Tirth Kanani

🧵 I got tired of Claude Code re-reading my entire codebase on every single task. So I built code-review-graph: a persistent structural map of your code that cuts token usage by 6.8x on average. 📉 49x reduction on monorepos. 🟢 Open source. Local-only. Zero cloud. 👉 github.com/tirth8205/code… Here's how it works 👇

Open call. I need one person to build something with me. Not a co-founder, not staff. Someone who reads the footnotes and builds by feel. Anywhere in the world. The rest is here: tirthkanani.com/one

you can outsource your thinking but you cannot outsource your understanding

Sovereign AI. Launching Tonight. 🇬🇧

AI agents can now pay, trade, and sign transactions autonomously. The one thing they still can't do: earn yield on the capital sitting in their wallets. 400,000 agents. Millions in idle float. Zero yield. We built the infrastructure to fix that. 🧵

Built terminal games you play while Claude Code works on your tasks. Claude Runner has the sparkle mascot dodging Segfault, npm ERR!, and git merge conflicts. Snake is a GitHub contribution graph edition where the food is mini Claude logos. There are hidden easter eggs at certain scores that I won't spoil. The status bar connects to Claude Code via hooks; green pulse whilst it's working, yellow when it needs your input, blue when it's done. Auto-pauses your game so you don't miss it. Zero tokens. Zero API calls. Pure terminal rendering. Github: github.com/tirth8205/clau… Inspired by claude-arcade (claude-arcade.lovable.app) cc @AnthropicAI @bcherny

Microsoft employees watching Anthropic ship better AI features for office than Copilot.

Claude Code reads 27,000 files just to review a 3-line change. code-review-graph v2.1.0 fixes that — builds a persistent structural graph so your AI only reads what actually matters. → 8× fewer tokens → 49× on monorepos What’s new in v2.1.0: 🟢 Jupyter & Databricks notebook parsing (finally!) 🟢 20 languages supported (Lua added) 🟢 Zero-config install - one command, 7 platforms 🟢 JS/TS arrow function indexing fix Website: code-review-graph.com Discord: discord.gg/3p58KXqGFN pip install --upgrade code-review-graph Repost if this is useful to your dev workflow 🔥 #OpenSource #DevTools #AI #ClaudeCode #CodeReview #MCP #Python #Jupyter #Notebooks #Lua #TreeSitter #LLM #BuildInPublic #DeveloperExperience #AICoding

🙏 Thank you to everyone who has contributed to code-review-graph. Every issue filed, every pull request opened, every bit of feedback has made this tool better. Genuinely grateful for the community growing around it. 🚀 v1.8.4 is now live on PyPI. 🟢 Vue SFC + Solidity parsing (14 languages now) 🟢 Call targets auto-resolve to qualified names 🟢 Multi-word AND search for precise results 🟢 Blast radius pagination for large codebases 🟢 New find_large_functions MCP tool 🟢 Security hardening across the board 🟢 VS Code extension v0.2.0 with all 13 commands pip install --upgrade code-review-graph 🔗 github.com/tirth8205/code… ♻️ Repost if this is useful to your dev workflow #OpenSource #DevTools #AI #ClaudeCode #CodeReview #MCP #Python #VSCode #TreeSitter #LLM #BuildInPublic #DeveloperExperience

We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: anthropic.com/features/81k-i…

Claude Code was burning 739K tokens to review a change in Next.js. I built a knowledge graph that cuts it to 15K. Tree-sitter parsing → blast-radius analysis → only the files that matter get read. 6.8x avg token reduction. 49x on monorepos. Review quality went UP. Open source 👇 tirthkanani18.medium.com/i-built-a-know…

LabOS today goes into LabClaw🦞 Today we embedded Stanford LabOS AI-XR co-scientist into LabClaw with OpenClaw🦞 — turning it into the unified skill layer for biomedical AI agents and world model. No more separate systems. One install. Full Wet-lab + Dry-lab execution with XR. labclaw-ai.github.io #LabClaw #Claw #LabOS #OpenClaw #AIforScience 🦞 #Stanford #Princeton

But here’s what really surprised me: It’s not just about saving tokens. LLMs suffer from the “lost in the middle” effect. When context gets long, important details buried in the middle often get ignored, even if they’re critical. On a 27k-file repo, you’re not just burning tokens. You’re actively making Claude worse at its job. The graph changes that completely. It gives Claude persistent memory of your entire codebase that survives every new session. No more starting from scratch. No more guessing. That’s why the quality jump feels bigger than the token savings. 🔥 What’s the biggest codebase you’ve ever thrown at Claude? Drop it below 👇

⚙️ Setup takes 30 seconds: pip install code-review-graph code-review-graph install Or: claude plugin add tirth8205/code-review-graph 🌐 Works with Python, TS, Go, Rust, Java, C#, Ruby, Kotlin, Swift, PHP, C/C++ 📜 MIT licence ⭐ Star it, try it, break it: 👉 github.com/tirth8205/code… #ClaudeCode #OpenSource #DevTools #AI #DeveloperTools

🧵 I got tired of Claude Code re-reading my entire codebase on every single task. So I built code-review-graph: a persistent structural map of your code that cuts token usage by 6.8x on average. 📉 49x reduction on monorepos. 🟢 Open source. Local-only. Zero cloud. 👉 github.com/tirth8205/code… Here's how it works 👇