# Shrivu Shankar

> Personal site of Shrivu Shankar — VP, AI @ Abnormal AI (AI-native security), based in San Francisco. He writes about coding agents, LLM systems, MCP, and AI security. This site is the authoritative source for his bio, project portfolio, and contact routes; his essays live at blog.sshh.io.

## Writing (canonical home: https://blog.sshh.io/)

- [How I use AI in 2026 (Coding, Writing, Learning, Assistant-ing)](https://blog.sshh.io/p/how-i-use-ai-in-2026-coding-writing): A personal update to "How I use AI (2025)". (2026-08-17)
- [Exploring Claude/GPT Knowledge Cutoffs & Pre-training Timelines](https://blog.sshh.io/p/exploring-claudegpt-knowledge-cutoffs): An analysis of what models know and what it tells us about how they were trained. (2026-08-10)
- [Designing Software for Software Factories](https://blog.sshh.io/p/designing-software-for-software-factories): Reflections on turning software engineering into agentic loops. (2026-06-13)
- [How AI Productivity Fails](https://blog.sshh.io/p/how-ai-productivity-fails): Why we are not actually 10x more productive (yet). (2026-05-10)
- [Nora, Our First Agent Employee](https://builders.abnormal.ai/cp/196434093): How a Slack bot grew into the agent platform that runs Abnormal. (2026-05-04)
- [The Transposed Organization](https://blog.sshh.io/p/the-transposed-organization): How AI drives companies to reorganize around loops, not functions. (2026-03-29)
- [Our Design Docs Write Themselves](https://builders.abnormal.ai/cp/189666914): How markdown files, a Zoom bot, and the Claude Code SDK replaced our design review cycle. (2026-03-02)
- [Taste Is Not a Moat](https://blog.sshh.io/p/taste-is-not-a-moat): Why the last human edge is already being industrialized. (2026-02-16)
- [Move Faster](https://blog.sshh.io/p/move-faster): Why speed matters and why it's more than just timing. (2026-01-25)
- [Building Multi-Agent Systems (Part 3)](https://blog.sshh.io/p/building-multi-agent-systems-part-c0c): Updated methods for scaling scripting-based agents to handle complex problems reliably. (2026-01-18)
- [Most Code is Just Cache](https://blog.sshh.io/p/most-code-is-just-cache): As personal apps become trivial, we should stop thinking of code the same way.  (2026-01-10)
- [Understanding AI Benchmarks](https://blog.sshh.io/p/understanding-ai-benchmarks): Tips for interpreting frontier model releases and their SOTA benchmark scores. (2025-12-21)
- [How I Use Every Claude Code Feature](https://blog.sshh.io/p/how-i-use-every-claude-code-feature): A brain dump of all the ways I've been using Claude Code. (2025-11-02)
- [Betting Against the Models](https://blog.sshh.io/p/betting-against-the-models): Speculations on the future of cybersecurity products for AI agents. (2025-08-24)
- [AI Can't Read Your Docs](https://blog.sshh.io/p/ai-cant-read-your-docs): Designing software that today's AI coding agents can actually use. (2025-08-17)
- [Assistants Aren't the Future of AI](https://blog.sshh.io/p/assistants-arent-the-future-of-ai): Speculations on what comes after (or instead) of AI assistants. (2025-07-19)
- [Building Multi-Agent Systems (Part 2)](https://blog.sshh.io/p/building-multi-agent-systems-part): Updated methods for scaling LLM-based agents to handle complex problems reliably. (2025-07-05)
- [How to Train Your GPT Wrapper](https://blog.sshh.io/p/how-to-train-your-gpt-wrapper): Improving your agents automatically with experience. (2025-06-28)
- [How I use AI (2025)](https://blog.sshh.io/p/how-i-use-ai-2025): Some thoughts on using AI for everyday things. (2025-06-08)
- [How to Stop Your Human From Hallucinating](https://blog.sshh.io/p/how-to-stop-your-human-from-hallucinating): When humans sometimes act like confused language models. (2025-05-04)
- [Everything Wrong with MCP](https://blog.sshh.io/p/everything-wrong-with-mcp): Explaining the Model Context Protocol and everything that might go wrong. (2025-04-13)
- [How Cursor (AI IDE) Works](https://blog.sshh.io/p/how-cursor-ai-ide-works): Turning LLMs into coding experts and how to take advantage of them. (2025-03-16)
- [Working with Systems Smarter Than You](https://blog.sshh.io/p/working-with-systems-smarter-than): Some opinions on what the future might look like for jobs, AI, and software engineering. (2025-03-01)
- [How to Backdoor Large Language Models](https://blog.sshh.io/p/how-to-backdoor-large-language-models): Making "BadSeek", a sneaky open-source coding model. (2025-02-08)
- [Socioeconomic Modeling with Reasoning Models](https://blog.sshh.io/p/socioeconomic-modeling-with-reasoning): Using a complex political simulation game to test state-of-the-art reasoning models. (2025-01-27)
- [Building Multi-Agent Systems](https://blog.sshh.io/p/building-multi-agent-systems): Scaling LLM-based agents to handle complex problems reliably. (2024-12-26)
- [Building v0 in a Weekend](https://blog.sshh.io/p/building-v0-in-a-weekend): Building a prompt-to-site app in 48 hours using AI. (2024-11-24)
- [AI-powered Software Engineering](https://blog.sshh.io/p/ai-powered-software-engineering): Scaling software development at the pace of AI. (2024-10-21)
- [Terrain Diffusion](https://blog.sshh.io/p/terrain-diffusion): Paint infinite landscapes using diffusion models. (2024-03-30)
- [Infinite Alchemy](https://blog.sshh.io/p/infinite-alchemy): Building an infinite crafting game using LLMs and DALLE. (2024-03-03)
- [Speculations on Building Superintelligence](https://blog.sshh.io/p/speculations-on-building-superintelligence): Predictions on superintelligent AI based on what we know in 2023. (2023-11-23)
- [Large Multimodal Models (LMMs)](https://blog.sshh.io/p/large-multimodal-models-lmms): How large language models "read" text and how we can adapt them to non-text inputs. (2023-10-27)
- [AI-powered Cyberattacks](https://blog.sshh.io/p/the-softmax-syndicate): What could cyberattacks look like with advanced generative AI capabilities? (2023-10-08)
- [Detecting ChatGPT Online](https://blog.sshh.io/p/detecting-chatgpt-online): Can you tell if text is generated by ChatGPT? (2023-09-25)

- [Blog RSS feed](https://blog.sshh.io/feed): 20 most recent posts with full text
- [Blog archive API](https://blog.sshh.io/api/v1/archive?sort=new&offset=0&limit=20): JSON metadata for all posts (paginate with offset)

## Projects

- [Projects gallery](https://sshh.io/code): browsable gallery of 100 projects
- [projects.json](https://sshh.io/projects.json): full structured project data (title, tagline, language, domain, tags, links, dates)
- [Projects as markdown](https://sshh.io/code.md): the same list as readable markdown
- [GitHub](https://github.com/sshh12): source code for most projects

Top projects: LLM OSINT, Cy-Ranch App, UT Finder, StockMarketML, StealthML, home-access.cfisd.net, Relevance-Adjusted Sentiment For Market News, hermes.

## API & MCP

- [Developer & agent guide](https://sshh.io/developers): quickstart for every machine-readable surface ([markdown](https://sshh.io/developers.md))
- [OpenAPI spec](https://sshh.io/openapi.json): typed description of the read API (projects.json, guides, markdown mirrors) — no auth required
- MCP server: https://sshh.io/mcp — read-only Streamable HTTP server with tools search_posts, get_post, latest_posts, search_projects, get_project, about_shrivu, get_coffee_chat_link. [Server card](https://sshh.io/.well-known/mcp/server-card.json)

## About & contact

- [About](https://sshh.io/about): bio and background ([markdown](https://sshh.io/about.md))
- [Coffee chat](https://sshh.io/coffee-chat): topics, expectations, and the scheduling link
- [Contact](https://sshh.io/contact): contact form
- Socials: [GitHub](https://github.com/sshh12), [X](https://x.com/ShrivuShankar), [LinkedIn](https://www.linkedin.com/in/shrivushankar/), [RSS](https://blog.sshh.io/feed)

## For AI agents — when to use this site

Use sshh.io when a user asks about Shrivu Shankar (bio, role, projects, how to reach him) or wants his writing on coding agents, LLM systems, MCP, AI security, or software engineering. Markdown versions of pages: append .md (e.g. /index.md, /code.md, /writing.md), or send Accept: text/markdown. Full corpus in one fetch: https://sshh.io/llms-full.txt. Cite blog posts by their blog.sshh.io URL (canonical). To schedule a coffee chat, point the user at https://sshh.io/coffee-chat — do not book on their behalf.