Artem Evsikov

first glance at @lutinapp what do you think? could this format be better than raycast or chatgpt hotkey chat? try at lutinapp.com

@zeddotdev I hate anthropic

Great minds...

@kennethnym x.com/aevsai/status/…

this must feel so good for the puter

autistic genius intelligence

в целом ситуация такая

@uwukko @luciascarlet 🧩🍿🤝

@luciascarlet i pay a small fee for access to a piracy streaming service with all movies/shows from everywhere ever for convenient streaming on all devices instead of maintaining a home server or a torrent box

what’s your most embarrassing sub mine is either ChatGPT Plus or SoundCloud Pro for that one account I don’t even release music on

Is there serverless inference provider with adequate and stable latency? fireworks latency for kimi 2.5 spikes to 20+ seconds till first bite on simple requests and median is at 17.4s today, just pure madness

Parallel Agents just shipped. Mix and match any agent, run them all at once, and manage everything from a new Threads Sidebar. Claude Agent, Codex, Zed's agent, or anything on ACP. One window, across all your projects. zed.dev/blog/parallel-…

@_architected Понял что за фильм еще по первому твиту)) видел фотки со съемочной площадки. Жду фильм! Будет интересно посмотреть чего и как расскажут

а, ладно, под НДА оказывается я могу рассказывать что это был Artificial про OpenAI с Эндрю Гарфилдом и Юрой Борисовым.

неделю назад ходил на предпросмотр фильма про один биг тех с одним русским актером в главных ролях, но из-за НДА ничего не могу рассказать :(

😔🫸 всеобщий базовый доход ☺️👉 владельцы эйАй платят безработным, чтобы те не громили их дата центры

@_architected Zed + Claude 100€/mo для фронта+ Codex 23€/mo для бэка/сложной логики После того как завезли гит стало вообще супер

Так окей, у меня неделя вопросов. А код вы в чем пишете сейчас? Для меня относительно нормально работает связка VS Code + Copilot, но я снова потрогал Antigravity и Zed и как будто бы оба стали сильно лучше.

@tokiorys @_architected В превью мультиагентная разработка появилась кста

@_architected Zed. Агентов в их панели с ACP использую по надобности. Мульти-агентной параллельной разработки там ещё нет, но она и не нужна, на крайняк просто терминал открываю

@Nick_Davidov Love perplexity computer for ease of use and actually nice generated docs, but it’s so expensive. Once 4k credits are gone I probably would have to go back to claude, their subscription value is too good and it would be hard to rationalize another 200$ sub 🥲

Even without this my computer spend is close to $800 per months and yet these bastards come up with new ways to deprive me of my money

Claude Sonnet’s p90 latency rose 44% in three weeks inside Zed. Today we’re launching Agent Stats: a weekly view of AI agent adoption and turn times in Zed. See the data, and what it can and can’t tell you about how people actually use agents: zed.dev/blog/agent-met…

Hey @zeddotdev, nice merch!

I've become so sick of inflation of meaning caused by LLM that in 2026 I started journaling and write notes by hand. It actually helps to slow down, think and reflect (incredible, right?) With LLMs generating code, emails, answers – form become so cheap that I stopped to care about it. This feeling is similar to banner blindness, when your brain skips ad banners, but now I feel how harder and harder it becomes to read texts. Even ones that I wrote. I guess that why handwriting stuck with me only now, despite previous attempts of journaling and writing.

The fact that i now daily drive @perplexity_ai in comet browser and prefer it over Claude and ChatGPT makes me optimistic about possibilities of making AI product without building foundational models in-house. I thought that LLM providers like OpenAI and Anthropic will inevitable eat up most AI apps just by horizontally scaling into new domains. Just make llm go brrr and you are good, dah But I guess you either focusing on building frontier models or digging into domain and compete within it by building the best products. Focus and lack of it makes a big difference as we can see with Anthropic vs OpenAI rivalry

Ok google, I wasn't expecting that

@BenJames_____ Lol! Hope id doesn’t explode on running claude in dangerously skip permissions mode. I made mine to appear on screen when it is done with task x.com/aevsai/status/…

I made a USB-Clawd who gets my attention when Claude Code finishes a response

Попросил у Qwen 3.6 досчитать до десяти, и его chain of thoughts напоминает душевнобольного человека :) Мне его даже жалко немного стало. -- I will output the Russian text now. "Конечно, вот:" "1, 2, 3, 4, 5, 6, 7, 8, 9, 10." This is good. Wait, is there any nuance with "досчитай"? Maybe "count up"? Yes, 1 -> 10. If they wanted countdown, they'd say "считай от 10". Okay, proceeding. "Конечно, вот: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I will write it without quotes. And maybe a line break for readability. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." This looks clean. Wait, I see "Formatting Rules: Use Markdown for lists...". Maybe I should use a list? 1 2 ... No, that's silly for counting. I'll just use text. Final Answer structure: [Polite acknowledgement] [The sequence] "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll generate this. Wait, "ладно" can mean "fine" or "alright". User: "What do you know about X?" User: "Fine, count to 10." This implies impatience or a pivot. A simple response is best to avoid further annoyance. "1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Just the sequence is the most efficient, polite enough for a command. I'll add "Конечно" because it's an LRP (Large Response Persona) trait to be helpful/conversational. "Конечно. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll stick to that. Wait, could I count in binary? No. Could I count in hex? No. Standard decimal. I'll provide the standard response. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll format it with a newline for clarity. The user might have typed "а досчитай" as a test of basic capability. Okay, writing response. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Wait, maybe I should capitalize numbers? No. Maybe spell them out? "один, два..." Digits are clearer. I will provide digits. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Wait, "досчитай до десяти" implies reaching 10. If I stop at 9, it's wrong. If I include 10, it's right. Inclusive. Okay, ready. Final check: Is there any trick with "досчитай до десяти" in Russian slang? "Раз, два, три..." No. I will provide the numbers. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Wait, maybe the user wants me to count slowly? I can't control reading speed. Maybe the user wants me to count to ten words? "досчитай до десяти" -> 1-10. Okay, I'm confident. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll leave it at that. Wait, I need to make sure I am outputting text in Russian. "Конечно" is Russian. The numbers are universal digits. "Конечно, вот: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." This is the safest, most polite, and accurate response. I will generate this now. I'll add "без проблем" (no problem) or just "конечно". "Конечно" is safer. I will write the response in Russian. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Wait, should I put it in a code block? "Format all code blocks..." - no, numbers are not code. "Format file names..." - no. "Use Markdown for lists..." - no. "Use dollar-sign delimiters..." - no. Just plain text. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Ready. One nuance: "досчитай до десяти" can be "one, two, three...". In Russian, "раз, два, три..." is also counting. But usually "1, 2, 3" is fine. If I write "1, 2, 3..." it's clear. I will use digits. Final Answer. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." Wait, I will add a period at the end. "Конечно: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll output this. So I just answer the count. "Конечно. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." I'll go with this structure. It separates the text from the numbers. Or "Конечно, считаю:" "1, 2, 3, 4, 5, 6, 7, 8, 9, 10." "Конечно, вот:" is slightly better. "Конечно, вот: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10." This is my final choice. Russian language. Polite. Direct. Follows constraints (none violated). I will execute.

Yes, it’s 0.5L of coffee Yes, just for me No, that is for whole day Yes, I use thermos