https://suryansh.work Build-in-public logs, MacGet, and production AI systems. en-us https://suryansh.work/blog/mcp-servers-vs-agent-skills-tool-payload-design https://suryansh.work/blog/mcp-servers-vs-agent-skills-tool-payload-design Shipping an MCP server is the easy half. The work that decides whether an agent can actually drive your product is payload design — closed vocabularies, schema-constrained returns, and tool descriptions written for a reader who will not ask a follow-up question. Mon, 03 Aug 2026 00:00:00 GMT https://suryansh.work/blog/macget-1-3-0-bittorrent-book-catalogs-adaptive-concurrency https://suryansh.work/blog/macget-1-3-0-bittorrent-book-catalogs-adaptive-concurrency MacGet 1.3.0 adds BitTorrent and OPDS book catalogs to the macOS download manager — and rips out the adaptive connection ramp I shipped as a headline feature in 1.1.0, because the heuristic read a saturated link as a failed experiment. Thu, 30 Jul 2026 00:00:00 GMT https://suryansh.work/blog/automating-blog-cover-art-gemini-3-pro-image https://suryansh.work/blog/automating-blog-cover-art-gemini-3-pro-image I wrote a Node script to generate whiteboard-sketch covers for every post on this blog, and banned the model from drawing readable text. Nano Banana Pro ignored the ban, spelled everything correctly, and proved the constraint was a leftover assumption from 2023. Sun, 26 Jul 2026 00:00:00 GMT https://suryansh.work/blog/claude-opus-5-vs-fable-5-routing-langgraph-agents https://suryansh.work/blog/claude-opus-5-vs-fable-5-routing-langgraph-agents Opus 5 landed at half Fable 5's price and claims to approach it on most tasks. For a multi-agent research desk the real question isn't which model wins — it's which node gets which, and the two API changes that will silently break your existing graph. Sun, 26 Jul 2026 00:00:00 GMT https://suryansh.work/blog/free-build-log-blog-nextjs-portfolio https://suryansh.work/blog/free-build-log-blog-nextjs-portfolio How I added a full build-log blog to my portfolio — likes, email subscribers, and new-post notifications — without paying for anything. The stack, the code, and the one Vercel Hobby limit that broke my cron. Wed, 22 Jul 2026 00:00:00 GMT https://suryansh.work/blog/how-we-architect-a-financial-ai-analyst-multi-agent https://suryansh.work/blog/how-we-architect-a-financial-ai-analyst-multi-agent The system behind an AI that reads filings and answers like an analyst — specialized agents on LangGraph over a Vertex multimodal RAG data plane. What the architecture looks like, and the four things that broke on the way there. Thu, 16 Jul 2026 00:00:00 GMT https://suryansh.work/blog/langgraph-prompt-injection-hard-soft-gates https://suryansh.work/blog/langgraph-prompt-injection-hard-soft-gates Prompt injection isn't one problem, so it doesn't get one fix. How I layered deterministic hard gates and an LLM-as-judge soft gate across a LangGraph agent — and why the refusal itself turned out to be the leak. Wed, 15 Jul 2026 00:00:00 GMT https://suryansh.work/blog/grounding-financial-agents-with-vertex-multimodal-rag https://suryansh.work/blog/grounding-financial-agents-with-vertex-multimodal-rag LangGraph runs the control plane; Vertex AI's managed multimodal RAG Engine runs the data plane. How I got retrieval quality and groundedness where I needed it — because in financial filings, the number lives in the chart, not the paragraph. Sun, 12 Jul 2026 00:00:00 GMT https://suryansh.work/blog/how-i-code-with-claude-code https://suryansh.work/blog/how-i-code-with-claude-code The context-management habits that keep an AI coding agent sharp over a long session — when to /clear, when to /compact, and how to write a CLAUDE.md the model actually reads. Fri, 10 Jul 2026 00:00:00 GMT https://suryansh.work/blog/system-design-notes-for-ai-products https://suryansh.work/blog/system-design-notes-for-ai-products The infrastructure lessons behind running LLM features in production — streaming, caching, cost routing, and treating the model as an unreliable network call. Fri, 10 Jul 2026 00:00:00 GMT https://suryansh.work/blog/how-i-architect-langgraph-agents https://suryansh.work/blog/how-i-architect-langgraph-agents The structure I reach for when building production AI agents — explicit state, small nodes, tool boundaries, and knowing when NOT to let the model decide. Wed, 08 Jul 2026 00:00:00 GMT https://suryansh.work/blog/building-macget-download-engine https://suryansh.work/blog/building-macget-download-engine A build log on how MacGet downloads files fast on macOS — chunked HTTP parallelism, an actor-based engine in Swift, browser-capture over Native Messaging, and bundled yt-dlp for video. Sat, 27 Jun 2026 00:00:00 GMT