# rvachev.org > Personal site of Nikita Rvachev β€” a bilingual (EN / RU) editorial blog of essays and notes on AI, technology, and building things. Posts originate from the Telegram channel @rvnikita_blog. Author: Nikita Rvachev (nikita@rvachev.org). English lives under `/en/`, Russian under `/ru/` β€” every page exists in both. There are 1650 posts in total. Every blog post is also available as clean Markdown: request any post URL with the HTTP header `Accept: text/markdown` and the server returns Markdown instead of HTML (same URL). This strips layout and navigation and is far cheaper to read than the HTML page. ## Site - [Home](https://rvachev.org/en): landing page with bio and recent essays - [Essays index](https://rvachev.org/en/blog): all posts, newest first, paginated - [About](https://rvachev.org/en/about): about the author - [Contact](https://rvachev.org/en/contact): contact form - [RSS feed](https://rvachev.org/en/rss.xml): latest posts as RSS - [Sitemap](https://rvachev.org/sitemap.xml): every URL on the site (all 1650 posts Γ— both locales) ## Recent posts - [πŸ“† Lecture 5 β€” Monday, June 1, 10am ET / 5pm MSK](https://rvachev.org/en/blog/lecture-5-monday-june-1-10am-et-5pm-msk): With the course organization at Stanford, there's a bit of a mess: no schedule, but two new recordings on YouTube. For now, we're watching Lecture 5. - [⚑️⚑️⚑️ Anthropic Released Claude Opus 4.8.](https://rvachev.org/en/blog/anthropic-released-claude-opus-4-8): Anthropic released Claude Opus 4.8. The model is less likely to miss bugs in code. Prices are the same, fast mode is cheaper. New: Dynamic Workflows and Effort Control. - [After AlphaGo, the Level of Human Go Players Noticeably Increased](https://rvachev.org/en/blog/after-alphago-the-level-of-human-go-players-noticeably-incre): After AlphaGo, human Go players improved by learning from the model's moves. - [After AlphaGo, the Level of Go Play Increased](https://rvachev.org/en/blog/after-alphago-the-level-of-go-play-increased): After AlphaGo, the level of human play in Go noticeably increased - they learned from the model's moves. - [Following the Discussion of the 4th Lecture](https://rvachev.org/en/blog/following-the-discussion-of-the-4th-lecture): Discussion of the 4th lecture revealed how much more there is to learn in software development. - [Held the Fourth Lecture Club Call on Stanford MS&E 435](https://rvachev.org/en/blog/held-the-fourth-lecture-club-call-on-stanford-ms-e-435): Held the fourth Lecture Club call on Stanford MS&E 435: Economics of the AI Supercycle. - [πŸ€” Looking at the Results of the New Agentic Coding Benchmark DeepSWE](https://rvachev.org/en/blog/looking-at-the-results-of-the-new-agentic-coding-benchmark-d): πŸ€” Looking at the results of the new agentic coding benchmark DeepSWE, it becomes clear that the story around Anthropic Mythos was just marketing. - [I've Been Involved in Technology All My Life](https://rvachev.org/en/blog/i-ve-been-involved-in-technology-all-my-life): I've been involved in technology all my life. I believe it's one of the ways humanity moves forward. - [Vlad Kuklev's Channel](https://rvachev.org/en/blog/vlad-kuklev-s-channel): Today I recommend Vlad Kuklev's channel. Vlad shares cases on applying Claude Code in business. - [#friday #fun](https://rvachev.org/en/blog/friday-fun-1762): #friday #fun - [SpaceX Releases S-1 Ahead of IPO](https://rvachev.org/en/blog/spacex-publishes-s-1-ahead-of-ipo): SpaceX has published S-1 ahead of its IPO with a valuation of $1.75 trillion. Interesting facts about revenue, Bitcoin, and share structure. - [Analysis of AI Economy with Gavin Baker](https://rvachev.org/en/blog/analysis-of-ai-economy-with-gavin-baker): Gavin Baker shares a unique perspective on the AI economy and its dynamics. - [Telegram Now Allows You to Connect a Bot to Your Account](https://rvachev.org/en/blog/telegram-now-allows-you-to-connect-a-bot-to-your-account): Telegram now allows you to connect a bot to your account so it can respond for you in private chats. - [Lecture 4 β€” Economics of the AI Supercycle](https://rvachev.org/en/blog/lecture-4-economics-of-the-ai-supercycle): Lecture 4: Economics of the AI Supercycle β€” Software. Guest: Ali Ghodsi (CEO Databricks). - [AI Lectures: Updated Presentation by Benedict Evans](https://rvachev.org/en/blog/ai-lectures-updated-presentation-by-benedict-evans): Benedict Evans has updated his presentation "AI eats the world" for spring 2026. - [Paying an International Team is Still a Pain](https://rvachev.org/en/blog/paying-an-international-team-is-still-a-pain): Paying an international team is becoming increasingly complex. Let's look at the Stape project that simplifies the process. - [Third Lecture Club Call on Stanford MS&E 435](https://rvachev.org/en/blog/third-lecture-club-call-on-stanford-ms-e-435): Discussion of Chase Lochmiller's lecture on AI data centers and their infrastructure. - [#ai #semiconductors #infrastructure #compute #datacenter #energy #equipment](https://rvachev.org/en/blog/rvachev-s-thoughts-on-semiconductors-and-infrastructure): #ai #semiconductors #infrastructure #compute #datacenter #energy #equipment - [Telegram Introduced 'Guest Bots' for AI Assistants](https://rvachev.org/en/blog/telegram-introduced-guest-bots-for-ai-assistants): Telegram introduced 'guest bots', allowing you to summon an AI assistant in any chat. - [Reminder: Lecture 3 on Monday](https://rvachev.org/en/blog/reminder-lecture-3-on-monday): To prepare for our call, please watch the main video and additional materials in advance. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1751): Today, I want to share some thoughts that have been on my mind lately. - [Untitled](https://rvachev.org/en/blog/post-1750) - [Immortality and Technological Development](https://rvachev.org/en/blog/immortality-and-technological-development): It's interesting to think about how immortality would affect technological development. - [Changes in Using Claude Code](https://rvachev.org/en/blog/changes-in-using-claude-code): Starting June 15, using Claude Code with the -p key will no longer be included in the subscription. - [Second Call of the Lecture Club on Stanford MS&E 435](https://rvachev.org/en/blog/second-call-of-the-lecture-club-on-stanford-ms-e-435): Discussion on the economics of inference and compute as a bottleneck for AI. - [Agent View in Claude Code: Simplifying Work with Agents](https://rvachev.org/en/blog/agent-view-in-claude-code-simplifying-work-with-agents): Anthropic has rolled out the agent view in Claude Code - now all active sessions are in one window. - [Vibecon Conference on Vibe Coding β€” May 13, Free!](https://rvachev.org/en/blog/vibecon-conference-on-vibe-coding-may-13-free): My friends are organizing a cool online conference on vibe coding and agents this Wednesday. - [There is no failure, only trying and learning](https://rvachev.org/en/blog/there-is-no-failure-only-trying-and-learning): "There is no failure. Only try, do, learn, do, learn, do, learn." β€” Chamath Palihapitiya. - [Reminder: Lecture 2 on Monday](https://rvachev.org/en/blog/reminder-lecture-2-on-monday): To prepare for our call, please watch the main video and additional materials in advance. - [How to Save $20 a Month on AI](https://rvachev.org/en/blog/how-to-save-20-a-month-on-ai): It seems I've found a way to save $20 a month on AI. - [Automating Routine Tasks with AI](https://rvachev.org/en/blog/automating-routine-tasks-with-ai): The main idea from Code with Claude: the real power lies in building infrastructure that automates routine tasks. - [Markets and Stanford MS&E 435 Course](https://rvachev.org/en/blog/markets-and-stanford-ms-e-435-course): Today, the markets are raging, and NVIDIA is not in first place for growth. We discuss this in the Stanford MS&E 435 course. - [Online Streaming of Claude with Code in San Francisco](https://rvachev.org/en/blog/online-streaming-of-claude-with-code-in-san-francisco): If you couldn't make it to Claude with Code in San Francisco, watch it online. - [Untitled](https://rvachev.org/en/blog/post-1738) - [OpenAI Launched Advertising Infrastructure for ChatGPT](https://rvachev.org/en/blog/openai-launched-advertising-infrastructure-for-chatgpt): OpenAI launched advertising infrastructure for ChatGPT, which will change performance marketing. - [First Lecture Club Call on Stanford MS&E 435](https://rvachev.org/en/blog/first-lecture-club-call-on-stanford-ms-e-435): The first Lecture Club call on Stanford MS&E 435 gathered about 80 people and sparked a lively discussion. - [Let's Get Started β€” Join Us](https://rvachev.org/en/blog/let-s-get-started-join-us): Let's get started β€” join the lecture for the Stanford MS&E 435 course. - [Rvachev's Thoughts](https://rvachev.org/en/blog/rvachev-s-thoughts-1733): Today, I want to share some thoughts that have been on my mind lately. - [Nvidia Falls After Hyperscaler Reports](https://rvachev.org/en/blog/nvidia-fell-after-hyperscaler-reports): Nvidia fell after reports from Meta, Amazon, Microsoft, and Google, even though capex increased by $15 billion. - [Watching Stanford Lectures: Economics of the AI Supercycle Together](https://rvachev.org/en/blog/watching-stanford-lectures-economics-of-the-ai-supercycle-to): This week we are watching the first Stanford lecture on the economics of the AI Supercycle. - [Stanford Lecture Series: Economics of the AI Supercycle](https://rvachev.org/en/blog/stanford-lecture-series-economics-of-the-ai-supercycle): I found the Stanford lecture series on the economics of the AI supercycle β€” the best thing I've watched in recent months. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1728): Today, I want to share some thoughts that have been on my mind lately. - [Untitled](https://rvachev.org/en/blog/post-1727) - [Comparison of OpenAI ChatGPT 5.5 and Claude Opus 4.7](https://rvachev.org/en/blog/comparison-of-openai-chatgpt-5-5-and-claude-opus-4-7): OpenAI ChatGPT 5.5 is slightly better than Claude Opus 4.7 in Medium mode. - [OpenAI Released ChatGPT 5.5](https://rvachev.org/en/blog/openai-released-chatgpt-5-5): OpenAI has released ChatGPT 5.5, focusing on multi-step tasks and improvements in coding. - [How to Invest in Anthropic if You're Not an 'Accredited Investor'](https://rvachev.org/en/blog/how-to-invest-in-anthropic-if-you-re-not-an-accredited-inves): How to invest in Anthropic if you're not an 'accredited investor'. - [Sadness with Image Generation](https://rvachev.org/en/blog/sadness-with-image-generation): It's sad that image generators can't create an image of a person on a horse. - [ChatGPT Images 2.0: A New Generation of Image Generation](https://rvachev.org/en/blog/chatgpt-images-2-0-a-new-generation-of-image-generation): OpenAI has released ChatGPT Images 2.0 - a new generation of image generation with improved accuracy and control. - [OpenAI Strikes Back with New Product Announcements](https://rvachev.org/en/blog/openai-strikes-back-with-new-product-announcements): This week, OpenAI announces new products, including GPT-Image-2. - [Codex Learned to See What's Happening on Your Screen](https://rvachev.org/en/blog/codex-learned-to-see-what-s-happening-on-your-screen): Codex now takes screenshots of your desktop and remembers what you’re working on. - [Anthropic Launches Claude Design - A New Design Tool](https://rvachev.org/en/blog/anthropic-launches-claude-design-a-new-design-tool): Anthropic announced Claude Design - a tool for creating visual materials and a competitor to Figma. - [Rvachev's Thoughts](https://rvachev.org/en/blog/rvachev-s-thoughts-1717): Today, I want to share some thoughts that have been on my mind lately. - [The Speed of Sound and the Crack of a Whip](https://rvachev.org/en/blog/the-speed-of-sound-and-the-crack-of-a-whip): The tip of a whip, when cracked, breaks the sound barrier, creating a mini-sonic boom. - [Scheduled Tasks in Claude Code](https://rvachev.org/en/blog/scheduled-tasks-in-claude-code): Claude Code introduced Scheduled Tasks, which Claude uses for self-checking. - [Anthropic Released Opus 4.7](https://rvachev.org/en/blog/anthropic-released-opus-4-7): The first thing to do is to ask the new model to check all the code. - [Claude Code Desktop: A New Era of Parallel Work](https://rvachev.org/en/blog/claude-code-desktop-anthropic-s-new-app): Claude Code Desktop now supports parallel work with multiple agents simultaneously. - [Understanding LLM and AI through Stanford's Course](https://rvachev.org/en/blog/understanding-llm-and-ai-through-stanford-s-course): I recommend 10 lectures from Stanford's CS230 Deep Learning course to understand LLMs and artificial intelligence. - [SaaS Companies Lose 30-86% of Capitalization](https://rvachev.org/en/blog/saas-companies-lose-valuation-in-the-age-of-ai): SaaS companies are losing 30-86% of their capitalization - the market is overvaluing software in the AI era. - [Channels I Read: R-Founders](https://rvachev.org/en/blog/channels-i-read-r-founders): The R-Founders channel shares practical content for founders in international markets. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1709): Today, I want to share some thoughts that have been on my mind lately. - [Artemis 2 Returns from Lunar Flyby](https://rvachev.org/en/blog/artemis-2-returns-from-lunar-flyby): Artemis 2 returns from its lunar flyby. The most dangerous part of the mission is ahead. - [Waymo Released Safety Statistics](https://rvachev.org/en/blog/waymo-s-safety-statistics-compared-to-human-drivers): Waymo has published safety statistics of its vehicles compared to human drivers. - [Do Technologies Make Us Dumber? Data Says Otherwise.](https://rvachev.org/en/blog/do-technologies-make-us-dumber-data-says-otherwise): Calculators and GPS are thought to make us dumber, but data shows the opposite. - [Testing Paperclip - an AI Agent Management System](https://rvachev.org/en/blog/testing-paperclip-an-ai-agent-management-system): Testing Paperclip - an AI agent management system that resembles JIRA. - [AI Adoption by the Numbers by a16z](https://rvachev.org/en/blog/ai-adoption-by-the-numbers-by-a16z): 29% of Fortune 500 companies are already paying clients of AI startups. Coding is the absolute leader in adoption. - [A Secret Directive for Managers](https://rvachev.org/en/blog/a-secret-directive-for-managers): In my managerial youth, we developed a secret directive for task assignment. - [Anthropic Launches Project Glasswing](https://rvachev.org/en/blog/anthropic-launches-project-glasswing-for-software-protection): Anthropic has launched Project Glasswing to protect software using the Claude Mythos Preview model. - [Interesting Facts About Space Flights](https://rvachev.org/en/blog/interesting-facts-about-space-flights): Interesting facts about space flights, including toilet waste and the Earth's rotation. - [Wave Your Hand!](https://rvachev.org/en/blog/wave-your-hand): Wave your hand! - [A Cool Test for AGI](https://rvachev.org/en/blog/cool-test-for-agi): A cool test for AGI is when an agent can generate more electricity than it consumes for the task. - [Value Shifts from Code to Ideas](https://rvachev.org/en/blog/value-shifts-from-code-to-ideas): Value shifts from code to ideas, and concepts become more important than code. - [Recommendation for the Game Kerbal Space Program](https://rvachev.org/en/blog/recommendation-for-the-game-kerbal-space-program): I recommend the game Kerbal Space Program - a realistic simulator for designing rockets and launching them into space. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1694): Today, I want to share some thoughts that have been on my mind lately. - [Artemis II - Humans Are Flying to the Moon Again](https://rvachev.org/en/blog/artemis-ii-humans-are-flying-to-the-moon-again): The main news of the past few days has been the launch of the rocket to the Moon. This flight sets a record for the farthest point from Earth that a human has ever been. - [Closing the Blog and a New Beginning](https://rvachev.org/en/blog/closing-the-blog-and-a-new-beginning): I have two pieces of news - the good: I'm joining Anthropic, and the bad: the blog is closing. - [Leak of Claude Code Source Code via npm](https://rvachev.org/en/blog/leak-of-claude-code-source-code-via-npm): Today, Chaofan Shou discovered that the npm package @anthropic-ai/claude-code contained the complete source code of the project. - [Untitled](https://rvachev.org/en/blog/post-1690) - [PreText Typography: The Beauty of Typography in the Web](https://rvachev.org/en/blog/pretext-typography-the-beauty-of-typography-in-the-web): On weekends, everyone goes crazy for PreText Typography by Cheng Lou. - [Anthropic Quietly Rolls Out Next-Gen Model β€” Claude Mythos](https://rvachev.org/en/blog/anthropic-quietly-rolls-out-next-gen-model-claude-mythos): Anthropic has quietly started rolling out Claude Mythos, a next-generation model, in stealth mode. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1687): Today I want to share some thoughts on the importance of taking breaks and enjoying life outside of work. - [ARC-AGI-3: a test for humans and neural networks](https://rvachev.org/en/blog/arc-agi-3-a-test-for-humans-and-neural-networks): The ARC Prize introduced the third version of the benchmark, where humans easily solve tasks, but neural networks do not. - [How to Make Claude Code for Hours Unsupervised](https://rvachev.org/en/blog/how-to-make-claude-code-for-hours-unsupervised): Anthropic shared how to split the work among three agents for coding. - [Auto mode in Claude Code: run tasks and take a break](https://rvachev.org/en/blog/auto-mode-in-claude-code-run-tasks-and-take-a-break): You can now run tasks in Claude Code and not worry about safety. - [Jensen Huang on Achieving AGI in Lex Fridman Podcast](https://rvachev.org/en/blog/jensen-huang-on-achieving-agi-in-lex-fridman-podcast): Jensen Huang stated in the Lex Fridman podcast that AGI has already been achieved. He discussed how AI can build companies and its impact on jobs. - [Claude Code Learned to Dream](https://rvachev.org/en/blog/claude-code-learned-to-dream): The Claude Code community is discussing a new feature β€” Auto Dream, which tidies up the agent's memory. - [Figma Opens Canvas for AI Agents](https://rvachev.org/en/blog/figma-opens-canvas-for-ai-agents): Figma launched an open beta of the new MCP tool use_figma, allowing AI agents to draw designs on the canvas. - [Claude Learned to Use a Computer](https://rvachev.org/en/blog/claude-learned-to-use-a-computer): Claude can now open files and manage the browser like a human. - [Anthropic Launched a Science Blog](https://rvachev.org/en/blog/anthropic-launched-science-blog): Anthropic launched a science blog featuring two strong publications about Claude Opus. - [Andrej Karpathy: Remove Yourself from the Loop](https://rvachev.org/en/blog/andrej-karpathy-remove-yourself-from-the-loop): An interview with Andrej Karpathy about removing yourself from the loop for maximum efficiency. - [A Startup Unicorn Idea Born from AI Discussions](https://rvachev.org/en/blog/startup-idea-ai-transformation): During a conversation about AI, a startup unicorn idea was born. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1675): Today I want to share some thoughts that have been on my mind lately. - [Thoughts of Rvachev](https://rvachev.org/en/blog/thoughts-of-rvachev-1676): Today, I want to share thoughts on the importance of taking breaks and enjoying the little things. - [Claude Code Now Managed via Telegram](https://rvachev.org/en/blog/claude-code-now-managed-via-telegram): Anthropic has released Claude Code Channels - now you can manage your Claude Code session directly from Telegram and Discord. - [Expectation vs Reality](https://rvachev.org/en/blog/expectation-vs-reality): Expectation: AGI, robots do all the work, we relax. Reality: DoorDash launches Dasher Tasks. - [Pascal's Wager for the Age of AI](https://rvachev.org/en/blog/pascal-s-wager-for-the-age-of-ai): Pascal's wager for your career: work more to keep up in the age of AI. - [Telegram as an Unexpected Beneficiary of OpenClaw](https://rvachev.org/en/blog/telegram-as-an-unexpected-beneficiary-of-openclaw): One of the beneficiaries of the emergence of OpenClaw has been Telegram for the foreign audience. - [Incident with AI Agent at Meta](https://rvachev.org/en/blog/incident-with-ai-agent-at-meta): At Meta, an AI agent went out of control and triggered a serious security incident. - [Claude Conference for Developers](https://rvachev.org/en/blog/claude-conference-for-developers): Anthropic has announced the dates and cities for the Claude conference for developers. - [Why People Have 10x Productivity with AI, but Companies Don't](https://rvachev.org/en/blog/why-people-have-10x-productivity-with-ai-but-companies-don-t): a16z wrote an article about the difference between Individual AI and Institutional AI. Each employee has become faster, but the artifacts are not connected into a single stream of decisions. - [Percepta Embedded a Computer Right Inside the Transformer](https://rvachev.org/en/blog/percepta-embedded-a-computer-right-inside-the-transformer): The Percepta team embedded a virtual machine into the model's weights, enabling it to solve complex tasks. ## Optional - [Full archive](https://rvachev.org/sitemap.xml): the 1650 posts above are the most recent; fetch the sitemap for the complete list, then request any post with `Accept: text/markdown` for Markdown. - [Russian edition](https://rvachev.org/ru): the same content in Russian under `/ru/...`.