Not all the Chinese providers have the same vibe. DeepSeek and MoonshotAI under-market and over-deliver consistently. Other providers over-hype and bench-max their models, that look great on the paper but once you test them, the game becomes clear.

90% of the types vs tests debate is typing advocates thinking all dynamic languages are like JavaScript and dynamic devs thinking all type systems are OO-based.

I really respect what you do, and I read your blog like it is a superhero comic, but I think the framing is wrong. You show that Zig is an amazing and safe language if you have the kind of IQ needed to invent TigerStyle and the discipline to follow it strictly. I would like to be like you, and I think many others would too, and many will probably at least try. But not everyone is able to make that kind of intellectual or self-discipline effort. I am too dumb to manage memory by hand and too chaotic to just follow established patterns. My IQ is only just enough to handle Rust, and instead of discipline I only have enough willpower not to tell the borrow checker to fuck off when it yells at me.

We live in a society where people *assign* other people *books* to read *against their will*.

why: I am so tired of worrying about & spending lots of time fixing memory leaks and crashes and stability issues. it would be so nice if the language provided more powerful tools for preventing these things.

Anthropic employees are what happens when you stop bullying nerds

RT @C_Kazmierczak: W negocjacjach akcesyjnych do UE popełniliśmy jeden błąd - koncentrowaliśmy się na wynegocjowaniu jak najwyższych fundus…

dtach > tmux > screen

MOVING TO POLAND IS A TERRIBLE CHOICE FOR 95% OF PEOPLE Poland is the fastest-growing economy in the EU. It’s modern and has world-class services. But people who overhype it don’t tell you one thing - if you don’t have an EU passport (and 95% of people in the world don’t have one), Poland is gonna turn your life into a living hell. The reason for it is not its people or nationalism. The reason is bureaucracy. Poland wasn’t built to host a lot of foreigners. For the best part of its history after communism, it’s been a poor country, and hardly anyone wanted to move there. Then the Ukraine war happened. Millions of Ukrainians and Belarusians flew into the country. Poland’s immigration authorities became overwhelmed. 4 years flew by, but the situation didn’t improve. When you move to Poland, you need to get a temporary residence permit. It lasts between 1 and 3 years. Getting it now is not feasible. The wait for a Polish temporary residence permit is between 8 and 12 months. In most cases, you get your first residence permit, called pobyt czasowy, for a year. Then you need to renew it. You will need a long-term apartment in Poland and a source of income in Poland. Your company’s registered in the US or Singapore won’t do. After 3 years, you can get a pobyt stały (permanent residency) - and wait 12 months again. The situation is better in some regions like Krakow, but in Warsaw, where most people go, it’s the worst. While you are waiting for your residence permit, you can’t leave Poland. If you leave after your visa-free days in the EU or your visa expires, you won’t be able to come back without a Schengen visa. To get a residence permit, you need to file an application. This requires either sending all your documents by traditional mail or scheduling an appointment. Getting an appointment takes 8 months. You have a question? Get ready to wait up to an hour in the telephone queue. Hardly any immigration workers speak English. Most government websites and documents are in Polish only as well. There is a new system where you can file an application online. But using it requires a government ID number called PESEL. You can get it in Poland, but it requires a Polish address. Alternatively, you can convince Polish authorities that you have a valid reason why you need a PESEL. This typically requires employment or business. Both will have real consequences for your life and potentially make you liable for Polish taxes and health insurance contributions. This is what US and UK founders get wrong – you can’t just come to Poland with your online business and live here. Poland doesn’t have digital nomad or independent entrepreneur visas. In Poland, you need a visa sponsor. That means that you either need to find a regular Polish job and employer that will sponsor you or found a company with substantial capital, employees, and real service to sponsor yourself as a director. Want to start a company? All the tax and company registration portals are in Polish. Getting a Polish accountant and tax advisor will be your only option. The situation is not likely to improve in the future. The Polish politics are dominated by right-wing anti-immigration sentiment. Under the right-wing influence, even the center-left are introducing laws that make it harder to immigrate and get a Polish citizenship. For everybody cheering for anti-immigration and thinking it only applies to the people from the Middle East and India, bad news. If you have a UK or US passport, you will fall under the same rules. Some categories of people, mainly people with Polish roots, will have it much easier. But for most people without an EU passport, moving to Poland will turn into the toughest bureaucratic challenge of their lives.

Time to talk about this one. CopyFail (CVE-2026-31431) — a 732-byte Python script that roots every Linux distro shipped since 2017. 🧵

prawda, mogą Joãować tej straconej bramki. ale z drugiej strony to dopiero 40', nigdy Neves, co się wydarzy po przerwie.

Hipokryzja zwolenników poboru jest po prostu bezgraniczna. Państwo, Naród, Ojczyznę cenią tak bardzo, że bez mrugnięcia okiem złożą na ich ołtarzach czas, zdrowie, przyszłość, a może i życie tysięcy współobywateli. …ale wciąż nie na tyle, żeby chociaż zatrzymać się i rozważyć jakie konieczne warunki i jakie koszty należałoby ponieść by ludzie chcieli ich bronić dobrowolnie. Skoro grozi nam egzystencjalne zagrożenie to może 5% PKB na obronność to za mało. Może trzeba wydawać 8, 10, 15? Ba, i 20% nie brzmi źle jeżeli alternatywą ma być ruska niewola i anihilacja wszystkiego co przez 35 lat zbudowaliśmy. No ale wtedy trzeba zmierzyć się z pytaniem o koszt alternatywny: albo godne wynagrodzenia dla zawodowych żołnierzy albo transfery socjalne; albo czołgi albo nauka; albo wydatki na infrastrukturę cywilną albo na fabryki dronów i amunicji. Tej dyskusji zwolennicy poboru w ogóle nie chcą widzieć. Skumulowane wydatki krajowego sektora publicznego na poziomie 52% PKB to dla nich luźno interesująca anegdota. Tymczasem pobór to właśnie próba ukrycia sposobu finansowania obrony państwa przed budżetem. Bazując na nim nie ma po co zastanawiać się nad tym gdzie, na co i jak wydajemy pieniądze publiczne skoro ludzi można opodatkować „w naturze”. Bo przecież dokładnie tym jest pobór: przymusowym świadczeniem pracy na rzecz państwa, której wartość jest sztucznie zaniżana poniżej uczciwej, rynkowej stawki. Ze wszystkimi ekonomicznymi konsekwencjami tego stanu rzeczy dla poprawnej alokacji talentów, spadku efektywności i morale wreszcie straconej produktywności tak cywilnej jak i wojskowej. A jeżeli państwo nie jest w stanie przekonać obywateli do obrony poprzez godne wynagrodzenie, warunki i instytucje - to problemem wciąż nie będzie brak przymusu, tylko brak wartości, którą ktokolwiek chce chronić, a co dopiero za nią umierać. Z tego punktu widzenia pobór - a nawet sama o nim dyskusja - to po prostu dowód nie siły, a słabości państwa, które odsłania się jako niezdolne do przekonania obywateli, ze warto bronić go dobrowolnie.

Sci-Hub is an evil website that pirated 85M research papers and made them freely available And now they've added AI to their database to make Sci-Bot. It answers your questions using latest, full-text articles. But DO NOT use it. We should all try to make billion-dollar academic publishers richer. I'm putting the link below so you know how to avoid it.

If this was how the buttons looked, what portion of humanity would press blue? It'd probably be a large enough number due to mistakes, the young, altruists, etc., such that it remains wise to press blue.

i actually think it’s immoral for anyone to press the blue button because you are forcing everyone else into a situation where they have to take on risk of death to protect you from your own decision

But this is foolish. Allowing risky behavior, or even seriously considering the blue pill, pushes the rest of society toward the same risk. People like to see their own blue choice as a way of saving others, instead of seeing it as irresponsible behavior that forces other people to take risks in order to save them. Instead of thinking: "I choose blue to help other blue choosers survive." People should think: “I choose blue, which means I make myself a person in danger, and now I need others to choose blue too, or I will die.” Blue is what creates this hostage situation. The rule should be: "do not create a risk that other people will later have to help you deal with".

People who complain don’t know what DeepSeek used to be or how it developed. DeepSeek delivered everything it promised. They were supposed to solve context, and they did. Even though the models are undertrained, the result is good, and for Flash it’s honestly outstanding. We now have 1M context with good recall, at a flat price, with almost no drop in performance. But DeepSeek was never an outstanding product, neither as a chat app nor as a model. DeepSeek doesn’t seem especially interested in playing the sales game. They’re not doing this because it’s their business model, they’re doing it out of generosity. If they could, they probably wouldn’t leave the basement at all, they’d just chase AGI and do Science. Above all, DeepSeek is a compass for other companies. I’m convinced that thanks to DSV4 and papers produced by Moonshot about Kimi, the next Kimi k3 will be the best product on the market, even if maybe not for enterprise. The same goes for GLM or MiniMax. Who want GLM with 10mln context window? And this is something the complainers don’t seem to notice. Or maybe they just don’t remember how weak the first revisions of DSV3 were, how they fell into death loops, or how uneven R1 was in many ways. DeepSeek doesn’t create products, it creates the foundations for them. It creates Science, puts forward hypotheses, and then proves to the world that it’s possible to serve a 1.6T model on old hardware with a 1M context window at almost linear speed.