Rudraprasad Das

@aakrit Is this consumption in the product or by tools used to build the product?

India's AI Stack 🇮🇳 We surveyed 244 AI startup CTOs and developers in India to understand what is actually being used in production today. The survey focused on deployed tools, what is currently in their stack, LLM choices, and how their preferences are changing. 🧵

@alexwikstrom Well, the Minecraft modding community is pretty big and active. I am not surprised!

Nobody: Not a single soul: YC: Lovable for Minecraft

@GithubProjects TypeError: Cannot read properties of undefined

Scare a DEVELOPER with ONE word.

@LoicBerthelot But why? What's your reason for not using them?

Software I never use as a SaaS founder... - n8n → i'm an old make power user. - GitHub → Still gitlab - Clerk → Before oauth0 and now supabase. Which softwares do you avoid?

@codewithDiyaa Did you try @kilocode ?

Any cursor alternative?

@abhitwt I have been trying to connect with the tech community on X for so long. But nothing :(

Twitter is cool. But it’s 100x better when you connect with people who code. If you’re into tech, AI, or programming, say hi

Are you even vibe coding? What's your flex?

@auralix4 What did you do?

@bradcarryvc So, you ruined a kid's life with you advice?

Met with a young founder He’s in college and wants to build a startup Told him that Zuck, Gates, and Jobs all dropped out of college He needs to do the same if he’s serious about being successful Dropped out that day 6 months later, his startup failed, his parents disowned him, and he still has $60k in student loans with no degree He called me crying that he’s now homeless with -$500 in his bank account I deleted his contact info Lesson: not everyone is destined for greatness!

@ArmanHezarkhani Thanks for the guide, but are you using claude cowork?

@ericw_ai 60 pages! I will probably ask my agents to summarise it 🤓

@aakashgupta this unlocks scale, but discovery wasn’t the real constraint to begin with

Anthropic just solved MCP’s biggest lie: that 50+ tools means 50+ capabilities. The math tells a different story. Users with 7 MCP servers were burning 67k tokens before typing a single character. That’s 33% of Sonnet’s 200k context window gone. One Docker MCP alone consumed 125k tokens with 135 tools. Tool Search flips the architecture. Instead of preloading every tool definition upfront, Claude Code now loads a search index and pulls tools on-demand. The token savings are dramatic: from ~134k to ~5k in Anthropic’s internal testing. That’s an 85% reduction while maintaining full tool access. But here’s what makes this actually interesting. They’re not just optimizing context usage. They’re changing what “tool-rich agents” can mean. The old constraint forced a brutal tradeoff: either limit your MCP servers to 2-3 core tools, or accept that half your context budget disappears before you start working. Power users documented setups with 7+ servers consuming 67k+ tokens. The GitHub issue tracking this had 97 upvotes. People were building CLI tools just to toggle servers on and off between sessions. Tool Search reframes the problem. The bottleneck wasn’t “too many tools.” It was loading tool definitions like 2020-era static imports instead of 2024-era lazy loading. Every IDE figured this out years ago. VSCode doesn’t load every extension at startup. JetBrains doesn’t inject every plugin’s docs into memory. Claude Code was architecturally behind its own ecosystem’s best practices. The accuracy improvement matters more than the token savings. Opus 4 went from 49% to 74% on MCP evals with Tool Search enabled. Opus 4.5 from 79.5% to 88.1%. Fewer tools in context means less confusion between similar tool names like notification-send-user vs notification-send-channel. This is Anthropic doing infrastructure work that doesn’t generate headlines but compounds over time. The MCP ecosystem just got permission to build bigger.

@alexgilev @AgenticUi Clean interface, but it assumes agent behavior is stable enough to score. That's rarely true in practice

Agent dashboard 🤖 w/ @AgenticUi

@LangChain Everyone's building the "cowork." Very few are building agents you can actually trust with real business workflows.

DeepAgents is a general purpose agent harness, which makes it quite easy to build a general purpose “open cowork”

@meta_alchemist @meta_alchemist Interesting tool. Though this feels less like models lying and more like security defaults being misunderstood.

A big issue with vibe coding LLMs like Claude, Gemini, GPT, etc: They all suck at Supabase security. They say everything is secure when it's not. This will lead to many hacks. To counter this, I made a penetration tester using Ralph Wiggum. -> free at suparalph.vibeship .co

@sydneyrunkle @LangChain @sydneyrunkle Nice loop. How do you handle drift when yesterday’s top performing formats stop working?

🤖 I built a content creation agent using LangSmith Agent Builder from @LangChain 📚 I feed it context from our docs and GitHub repos. ✏️ It drafts tweets & LinkedIn posts about new features, releases, and content (blogs, videos). ✨ The best part: I retrain it with past posts + performance data — so it learns what resonates with my audience over time! 🔗 Try Agent Builder: langchain.com/langsmith/agen…

Most people think AI productivity comes from better prompts. It actually comes from better defaults. If your system needs constant human nudging, it’s not intelligent yet.

Speed doesn’t come from moving faster. It comes from removing the parts of your day that don’t need a human.

Most AI tools automate tasks. Very few change how decisions get made.

@1Umairshaikh Hi! I am Rudra! Building and learning in public. Currently, building a multi-agent orchestration system where agents can collaborate and get real work done - starships.ai Happy to connect 👋

Founders: It’s Friday. Network day. Startups grow faster with the right people around you. Introduce yourself and network 🤝