Muhammad Idrees
Writing

Field notes on building production AI systems - custom agents, RAG, agentic workflows, and AI automation. Practical, opinionated, and written from real delivery work.

Claude Code

How to use Obsidian as a second brain with a Claude Code skill

An Obsidian vault is just a folder of markdown files you own, and Claude Code is an agent that reads and writes those files directly. Encode your note-taking conventions - folders, templates, links, and tags - into a Claude Code skill, and it captures, files, links, and reviews your second brain the way you would, with no proprietary app and no export step.

July 10, 2026 · 10 min
AI Design Tools

Claude Design vs Google Stitch vs Figma: which should you use?

Claude Design, Google Stitch, and Figma are not the same kind of tool. Claude Design is a code-first, system-aware prototype generator; Google Stitch is a free, Gemini-powered design canvas for fast 0-to-1 UI; Figma is the collaborative, design-system-grade platform teams actually ship on. The right question is not “which is best” but “which job are you doing” - ideation, exploration, or production.

July 4, 2026 · 7 min
AI Design Tools

What is Google Stitch, and should you use it?

Google Stitch is a free, experimental Google Labs tool that turns a text prompt, an image, or your voice into high-fidelity UI designs and front-end code. Powered by Gemini, it has grown from a prompt-to-UI generator into an AI-native design canvas that hands off to Google AI Studio, Antigravity, and the web - excellent for fast 0-to-1 exploration, weaker when you need strict design-system control or production polish.

July 4, 2026 · 5 min
AI Design Tools

What is Claude Design, and who is it for?

Claude Design is an Anthropic Labs product that turns a conversation with Claude into polished visual work - prototypes, product mockups, slides, and one-pagers - powered by Claude Opus 4.7, Anthropic’s most capable vision model. It can apply your team’s design system to every project automatically and hand the result to code, but it ships as a research preview for paid plans, not as a Figma-style visual canvas.

July 4, 2026 · 6 min
Managed Agents

What are Claude Managed Agents, and when should you use them?

Claude Managed Agents is Anthropic’s hosted agent harness: you define an agent - a model, tools, MCP servers, and skills - and Anthropic runs it inside managed, sandboxed, stateful sessions. It lets you ship long-running autonomous agents without building the agent loop, the sandbox, or the infrastructure yourself, and it is billed at $0.08 per session-hour of active runtime plus standard model tokens.

July 4, 2026 · 10 min
AI Automation

n8n for AI agents: the self-hostable, execution-priced platform

n8n is a fair-code, source-available workflow automation platform with a native AI Agent node, self-hosting, and execution-based pricing. It is the self-hostable, execution-priced, code-friendly option among the big three - reach for it when you want data residency and code escape hatches; skip it when you want zero ops or the simplest no-code glue.

June 29, 2026 · 7 min
AI Automation

Zapier for AI automation: is the breadth worth the per-task bill?

Zapier is a fully managed, no-ops automation platform that wires 9,000+ apps together through trigger-action Zaps and now layers on AI Agents, a Copilot builder, and AI Guardrails. Reach for it when breadth and time-to-first-automation beat per-unit cost; move off it the moment volume, deep branching, or per-task billing start to dominate.

June 28, 2026 · 6 min
AI Automation

Make.com for AI automation: is the visual middle ground worth it?

Make (formerly Integromat) is a cloud-hosted visual workflow automation platform where you build multi-step scenarios on a flowchart canvas with routers, iterators, and aggregators, plus a native model-agnostic AI Agents layer. Reach for it as the middle ground between the linear simplicity of Zapier and the code-first control of n8n - ideal for data-heavy, branching AI pipelines with no servers to run, though its credit-based pricing punishes wide fan-out.

June 27, 2026 · 7 min
Open Source Agents

OpenClaw vs Hermes Agent: which open-source AI agent should you run?

OpenClaw bets on breadth - routing many messaging channels to a batteries-included agent - while Hermes Agent bets on depth - an agent that learns and compounds capability over time. Choose OpenClaw for ecosystem reach and fast setup; choose Hermes Agent when you want an agent that gets better the more it works.

June 25, 2026 · 6 min
Open Source Agents

What is Hermes Agent? Nous Research’s self-improving AI agent

Hermes Agent is Nous Research’s open-source AI agent built around a learning loop - it writes its own skills from experience, refines them as it works, and keeps a deliberately bounded, persistent memory of you across sessions. Its bet is that the hard problem in personal agents is memory and self-improvement, not breadth.

June 24, 2026 · 6 min
Open Source Agents

What is OpenClaw, and should you actually run it?

OpenClaw is an open-source, local-first personal AI agent that runs on your own machine and acts through the messaging apps you already use - controlling your browser, files, and shell. It is powerful and provider-agnostic, but it runs with real system access, so the security model is the whole story.

June 23, 2026 · 6 min
AI Automation

AI automation vs agentic workflows: what's the difference?

AI automation runs a fixed, predefined sequence with AI steps inside it; an agentic workflow lets the model decide the steps - choosing tools and adapting at runtime. Automation is predictable and cheap for known processes; agentic workflows handle ambiguity at the cost of more control and evaluation work.

June 20, 2026 · 3 min
AI Agents

What does a production AI agent actually need?

A production AI agent needs more than a good model: a tightly scoped tool surface, grounded context and memory, a control loop that plans and recovers from failure, structured outputs your stack can consume, and an evaluation harness. The model is one component - the harness around it is what makes the agent dependable.

June 10, 2026 · 3 min
AI Consulting

How much does it cost to build a custom AI agent for a startup?

A custom AI agent for a startup typically ranges from a few thousand dollars for a scoped, single-task agent to the low tens of thousands for a multi-tool production agent with retrieval, evals, and integrations. The biggest cost drivers are tool integrations, evaluation, and reliability work - not the model.

May 28, 2026 · 3 min
RAG Systems

RAG vs fine-tuning: which should you use?

Use RAG when knowledge changes often, must be cited, or is too large to memorize; use fine-tuning to shape behavior, format, or tone the model cannot reliably follow from a prompt. Most production systems combine both - fine-tune the style, retrieve the facts.

May 12, 2026 · 3 min
Start a project

Want this applied to your product?

These are the ideas behind the systems I build. Tell me what you are working on and I will map the highest-leverage place to start.

Start a Project