Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 18/03/2025MVP - Maximum Vibe Product 0111
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 07/04/2025Important change with AI, that it is ok to re-roll the dice to wipe all previous work, and re-do it with updated prompt, rather then trying to fix whats already done. The work itself becomes very cheap, the most expensive part here is capacity of human to review and think. 051
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 17/04/2025The recent agent-to-agent protocol more than just a technological solution, it is attempt to "fix" colapsing SaaS economy and API market, and moving to value based pricing. 151
Leonid Bugaev @leonidbugaev.bsky.social · 17/04/2025The recent agent-to-agent protocol more than just a technological solution, it is attempt to "fix" colapsing SaaS economy and API market, and moving to value based pricing. 151
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 08/04/2025Now you can chat with documentation or code, via MCP, just by specifying github url - all search happens locally! More over you can create your own MCP servers, with pre-baked data! The easiest way to create MCP so far! All free and Open Source. github.com/buger/docs-mcpgithub.comGitHub - buger/docs-mcpContribute to buger/docs-mcp development by creating an account on GitHub. 051
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 08/04/2025Chatting with AI changed the "search" paradigm - simplifying it to Question -> Answer format. No wonder it become so addictive. MCP changes the game again, bringing it local and adding action to the flow: Question -> Answer -> Action Don’t underestimate its power. 031
Leonid Bugaev @leonidbugaev.bsky.social · 08/04/2025Chatting with AI changed the "search" paradigm - simplifying it to Question -> Answer format. No wonder it become so addictive. MCP changes the game again, bringing it local and adding action to the flow: Question -> Answer -> Action Don’t underestimate its power. 031
Leonid Bugaev @leonidbugaev.bsky.social · 08/04/2025Now you can chat with documentation or code, via MCP, just by specifying github url - all search happens locally! More over you can create your own MCP servers, with pre-baked data! The easiest way to create MCP so far! All free and Open Source. github.com/buger/docs-mcpgithub.comGitHub - buger/docs-mcpContribute to buger/docs-mcp development by creating an account on GitHub. 051
Leonid Bugaev @leonidbugaev.bsky.social · 07/04/2025Important change with AI, that it is ok to re-roll the dice to wipe all previous work, and re-do it with updated prompt, rather then trying to fix whats already done. The work itself becomes very cheap, the most expensive part here is capacity of human to review and think. 051
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 02/04/2025Recently started spending more time working outside. Living in a sunny place, it’s almost impossible to use a laptop outdoors. I’ve been using these Xreal One glasses - like portable monitors, not VR. They look a bit odd on calls, but honestly, it’s working really well for me. 031
Leonid Bugaev @leonidbugaev.bsky.social · 02/04/2025Recently started spending more time working outside. Living in a sunny place, it’s almost impossible to use a laptop outdoors. I’ve been using these Xreal One glasses - like portable monitors, not VR. They look a bit odd on calls, but honestly, it’s working really well for me. 031
Leonid Bugaev @leonidbugaev.bsky.social · 17/03/2025It's been over two years since the new cursor was released. We now have juniors who've only programmed with AI, never experiencing coding without it. Would you hire these juniors today? Do you think they can become capable engineers? 230
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 15/03/2025Have become top 3 trending developer on Github, with @mitchellh right behind me 😱 151
Leonid Bugaev @leonidbugaev.bsky.social · 16/03/2025Some future AI revolutions happen in some random discord discussions like this discord.com/channels/133...discord.comDiscord - Group Chat That’s All Fun & GamesDiscord is great for playing games and chilling with friends, or even building a worldwide community. Customize your own space to talk, play, and hang out. 020
Leonid Bugaev @leonidbugaev.bsky.social · 15/03/2025Have become top 3 trending developer on Github, with @mitchellh right behind me 😱 151
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 11/03/2025Why artifact storage products are so expensive, like PackageCloud or CloudSmith. We pay a LOT for it now. Essentially it is a file storage with some layer on top of it... However configuring own deb/rpm repo also a pain. I wonder if it is a solved problem? What do you use? 021
Leonid Bugaev @leonidbugaev.bsky.social · 11/03/2025Why artifact storage products are so expensive, like PackageCloud or CloudSmith. We pay a LOT for it now. Essentially it is a file storage with some layer on top of it... However configuring own deb/rpm repo also a pain. I wonder if it is a solved problem? What do you use? 021
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 10/03/2025It's always tempting to chase the hype—another flashy AI code editor, another trendy framework. But true innovation often hides in the quieter spaces, the "plumbing." These foundational tools quietly enable everything else to run smoothly and grow. 🧵 241
Leonid Bugaev @leonidbugaev.bsky.social · 10/03/2025It's always tempting to chase the hype—another flashy AI code editor, another trendy framework. But true innovation often hides in the quieter spaces, the "plumbing." These foundational tools quietly enable everything else to run smoothly and grow. 🧵 241
Leonid Bugaev @leonidbugaev.bsky.social · 10/03/2025A lot of problems can be solved by bruteforcing them. Imagine if we had Claude 3.5 10x faster and 10x cheaper, or 100x. Now imagine it is Grok 3. Our current models already good enough, for example see Manus. AGI is so close! 120
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 07/03/2025The more I get into the Rust and its ecosystem the more I think it can be one of the dominant languages in the AI era. Claude 3.7 knows it very well, built-in tooling is amazing and integrate very well with AI (like very detailed error messages). Type safe and super fast. 151
Leonid Bugaev @leonidbugaev.bsky.social · 07/03/2025The more I get into the Rust and its ecosystem the more I think it can be one of the dominant languages in the AI era. Claude 3.7 knows it very well, built-in tooling is amazing and integrate very well with AI (like very detailed error messages). Type safe and super fast. 151
Leonid Bugaev @leonidbugaev.bsky.social · 05/03/2025Its quite facinating that I actually stopped writing the code - I just think and review. 020
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 27/02/2025Rolling back to Claude Sonnet 3.5. The new 3.7 is too unpredictable, constanlty goes to loops in agentic mode. It defo better understand code (like knowing Rust much better), but worse at following instructions and history of conversation. 142
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 26/02/2025It’s fascinating that well-known engineering best practices—like domain-driven design, separation of concerns, unit testing, low cyclomatic complexity, detailed architecture docs, feature specs, and descriptive file names—are exactly what enable AI to scale in large codebases. 041
Leonid Bugaev @leonidbugaev.bsky.social · 27/02/2025Rolling back to Claude Sonnet 3.5. The new 3.7 is too unpredictable, constanlty goes to loops in agentic mode. It defo better understand code (like knowing Rust much better), but worse at following instructions and history of conversation. 142
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025It’s fascinating that well-known engineering best practices—like domain-driven design, separation of concerns, unit testing, low cyclomatic complexity, detailed architecture docs, feature specs, and descriptive file names—are exactly what enable AI to scale in large codebases. 041
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Just added internal Chat bot, which acts as internal R&D consultant, which helps you to shape R&D claims. It probably can help us shave $1M a year. Build boring stuff. 020
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Reliable chat with a code is even more important then actually writing the code. It helps to architect before build (especially with AI), generate test cases, debug issues, and way more. And it can be used by everyone and not just engineers: QA, Docs team, Support, Sales end etc. 010
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025We made it mandatory to use internal chat with our documentation, and some internal data sources, before going to engineers. This way we easily can find documentation gaps. AI learns based on our docs, if it does not know the answer - find it elsewhere and improve the docs. 020
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Grok 3 thinking mode has become my go to Architect model, and replaced O1 Pro for this case. O1 Pro still smarter, but it is very slow, and Grok 3 is smart enough to be an exelent architect, and I do not need accurate code generation out of it - Claude doing all the job there. 010
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Everyone obsesses over growing revenue and MRR, which is fair, but "boring" actions can save you even more money. Tax optimization and R&D claims can significantly reduce your taxable amount. A quality R&D claim might actually be the key to profitability. 130
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Rust complexity is fixed by Sonnet 3.7, and to be frank I'm quite amazed with the quality of Rust tooling and how AI ready it is - and thats where a lot of languages should learn from it. All the compiler errors are super descriptive, and actionable - exactly what AI needs. 150
Leonid Bugaev @leonidbugaev.bsky.social · 26/02/2025Sonnet 3.7 is quite good at Rust! Which means its time build some cool extreme performance stuff. 020
Leonid Bugaev @leonidbugaev.bsky.social · 21/02/2025Seriously, I do not know how Cusor, Windsurf, Lovable and other AI products survive. It feels like a 🔥💸. I reached $800/m in costs for Claude 3.5, while using it with Cline (e.g. bring your own API token). And it is with prompt caching! Thats active daily agentic work. 040
Leonid Bugaev @leonidbugaev.bsky.social · 21/02/2025I was doing just fine with my M1 Mac for a few years already, but with all this AI stuff, I urgently need an upgrade. And I'm not even talking about running LLMs locally. 4 different Agents running in IDEs in parallel, crunching the data, linters, compilations, its too much 😅 000
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 20/02/2025This is how debugging with MCP and Agent in Cline looks like: Using this curl call, and SQLite MCP tool you going to help me debug analytics in my app curl -X POST .... The first goal will be "Chat Interactions" chart on Dashboard.js, which should increment each call. 😍 011
Leonid Bugaev @leonidbugaev.bsky.social · 20/02/2025This is how debugging with MCP and Agent in Cline looks like: Using this curl call, and SQLite MCP tool you going to help me debug analytics in my app curl -X POST .... The first goal will be "Chat Interactions" chart on Dashboard.js, which should increment each call. 😍 011
Leonid Bugaev @leonidbugaev.bsky.social · 18/02/2025"Find which zedndesk ticket blocking PoC of the prospect <x>, and for each give me detailed summary and actions. Do not forget to check attached Jira tickets. Also check the latest call notes from hubspot notes" We are very close to reliably achive this internally 😍 010
Leonid Bugaev @leonidbugaev.bsky.social · 18/02/2025With "bring your own" api key for AI code editors (Cline), with agentic mode, it costs at least $300/m per developer, while using Claude Sonnet 3.5, assuming active usage. From the company expense level it's still not much, but I can imagine "profitability" of tools like Cursor 010
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 13/02/2025If MCP protocol from Anthropic succeed, it will democratize the connector market, and the ones who claimed that they have "100+s" connectors will loose their moat. Everyone will get access to everything. Can't wait for its server support with Auth! 0104
Leonid Bugaev @leonidbugaev.bsky.social · 14/02/2025Just build MCP tool to work with local SQLite db. Now while working on features I can ask to populate some data, or while fixing a bug make agent dig to db level. All build with Cline, which has built in optimized flow to build own MCP tools. 000
Leonid Bugaev @leonidbugaev.bsky.social · 13/02/2025MCP tools with IDE integration, like Cline or Cursor is game changer. I just implemented a "Big Brain" tool, which makes Cline to offload complex task to the "Big Brain", which actually generates RepoPrompt compatible request, sends to O1 Pro, and which do the actual work! 🔥 010
Leonid Bugaev @leonidbugaev.bsky.social · 13/02/2025If MCP protocol from Anthropic succeed, it will democratize the connector market, and the ones who claimed that they have "100+s" connectors will loose their moat. Everyone will get access to everything. Can't wait for its server support with Auth! 0104
Reposted by Leonid BugaevLeonid Bugaev @leonidbugaev.bsky.social · 12/02/2025If you’re getting something for free it means that you’re the product. 151
Leonid Bugaev @leonidbugaev.bsky.social · 12/02/2025If you’re getting something for free it means that you’re the product. 151
Leonid Bugaev @leonidbugaev.bsky.social · 07/02/2025Cline can build ITS OWN MCP tools. Just just by telling in chat like "Build me Jira tool". It just implemented it in one-shot and already connected to my org Jira. And now I can just do this: "Use JIRA tool to find info about TT-14013, and implement it" 🤯 120
Leonid Bugaev @leonidbugaev.bsky.social · 07/02/2025Forget about Cursor and Windsurf. Cline is absolutelly killing it! Open source free VS Code extension, with bring your own API token flow. It just one-shotted feature, in complex codebase, modifying 20 different files, and in process fixing tests. 🚀 230