An interesting story has just emerged about Huawei: a few hiring managers were caught selling positions at the company for $2,800 and then charging the hired employees a monthly fee of $420 to keep their jobs (allegedly). Many people were punished today, and even more were fired immediately.

What can we learn from this? First, Huawei is a company where people are willing to pay just to work there. Second, Huawei is capable of responding to such situations swiftly and effectively. Third, and perhaps most concerning, Huawei is likely not the only major tech company where this kind of corruption may occur.

Это краткое содержание Angry Tests, моей новой книги о тестировании. Будет опубликована 22-го апреля на Amazon. Можно заказать со скидкой 10% уже сейчас (если готовы получить ее через CDEK). Отправим 22-го апреля!

I'm reading Clean Coder (2011) by Robert Martin and catch this (p.159): "We didn’t become programmers because we like working with people. As a rule we find interpersonal relationships messy and unpredictable. We like the clean and predictable behavior of the machines that we program." Agree?

In 1995, when I was a student, I've become a member of ACM, for $35/year. Today, I've renewed my membership for the 30th time, for $99. Does it make me a scientist? I don't think so. Does it help the organization that refused to support ICCQ in 2022? Well, a little. Science has no borders, I believe :)

После просмотра первого сезона Натали и Александр (2025) от Владимира Щеголькова и Владимира Битокова захотелось перечитать Пушкина (ладно, кого я обманываю, прочитать впервые). Никогда ранее, сколько ни водили меня в детстве по его домам-музеям, мне не были так интересны его произведения. Оказалось, что он не только памятник на площади, но и пьяница, ревнивец и ловелас, в хорошем смысле. Александр Зарядин в роли поэта и Ксения Трейстер в роли жены — режиссерская удача. Вечное безделие и праздность самодержца Николая I — режиссерский промах или замысел? Спойлер: тема Дантеса не раскрыта, ждем второй сезон. #кино

This is Huawei office in Dongguan (Shenzhen, China). While filming, I'm waiting for a train to take me to another part of the campus. Nice, huh? (more videos here: @bloghacks)

Finally, 5K GitHub followers, after 15 years of contribution.

Появились детали лекции в НГУ 11-го апреля, можно регистрироваться: в 16:20, аудитория 2322.

Онлайн трансляция будет на RuTube.

Франция проголосовала против закона, разрешающего платформам типа WhatsApp и Telegram читать наши зашифрованные сообщения. А как проголосовали бы вы?

А не организовать ли нам очередное совещание в Москве? Обсудим, среди прочего, наиболее важные главы из Angry Tests. Предварительно 6-го июня в 18:30 где-то внутри Садового. Заполните, пожалуйста, форму, если хотите попасть (не всех сможем пригласить). Вход бесплатный, с меня угощение и компот.

How do you know which parts of your website are scanned by OpenAI and used to train ChatGPT? You don’t. But soon, you will. The IETF is working on a standard—most likely a web URL, similar to /robots.txt—where you’ll be able to set your preferences. These preferences will be respected by all AI scanners. Maybe.

Перед покупкой и прочтением Angry Tests, рекомендую посмотреть интервью (58 минут) с Michael Feathers, автором книги, по которой я учился тестированию: Working Effectively with Legacy Code (2004). На первой странице я цитирую его: "Legacy code is simply code without tests."

Do you know the game "Chinese whispers" (испорченный телефон)? I suggest a benchmark for code-editing AI models based on the same idea: the more transformations a piece of code can survive, the better the model. Take a simple Java algorithm, give it to Claude, and ask to improve it. Then take the result and repeat the process. Eventually, the algorithm will be broken. The question is: how soon?

Интервью с Алексеем Гусаковым, CTO группы поиска в Яндексе. Обсудили их продукты, работу с кадрами, борьбу за качество, конкуренцию с мировыми лидерами и вечную боль поиска кадров. Получилось интересно и нашлось чему позавидовать. Смотрите на YouTube и на VK (84 минуты).

Last year, I taught object-oriented programming to BSc students at Innopolis University. The course included eight 80-minute lectures. In the newly published video on YouTube and VK (102 minutes, with English and Russian subtitles), you’ll find the most important and engaging moments from all eight sessions. Enjoy!

В ближайшее время собираемся открыть дискуссионную группу в Телеграм под названием "Elegant Sexists". Будем обсуждать проблемы взаимоотношения полов в среде программистов и около. Где найти вторую половинку? Как познакомиться? Как вступить брак (и нужно ли)? Как строить семью, будучи программистом? Что скажете, дело стоящее? И кстати, из какого кинофильма картинка?

I'm planning to revive the Shift-M podcast, where I previously spoke with software engineering experts from around the world. The next guest will be Richard Pawson from the UK, the author of the Naked Objects paradigm (and book) and the creator of the Elan programming language. I believe he'll have many interesting insights to share about object-oriented programming. What questions would you ask him?

Every so often, ThoughtWorks—where Martin Fowler is Chief Scientist—publishes a comprehensive study on the current state of the art, called TechRadar. A new edition was released yesterday (PDF here). I highly recommend reading it if you want to stay informed about what matters in our craft right now. Their opinion on AI coding assistants like Claude-Code is this: 'We remain cautious about AI-generated code; despite some very good results, we still see a strong need for steering and vigilance during code review; with great power…"

I just finished watching an interview of Yann LeCun—Chief AI Scientist at Meta—where he basically echoes what we discussed earlier in the M188 video: Large Language Models can't truly reason, because they lack an abstract model of reality inside them. They just repeat what's been said before. On the other hand: how many human beings actually reason, rather than just repeating what they’ve heard?

How do you make a good-looking website without hiring expensive (and often annoying) designers? Back in 2015, we created a simple CSS framework to solve exactly that. Tacit gives you a single CSS stylesheet to attach to your site—and it just looks great and mobile-friendly. The best part? You don’t use any CSS classes in your HTML5. You focus on the data; Tacit handles the look and feel. The UI of Zerocracy, for example, is built this way. Some of you may have already used it—what’s your opinion?