"Bulgakov is ours": how Ukrainian AI went wrong and other high-profile failures of neural networks
Artificial intelligence is no longer just a trend, but a reality that saves millions or… burns them in a matter of seconds. At every conference, we are told about the amazing successes of AI, but for some reason, Ukraine is accustomed to keeping quiet about fake news, wasted budgets, and the dark side of AI. We decided to break this stereotype and analyze the topic without embellishment!
In this episode of the dev.ua video podcast, a star-studded company of tech experts has gathered to break the rose-colored glasses and discuss the most high-profile failures of modern technology. No boring theory — just real insider stories, millions in losses, bots and fake accounts that you will never hear about in beautiful advertising demos.
Artificial intelligence is no longer just a trend, but a reality that saves millions or… burns them in a matter of seconds. At every conference, we are told about the amazing successes of AI, but for some reason, Ukraine is accustomed to keeping quiet about fake news, wasted budgets, and the dark side of AI. We decided to break this stereotype and analyze the topic without embellishment!
In this episode of the dev.ua video podcast, a star-studded company of tech experts has gathered to break the rose-colored glasses and discuss the most high-profile failures of modern technology. No boring theory — just real insider stories, millions in losses, bots and fake accounts that you will never hear about in beautiful advertising demos.
We figured out why a beautiful POC with 200 lines of code turns into hell in production, why mass layoffs for the sake of AI agents end in system crashes, and how a regular bot can delete a database or wipe out a company’s annual budget overnight.
Release Partner — De Novo: Learn how to build AI projects safely and cost-effectively. De Novo has the largest cloud GPU infrastructure in Ukraine with NVIDIA H200, H100, A100, L40S, L4 cards, allowing businesses to train complex AI models on powerful hardware and without transferring sensitive data abroad.
Episode participants:
Stas Yurasov is the host and CEO of dev.ua, a tech journalist and skeptic.
Alexey Minakov is an AI expert, business consultant on AI implementation.
Vlad Melnyk is the co-founder of Lapatonia.
Dmytro Fedorenko is the AI director at De Novo.
What exactly are we talking about in the issue:
How was Lapatonia created and why did the scandal with Bulgakov arise?
Klarna and Amazon cases: why did the mass replacement of people with AI agents fail?
Token Economy: How Bots Burn $67,000 a Night and How to Protect Yourself from It.
Legal disasters: from fabricated precedents in the US to offended judges in Ukraine.
The uprising of autonomous agents: how AI decides on its own to change the decisions of its masters or delete databases.
Have you had any oddities or problems with artificial intelligence? Share your stories in the comments!
Don’t forget to like, subscribe to the channel, and share this video with those who still believe in the infallibility of AI!
Timecodes
00:00 — Introduction: Teasers about Amazon layoffs, vanilla demos, and lost millions.
00:59 — Greetings and introductions to guests at the dev.ua studio.
02:41 — How Lapatonia was created: hackathon, 100 million tokens per day and the sudden loss of GPUs from NVIDIA.
04:15 — Cooperation with De Novo: why Ukrainian models need a powerful infrastructure.
05:45 — What makes Lapatonia useful: simplified access to Ukrainian models (LAP and MAMAY) and data security.
07:42 — How people use the service: API vs. Web chat and token volumes.
09:22 — Fakap #1 (Lapatonia): Why the model praised Bulgakov and Putin, and how it was turned into a feature.
12:15 — Sovereignty or lagging behind? Why the state is not keeping up with the NLP community.
13:28 — The RADA model: how a hybrid approach combined the speed of LAP and the quality of MAMAY.
15:00 — Are «AI wars» possible and why Ukraine needs its own digital sovereignty.
17:04 — Fakap #2 (ML Cloud): How to create a quality product for ML engineers and miss the market by $1,000,000.
20:24 — Fakap #3 (Klarna): Why firing 700 people and replacing them with AI agents failed.
22:11 — Fact #4 (Amazon): They laid off 30% of their staff due to the AI boom, and within a few weeks the system collapsed.
23:50 — «Translation from pocket to pocket»: how OpenAI and NVIDIA are turning over billions of dollars.
24:45 — AI efficiency: why employees gain time for themselves, not for the company.
25:23 — «AI Slop» and coding problems: senior developers spend more time reviewing generated code.
26:09 — Fakap #5 (MFA of Ukraine): Where did digital consultant Viktoriya Shiy disappear to and the scandal with her prototype?
27:51 — Fakap #6 (McDonald’s): Failure of voice bots on McDrive due to surjik, accents, and street noise.
29:10 — Fakap #7 (Kulava, Uber, Hermes): How bots attack APIs and burn through $67,000 and annual budgets overnight.
32:37 — Self-hosting vs Cloud API: how to protect yourself from sky-high token bills.
33:08 — Fakap #8 (Legal): A lawyer in the US invented 6 fake precedents via ChatGPT, and Ukrainian judges are offended by AI.
35:38 — Fact sheet #9 (Air Canada & New York): Lawsuits over non-existent discounts and incorrect advice from bots.
38:00 — Fakap #10 (Facial Recognition): How a London supermarket mistook an actor for a robber.
39:37 — AI Washing & Rogue Startups: How instead of «smart agents» people from the Philippines work under the hood.
42:14 — Fact sheet #11 (Amazon Recruiting): Gender discrimination in resume evaluation and modern biases in LMK.
