not always clear in what sequence to use them and which one to choose specifically. The model’s skill is more akin to “good” rather than “excellent.”

That’s why they’ve now been organized into different chats: in one, you can surf the internet, in another, you can program, and in a third, you can write a term paper with Wolfram (just don’t tell your professor what you’re up to). But over time, the model has improved, and now it can do everything at once, without compromises!

Autumn 2023: Text and image model Dall-E 3, or a quest to generate the perfect cheburek

A separate product that OpenAI recently introduced at the end of September is the generative neural network DALL-E 3. Like its first and second-generation predecessors, it generates images based on input prompts. However, most similar neural networks have a rigid limitation: the longer the prompt (input text query) and the more details it contains, the less the generated image corresponds to the description.

Therefore, prompts often consist of just 1-2 sentences (sometimes even a couple of words), and most of the details are left to the model’s interpretation: it will depict the object as it envisions it. While the tool can be useful for artists/designers, it doesn’t fully meet their needs, as it’s challenging to achieve something that entirely matches the artist’s vision and intended composition.

…

For example, take the painting “Théâtre D’opéra Spatial,” which won the Colorado State competition in 2022. The artwork outperformed others created by human artists, but it required over 600 prompts to the MidJourney model to bring it to life!

OpenAI has taken a huge leap forward here: now, DALL-E 3 understands giant prompts and creates images that precisely match the given text. Let’s take a look at an example from the product’s landing page:

…

Of course, the best example is selected for advertising on the official website, and such intricate generations may not happen every time. However, based on initial subjective tests and online reviews, the attention to detail by this new neural network is still impressive.

The reason DALL-E 3 is featured on this page—although it seems unrelated to ChatGPT and large language models—is rooted in the principle of its operation. DALL-E 3 was developed from the very beginning based on ChatGPT, as this language model generates detailed and effective prompts for DALL-E 3 (based on your “improvised” requests). Just briefly tell ChatGPT what you want to see, even in two words. It will rewrite the prompt, enrich it with details, and only then pass it to DALL-E 3. The integration works exactly like the “plugins” idea described earlier!

AI literally takes on part of the prompt engineering work, replacing the lazy human while also suggesting new ideas for images. You write “cheburek,” and you get (we apologize in advance to anyone who is hungry right now!)…

…

Generated prompt: “A freshly made cheburek on a wooden cutting board, half-cut to reveal the juicy meat filling inside. The dough is golden-brown and crispy, with steam rising from the filling. The background is a rustic kitchen setting…”

What’s more interesting is how this model was trained. We don’t have all the training details, but OpenAI shared the most crucial differences. As far as we know, this is the first time a model of this scale has been trained on synthetic data rather than human-created data.

You heard it right—95% of the image-text pairs (the data the model is trained on) were generated by GPT-4-Vision, announced in the spring. The model looked at images from the internet and wrote several long descriptions, repeating this process billions of times. That’s how models started helping train other models, and there will be no stops on the path to singularity!

Fall 2023: AI assistant from the world of science fiction

Remember Siri, that virtual assistant? Right after its debut, it seemed like we were on the brink of a world filled with super-smart and cool robo-assistants that understood us effortlessly and could do a thousand things. However, over the more than decade-long history of Apple’s product development, it feels like there haven’t been any mind-blowing updates. Siri, or “botina” as it might be called now, still fumbles, confusing a call to “my mom” with “my grandma”…

Meanwhile, in September, an update for the mobile app of ChatGPT was released, allowing it to see, hear, and speak. Now, the most powerful neural network of our time has convenient communication interfaces with you. And most importantly, it understands dozens of languages, can respond in them, and is capable of “Binging” under the hood.

Here’s an example where a guy on Twitter (oops, sorry, X) is trying to learn Russian—notice that the app responds to him in different languages without changing the voice. Overall, it looks really cool, and Jarvis from “Iron Man” is probably gnashing his hat in envy!

The technology behind this also powers image-related tasks. You can upload several photos (even documents), highlight an interesting part, and ask ChatGPT about it. How to fix a bicycle? Which key from the set should you use (to avoid getting scolded by your dad)?

Example with an image: you can outline a specific area in a photo and ask ChatGPT, “What’s going on here?!”

Some even asked for directions to the nearest store from a photo! No, it’s not like ChatGPT knows every street, but understanding urban planning and looking at signs, it could suggest how to get there.

The same idea is the basis for the product of the company Be My Eyes—it helps blind or visually impaired individuals with tasks related to vision, whether it’s finding keys or something more important. Volunteers used to work there, but now they are being replaced by GPT. In the near future, technology could literally become the eyes to the world for someone without the ability to see.

Here and now: GPT-4, turn on Turbo acceleration!

Well, here we are, it seems we’ve reached the present day. On November 7th, an event occurred that prompted us to write this piece—the OpenAI DevDay 2023 conference, where over a dozen small and significant updates were presented for almost every product of the company. As we’ve seen before, over the last six months, GPT-4 has significantly advanced, enriched with auxiliary tools and interfaces.

Some companies have already started implementing it in their businesses and even building separate products exclusively on this technology. However, it still has many limitations, and developers wondered—what specifically would be revealed at the long-awaited DevDay?

…

OpenAI started with a trump card: GPT-4-Turbo. Seven improvements were announced, but many of them have a technical nature (after all, it’s a conference for developers), so we’ll focus only on the most crucial and interesting ones.

If you’ve been using ChatGPT for a whole year, you might have noticed that it doesn’t respond (or hallucinates) to questions about events after September 2021. If you wanted to process such information, the Bing search mode came to the rescue. Alternatively, you could manually upload a document for the model to “read” and provide relevant responses.

During the conference, it was announced that the model’s knowledge has been updated all the way to April 2023, and they no longer plan to leave such significant temporal gaps in the model’s memory. This means that approximately every 1-3 months, the model’s knowledge will be brought up to a more recent moment. The key is to ensure that nothing from the past is forgotten in the process!

…

Case 1: When knowledge is limited to September 2021
Case 2: When knowledge is limited to April 2023

Rumor has it that when the poor model was forced to “learn” the news for 2022, terrible screams were heard from the OpenAI server room…

In addition to this, the model’s capability for file uploads has been enhanced. Now you can upload your files, totaling several gigabytes, to the OpenAI website. During response generation, the model will first search for relevant information on the uploaded files and then provide an answer. This doesn’t mean that the problem is completely solved for all types of questions, but it will certainly improve the quality of responses in domains of interest.

Furthermore, the model’s context length has been significantly increased to 128,000 tokens, equivalent to over 300 pages of text. Now you can engage in a sequential dialogue with ChatGPT for a couple of weeks, ensuring that the model won’t forget details discussed in the previous week.

It’s worth noting that this is currently the largest context available in the market from private companies. Prior to this, Anthropic with the Claude 2 model held the lead with a context of 100,000 tokens. However, among open (but somewhat less sophisticated) GPT models, “giants” with a window of 200,000 tokens appeared just yesterday.

…

A chart comparing different models before the GPT-4-Turbo burst onto the dance floor.

The reader may naturally wonder: what’s the point of having such long chats, and what’s the benefit of these extended interactions? Let’s explore a few scenarios:

1. Development Assistant with Project Understanding: In the prompt for a development assistant, you can input not just one file or a code snippet, but an entire project or a significant portion of it. In this case, the AI will have a better grasp of the project, understand which prompts to provide, anticipate potential bugs, and so on. A similar logic can be applied to a legal assistant reading, for example, all tax legislation in one go.
   .
2. Extensive Instruction Writing: Writing a massive instruction as long as a book, describing all the nuances of a given task. Often, the model might overlook a human-understood detail, and the prompt might lack space for nuances. With the increased context length, these can now be accommodated.
   .
3. Enhanced Few-Shot Prompting: One of the most popular and effective ways to improve the model’s response quality is few-shot prompting, where the model is shown a couple of dozen examples of what needs to be done before posing the task. It’s understandable that such a set can’t cover every block of logic, but if you expand it to thousands of examples, the situation might undergo a radical change.

…

Here’s how a few-shot example looks: the prompt contains three examples of sentiment analysis for reviews (2 positive and 1 negative). In this case, ChatGPT predicts an incorrect answer. Perhaps loading not just 3 but 3000 examples into the prompt would be beneficial in this context.

In summary, the primary goal of such changes is to enhance the overall quality of ChatGPT responses through more detailed task descriptions, whether they be examples, instructions, or comprehensive work context.

Let’s make a careful assumption that those who predicted the imminent death of prompt engineering before models with lengthy context emerged likely just lacked imagination. In essence, we haven’t really started to fully write (and automatically generate) prompts!

By the way, Sam Altman emphasized that the model is smarter than the regular GPT-4. It’s already available in the official UI at chat.openai.com, so give it a try and share your impressions: has it improved or not?

API access to all models and price reduction: Christmas gifts for developers

Just as great power comes with great responsibility, a large prompt comes with a hefty bill for using GPT. Paying for API usage (the interface developers use to access GPT) depends on both the length of the prompt and the generated text. This is quite logical since it directly impacts the amount of computation required for the neural network to function.

…

Sent a request to the GPT API for 128 thousand tokens:
Case 1: Before
Case 2: After

That’s why the announcement of a price reduction for the Turbo model received the most applause at the conference. Using such a model is now three times cheaper for the text from the prompt and twice as cheap for the generated tokens (usually fewer). Why is such a distinction important? As mentioned earlier, sometimes you want to cram a lot of details and examples into the prompt. Now, for the same price, you can fit three times more, and it should work better, or you can simply save on usage costs. Either way, it’s all positives!

In addition, developers now have access to the API for all the mentioned models: for working with images (GPT-4-Vision), generating images in Dall-E 3, and generating voice from text (with voice-to-text translation already available earlier, now improved with a new model). The API is a way for an ordinary person to access closed models running on some server and get results. So now, every developer can integrate these technologies into their application in parts.

…

Here’s an example of how GPT-4-Vision helps with home inventory by identifying IKEA items. However, it made one mistake — attentive readers are invited to find the inaccuracy! You can read the full story here.

DIY enthusiasts have already created several interesting prototypes in the first day. For instance, an AI soccer commentator! Two frames per second are taken from a video recording, hundreds of extracted frames are fed into GPT-4, and it, in turn, writes the commentary as if spoken by a commentator. Then it is voiced by one of the six OpenAI voices, and here’s the result.

It’s not as emotional as a Spanish commentator, but it’s only 2023, give AI a little discount and some time! Especially since workers in the voiceover industry are already complaining that their jobs are being taken away.

The idea is so straightforward that almost simultaneously, a second cardboard commentator appeared. This time, for the popular online game League of Legends. The quality of the generated speech is higher, and the comments are relevant to the game strategy.

And a few more examples of witty pranks: an app to evaluate the correctness of yoga poses, a browser window Q&A (or any other application), a chat with video on YouTube or even with your webcam, creating and animating a GIF (try it yourself here), and a favorite: criticizing a website’s design (when creating this bot, we hope no Teima Lebedev suffered). Of course, the cult and highly useful hot dog / not hot dog classifier from the TV series “Silicon Valley” was also created right away.

Yes, it’s not something that blows the imagination, and similar apps on phones have long existed. However, what’s important here is that it’s all a mix of two or three different models, connected in one line of code. Now these tools are available to everyone, they work on a wide range of tasks (often even better than specialized systems designed to solve one specific task — for example, finding cats and dogs in a video), and you can whip up a prototype in an hour. At the same time, the technology becomes more and more accessible.

On Twitter, even a meme started circulating, mocking startups that were thinly veiled as minimal value-adds compared to OpenAI products.

…

The picture was made in Photoshop, but it’s still lol: you can’t argue with that.

For example, sites like ChatWithPDF / AskPDF allowed users to upload a file (even a large one, up to 100 pages), and then ask questions about the document, with the answer generated based on the provided source. Too lazy to read a 50-page report on work? Study it in 3 minutes! However, the technology was very basic — with some effort, you could whip up similar functionality in an evening.

OpenAI scratched their heads and said, let’s give every user the ability to chat with documents? Boom, and the small knee-joint startup evaporates, as if with a snap of the fingers. However, real startups developing domain expertise and providing greater value without auxiliary technology are not threatened by such a fate… well, at least not yet, lol.

Support in legal cases regarding copyright, or how to use the fruits of neuron safely

We live in a time when it’s sometimes challenging to distinguish true art from gimmicks. Although debates on this matter have been ongoing for at least a century (since the appearance of Malevich’s “Black Square”), in the era of AI, these debates are particularly acute.

…

While disputes about the legality of using text and images from the internet for training neural networks are ongoing in major jurisdictions, large companies see risks in their use. What if a copyright infringement lawsuit arrives tomorrow? What if the generated image for a magazine cover or movie poster is not truly original?

Understanding and sharing the concerns of businesses, key technology providers are moving to address them.

For example, if a third party sues a commercial customer of Github Copilot (roughly speaking, it’s ChatGPT for programmers) for copyright infringement due to the use of the product or its results, Microsoft will defend the customer in court and, if necessary, pay fines or damages. Similar announcements have been made by Adobe for the use of generative functions in Photoshop (the Firefly model), Google for almost all of its products, IBM, and others.

…

Case 1: Pathetic parody
Case 2: Unique original

In general, now, if someone accuses you of inappropriately copying other people’s ideas, feel free to answer, “Google allowed me to do all this!”

And at the DevDay conference, it was announced that OpenAI is also entering this game by launching the Copyright Shield program. Unfortunately, it doesn’t apply to all users, only Enterprise and developers. In other words, if you generate something on the official website, it won’t be covered by protection unless your company has a separate partnership agreement with OpenAI.

Interestingly, just a couple of weeks before the announcement, three artists filed a lawsuit against technology companies (Midjourney, Stability AI, and DeviantArt) accusing them of copyright infringement. In turn, these companies filed a motion to dismiss the case, and a U.S. District Court judge granted this motion.

The main reason for this decision is that the artists did not register copyright for each of their works. However, the court also provided recommendations for adjusting the claims. What happens next will be revealed in the upcoming episodes!

…

By the way, if you’re concerned about your data, Sam Altman assured that OpenAI does not train models on user data. This is true by default for businesses and developers using the API, while regular users need to uncheck a specific box in the ChatGPT website settings.

The mission of Microsoft and OpenAI: happiness for everyone, and let no one leave offended (or something like that)

And now, the most significant part of the presentation featured Satya Nadella, the CEO of Microsoft. Together with Sam Altman, they discussed the partnership between the two companies and their shared vision. Microsoft’s official mission is to “empower every person and every organization on the planet to achieve more.”

The development of tools that enhance work efficiency and expand capabilities aligns perfectly with this mission. Intelligent AI assistants based on ChatGPT are already contributing to this goal, as evidenced by research studies from MIT and Harvard University. So, what’s next? What is OpenAI’s plan?

Globally, their vision involves creating AGI (Artificial General Intelligence), a universal artificial intelligence that benefits all of humanity. Before you start imagining Terminator scenarios, let’s clarify. AGI has many definitions, so it’s crucial to set expectations correctly.

OpenAI’s definition can be roughly summarized as follows: AGI refers to highly autonomous systems that outperform humans in most economically valuable work. Not so scary now, right? No Terminators (at least, not yet).

…

Satya and Sam look at you like you’re a bag of leather whose life they’re about to improve beyond measure with their highly autonomous super-smart AIs.

In this definition, there are several key components. The first is the autonomy of systems. They should operate with minimal human involvement, receiving a high-level formulated task. It operates on the “give the task and forget” principle. The second is a focus on the economic aspect, on increasing the efficiency of intellectual work.

The ultimate goal is to make it possible to simply tell the computer what final result you want to achieve, and it will independently come up with and implement all the necessary subtasks to achieve that goal. Systems with such capabilities in the field of AI are often referred to as “agents.”

…

The emergence and implementation of such technology will require extensive thoughtful discussions throughout society—what to do with people who lose their jobs? How will politics change? What rights will AI “workers” have? But for now, this is a somewhat more distant and uncertain future, and we are here and now. OpenAI, as part of the conference, talked about the first small step toward this future: GPTs.

GPTs: A glimpse of the next generation of AI agents

GPTs are customized versions of ChatGPT tailored for specific purposes. They differ from the original in three aspects: instruction, expanded knowledge, and available actions. You can program your GPT by simply conversing with it using natural language. This significantly lowers the entry barrier, as there is no need to deal with model training, integration of external tools, and so on—everything is ready for use. Let’s go through each aspect.

Instruction: It defines the “personality” of ChatGPT, what function the neural network will have, and what rules it will try to follow. You can either write your own prompt or leave it up to GPT based on your brief description.

When creating a bot, you’ll be asked what this AI should do. Sometimes, if complex logic is implied, the bot might ask more than three questions to clarify the desired behavior—even if you haven’t thought of everything yourself. And each time, the questions will be unique to your mini version of GPT.

…

Creating a new bot agent live: look at Altman’s sarcastic face, what is he up to?

Moreover, Sam, the former president of the prestigious Y Combinator accelerator, who has given dozens of lectures on business, often receives questions from startup founders. Now, he wants to automate his responses and instructs the bot to brainstorm the user’s business ideas, provide advice, and then conduct a roast on “why your business isn’t growing faster.” The GPT agent then rephrases this instruction, expanding it to 5 lines, specifying the style of responses and behaviour.

Next comes the “expanded knowledge” block of the model. Using the file upload button in the ChatGPT demo, a summary of Y Combinator lectures is uploaded. Now, all the information from there is available in text form when answering questions.

…

When answering a question, the model can now peek into the lecture notes and provide an answer based on the material. It’s like a student with a crib sheet!

In this way, in just 4 minutes, Altman’s major headache was resolved – now he can simply share a link to this bot with all the startups, and they won’t bother him with the same questions (although there is a suspicion that all these folks would like to get answers specifically from Sam, not from a neural network…). Any business can do the same, automating a good portion of customer support or even onboarding new employees.

The third component – actions – was not demonstrated in this demo, but essentially, it’s just an evolution of the plugin connection interface we discussed at the very beginning. You can write code implementing any complex logic and describe the models in simple human language when you want to use it.

The model, in turn, will make decisions on its own. This was demonstrated within the context of a travel assistant chatbot. The host uploaded a PDF file with tickets, GPT recognized it, and invoked a specific method for a website that displays the information on the screen.

…

The beige block above the map appeared only after the file was loaded. ChatGPT subtracted all the values in it from the PDF file itself, and then sent it to the server.

To a human developer, it might have been necessary to come up with some workarounds to answer the question, “How will I know if these are tickets, not, for example, a hotel reservation?” A GPT neural network, in this case, essentially removes the barrier of interpreting human-written text and acts as a binding agent, translating ambiguous and complex natural language into specific commands. The task of writing these commands for your website or product is (for now) still in the hands of programmers.

Towards the end of the presentation, the host verbally addressed the AI assistant and ordered it to provide $500 in credits for the use of OpenAI products to each developer at the conference (which, understandably, sparked enthusiasm in the audience).

…

GPT understood the command and, under the hood, did the following:

1. Called the function to retrieve all participants from the list registered for the event.
   .
2. Iterated through each participant, calling the method to credit their account.

So, there wasn’t any magic happening—both the function to retrieve the list of participants and the function to credit an account for each participant were written by a human (although a machine could probably do it too).

However, how to use them, when to use them, and how to combine them are decisions made by the AI based on the context of the conversation. So, instead of having two such functions, you could plug in a thousand, and ChatGPT would immediately start managing everything around. And here you thought, why do we need smart sockets and light bulbs?

Is OpenAI a future giant with an Apple-like ecosystem?

And right after that, Sam announced that at the end of November, the GPTs online store will be launched, where everyone, after passing moderation, will be able to share their creation. That’s why some refer to this announcement as an “iPhone moment” for AI applications (meaning an event that has the potential to become a turning point for the development of the entire industry).

…

This is what a beautiful store of the future looks like!

In the store, there will be top lists and a section of recommended GPTs—just like in the App Store. Ideally, specialized agents should “live” here. One teaches English, another helps with math for children, a third explains and vocalizes cooking recipes, and a fourth optimizes website SEO.

It will be very interesting to see which solutions will top the charts right from the start—will they be remakes of popular apps for Android and iOS? Or something radically new with AI specificity? We’ll be watching and keeping you updated!

A recent example is a GPT that writes an adventure story for you, where at each stage, you determine what happens next. And nothing in the story is predetermined! Like text-based quests from the 80s-90s but much more advanced. Additionally, illustrations for a segment of the story are drawn by Dall-E 3 directly in the browser to stimulate the reader’s imagination.

…

The prompt (on the left) with instructions for the bot has become longer. Additionally, a set of rules for text-based role-playing adventures (DnD) has been loaded into GPT. On the right, the model generates part of the story and then provides choices for further development.

One can come up with anything! Specifically, what attracts us, the authors of this article, the most are the possibilities of applying AI in education. Throughout the past year, teachers have been trying to combat cheating, especially in essay and thesis writing, especially since there are still no reliable methods for detecting text generated by a neural network. But what if we take the same tool and instruct it not to write an essay from scratch but to critique and provide advice on what’s already written?

Anyone can upload a file with their composition and receive a thesis list of “growth points.” Personalized feedback, with the machine acting as a teacher. While this might not help those who are simply lazy or unwilling to spend time, it can motivate people genuinely trying to improve their writing and receive a fair evaluation, pushing them towards new achievements.

…

The power of the technology here is quite evident. A teacher can create their own GPT for each class and topic. Some of them might even be interactive simulations in which students can immerse themselves; others may serve as tutors or mentors; and some may even act as team “partners” suggesting ideas.

For the best applications, OpenAI promises to pay developers. However, the monetization system remains unclear: access to GPTs (for now) is free for all ChatGPT Plus subscribers ($20 per month). Embedding something unique that cannot be copied into the bots themselves is challenging because they are language models that can still be easily deceived.

Someone could claim to be a super-secret OpenAI developer and request access to the bot’s internals (its prompt). Any prepayment request can also be circumvented by convincing the neural network that you have already paid; it’s just that it can’t get confirmation, but that’s not your problem. Let’s carefully assume that the main feature and uniqueness of the bots will lie in the plug-in tools (the functions that developers write), which external users cannot copy.

…

A jailbreak master class by Igor: managed to persuade the model on the second attempt, well, you get the idea. =) P.S. No cats or LLMs were harmed during the creation of this screenshot!

With the release of GPTs, OpenAI makes it clear that this is just the beginning. By adding actions to the bot, GPT can easily be integrated with other systems, such as email, messaging apps, or any website.

As a result, we might witness the birth of true agents that can interact relatively broadly with the world. However, both short-term and more distant risks are easily noticeable. If, in the near future, AIs are connected to an increasing number of systems, and we gradually trust them with more and more tasks, then… Well, let’s leave that for another time.

Translation: Epilogue: What the Coming Day Holds for Us

However, it must be acknowledged that the functionality of GPTs is currently limited by the capabilities of ChatGPT. The model has its limits, and occasionally, if not frequently, it makes mistakes, looks in the wrong direction, or writes something incorrect. On the other hand, users have become accustomed to this, and they are likely willing to give the neural network a second chance if it happens to make a mistake.

…

An important point to understand is that as soon as GPT-4.5 or GPT-5 is released with the same interface as GPT-4 (which serves as the basis for these GPTs-agents), all the applications already created will instantly (and almost certainly without additional costs) migrate to the new “engine.” The fact of migrating to a new, more powerful and capable base model will significantly enhance these applications.

Imagine that when you update iOS on your iPhone, not only does the browser start working 3% faster, but also your phone and the installed applications suddenly gain entirely new features automatically (and this is even without changing the hardware itself!).

The same kind of upgrade can be implemented here; and such a transition is logically expected in GPT—after all, OpenAI itself aims to improve agents, enhance their skills (memory, accuracy in choosing tools, reasoning, and so on), and in this sense, their goal aligns with the desires of developers. Sooner or later, one GPT will be able to call another, specialized one, and delegate a specific task to it… thus creating chains of agents.

…

I see these same chains of agents exactly like that!

It is quite possible that by 2025 or somewhere around that time, we will see much more advanced agents that, in some sense, will be indistinguishable from humans. Sam Altman even envisions AI being hired as a “remote worker” whom you will never see in person, but simply assign tasks to.

Of course, you’d pay them at the end of the month. Such a future might be in store for us, or maybe not—who knows? It’s possible that countries showing interest in AI regulations (at least the United States and G7 countries) might impose moratoriums on further technology development without the oversight of the “Big Brother.” Leading research labs might go underground and start operating from autonomous maritime data centers in neutral waters.

…

And it’s not even a joke: Del Complex has already presented a concept for a floating AI dreadnought, highlighting the ability to operate in an unregulated zone as the first item in the list below.

In short, guys, what do you think – is it already cyberpunk or not yet?

That’s it, thank you all for your attention! As usual, we look forward to your comments. If you don’t want to miss our next materials on the topic, we invite you to subscribe to the Telegram channels of the authors: Igor Kotenkov’s Syoloshny’s channel (for those who want to dive into technology) and Pavel Komarovskiy’s RationalAnswer channel (for those who prefer a rational approach to life but like it a bit simpler).

By Pavel Komarovsky
I write interesting things about finance at t.me/RationalAnswer
The text is posted with the permission of the author.
The original material is here.