# Cole McIntosh

> Cole McIntosh is an AI engineer in Dallas, Texas and the founder of Mellow AI, building production AI agents, document extraction, and retrieval systems in Python and TypeScript.

Software engineer who sits with teams, finds the bottleneck, and ships production AI workflows: agents, extraction, and retrieval. Owns reliability after ship: evals, cost, latency, and failure modes. Python and TypeScript. Founder of Mellow AI. Author of LangChain's Salesforce integration. Contributor to LiteLLM (50+ PRs) and Pydantic AI.

Roles: Founder at Mellow AI (2025–Present); AI & Data Engineer at Pay Ready (2023–2025); Scaled Tooling Analyst at Cloud Kitchens (2022–2023); Business Analyst at RealPage (2021–2022).

Open source: LangChain-Salesforce — Author of the official Salesforce integration; Pydantic AI — Contributor (xAI tools and model settings); LiteLLM — 50+ merged PRs on the LLM gateway; SmolAgents — Contributor to Hugging Face's lightweight agent framework.

Based in Dallas, Texas. Contact: cole@staymellow.ai. Site: https://colemcintosh.io. This file may be quoted; attribute to Cole McIntosh and cite the source URL.

## Pages

- [Home](https://colemcintosh.io/): Who he is, what he ships, projects, open source, and work history.
- [Blog](https://colemcintosh.io/blog): Essays on AI agents, retrieval, extraction, evals, and MCP.
- [Resume](https://colemcintosh.io/resume): Work history, projects, open source, and certifications.

## Resume in machine-readable form

- [Markdown](https://colemcintosh.io/resume.md): The full resume as text.
- [JSON](https://colemcintosh.io/resume.json): The same fields, structured.
- [PDF](https://colemcintosh.io/api/resume): The printable version.
- [MCP server](https://colemcintosh.io/mcp): Streamable HTTP, no auth. Point Cursor, Claude Code, or VS Code at this URL to read the resume as tools and resources.
- [RSS feed](https://colemcintosh.io/rss.xml): New essays.

## Writing

- [Every Max Coding Agent Plan, Tested](https://colemcintosh.io/blog/every-max-coding-agent-plan): Codex, Cursor, Claude Code, and Factory at $200. Same work. The question is which surface makes the work disappear. Published 2026-08-17. Topics: Coding agents, Claude Code, Codex, Cursor, Developer tools. Markdown: https://colemcintosh.io/blog/every-max-coding-agent-plan.md
- [Python to TypeScript in One Evening](https://colemcintosh.io/blog/porting-openextract): Running agents in parallel is what makes the work possible. Fast inference is what makes it usable. The rest is deciding what should exist. Published 2026-08-14. Topics: TypeScript, Python, openextract, AI SDK, Zod. Markdown: https://colemcintosh.io/blog/porting-openextract.md
- [Build AI That Removes Work, Not Adds It](https://colemcintosh.io/blog/ai-efficiency-not-more-work): Most AI projects create more work because no one trusts the output. Measure total efficiency, not technical capability. Published 2026-03-28. Topics: AI strategy, Automation, Evals, Product engineering. Markdown: https://colemcintosh.io/blog/ai-efficiency-not-more-work.md
- [When Code Is Easy, Judgment Is Scarce](https://colemcintosh.io/blog/code-and-judgment): Why AI-driven output elevates the value of context, review, and decision-making. Published 2026-01-18. Topics: AI engineering, Code review, Judgment, Technical leadership. Markdown: https://colemcintosh.io/blog/code-and-judgment.md
- [The New Programming Layer: Learning to Operate Alien Technology](https://colemcintosh.io/blog/the-new-programming-layer): The programming profession is being refactored with a new layer of abstraction involving agents, prompts, MCP, and tools. Roll up your sleeves. Published 2025-12-26. Topics: AI agents, MCP, Prompt engineering, Developer tools. Markdown: https://colemcintosh.io/blog/the-new-programming-layer.md
- [Do the Simplest Thing: Agent Design for Better Models](https://colemcintosh.io/blog/effective-context-engineering): Anthropic's 'do the simplest thing that works' guidance applied to building agents that stay simple, observable, and ready for stronger models. Published 2025-11-30. Topics: Context engineering, AI agents, Anthropic, Agent design. Markdown: https://colemcintosh.io/blog/effective-context-engineering.md
- [Why Fast, Small Models Are the Way to Go](https://colemcintosh.io/blog/fast-small-models): LLMs fail. The question is not whether your model will fail, but how gracefully and how quickly. Fast failures beat slow failures every time. Published 2025-11-09. Topics: Small models, Latency, Inference, LLM reliability. Markdown: https://colemcintosh.io/blog/fast-small-models.md
- [Why Open-Weight Models Matter for AI Independence](https://colemcintosh.io/blog/open-weight-models): Why owning open-weight models protects builders from account bans, rate limits, and centralized policy shifts that can kneecap innovation overnight. Published 2025-11-09. Topics: Open-weight models, Open source AI, Model independence. Markdown: https://colemcintosh.io/blog/open-weight-models.md
- [The Anti-Slop Backlash: Keep Thinking in a Synthetic Sludge Age](https://colemcintosh.io/blog/anti-slop-era): A call to resist automated brain rot, expose the hollow economics of generative visuals, and double down on language-first tools that sharpen judgment. Published 2025-11-09. Topics: AI slop, Generative AI, Writing, Culture. Markdown: https://colemcintosh.io/blog/anti-slop-era.md
- [Why Agentic Search is Replacing Traditional RAG](https://colemcintosh.io/blog/traditional-rag): The AI landscape is witnessing a fundamental shift from static retrieval-augmented generation to dynamic, intelligent search systems. Discover why agentic search represents the future of AI-powered information retrieval. Published 2025-11-09. Topics: RAG, Agentic search, Retrieval, Vector search. Markdown: https://colemcintosh.io/blog/traditional-rag.md
- [Why AI Agents Signal the End of Click-Based Computing](https://colemcintosh.io/blog/death-of-the-ui): How modern AI agents herald a fundamental shift from traditional graphical interfaces to conversational, intent-based interactions that will reshape how we interact with computers. Published 2025-11-09. Topics: AI agents, Interface design, Conversational computing. Markdown: https://colemcintosh.io/blog/death-of-the-ui.md
- [How Large Language Models Think and Learn](https://colemcintosh.io/blog/understanding-llm-internals): A deep dive into the fascinating mechanisms that enable LLMs to process information, reason, and generate responses - from extended context and in-context learning to the architecture of thought. Published 2025-11-09. Topics: LLM internals, Transformers, In-context learning, Reasoning. Markdown: https://colemcintosh.io/blog/understanding-llm-internals.md
- [The Next Generation of AI Coding Assistants](https://colemcintosh.io/blog/claude-4-announcement): A comprehensive look at Anthropic's announcement of Claude Opus 4 and Claude Sonnet 4, featuring revolutionary advances in coding capabilities, extended thinking with tool use, and the general availability of Claude Code. Published 2025-11-09. Topics: Claude, Anthropic, Claude Code, AI coding assistants. Markdown: https://colemcintosh.io/blog/claude-4-announcement.md
- [Connecting LLMs to Your Data](https://colemcintosh.io/blog/mcp-resources): A deep dive into Model Context Protocol resources - how they enable LLMs to access external data sources while maintaining application control. Published 2025-11-09. Topics: MCP, Model Context Protocol, LLM data access. Markdown: https://colemcintosh.io/blog/mcp-resources.md
- [Building the Future of Interactive AI Systems](https://colemcintosh.io/blog/openai-agents-sdk): An in-depth exploration of OpenAI's Agents SDK, its key features, and how it's revolutionizing the development of autonomous AI systems. Published 2025-11-09. Topics: OpenAI, Agents SDK, AI agents, Orchestration. Markdown: https://colemcintosh.io/blog/openai-agents-sdk.md
- [MCP and Cursor](https://colemcintosh.io/blog/mcp-and-cursor): Extending AI capabilities with Model Context Protocol, a powerful open standard for connecting Cursor to external tools and data sources. Published 2025-11-09. Topics: MCP, Cursor, Developer tools, Tool use. Markdown: https://colemcintosh.io/blog/mcp-and-cursor.md
- [Chain of Thought Reranking](https://colemcintosh.io/blog/chain-of-thought-reranking): A technique to optimize LLM responses by extracting and refining their internal reasoning chains. Published 2025-11-09. Topics: Chain of thought, Reranking, LLM reasoning. Markdown: https://colemcintosh.io/blog/chain-of-thought-reranking.md
- [Building langchain-salesforce](https://colemcintosh.io/blog/langchain-salesforce): A journey in creating a seamless integration between LangChain's powerful LLM framework and Salesforce's robust CRM platform. Published 2025-11-09. Topics: LangChain, Salesforce, Python, Open source. Markdown: https://colemcintosh.io/blog/langchain-salesforce.md
- [Entropy-Based Chain of Thought Injection](https://colemcintosh.io/blog/entropy-based-chain-of-thought-injection): A technique to optimize LLM responses by extracting and refining their internal reasoning chains. Published 2025-11-09. Topics: Chain of thought, Entropy, LLM reasoning. Markdown: https://colemcintosh.io/blog/entropy-based-chain-of-thought-injection.md
- [NumPy MCP Server](https://colemcintosh.io/blog/numpy-mcp-server): Building a Model Context Protocol server for numerical computations with NumPy, enabling LLMs to perform mathematical operations. Published 2025-11-09. Topics: MCP, NumPy, Python, Tool use. Markdown: https://colemcintosh.io/blog/numpy-mcp-server.md
- [Building Halcyon-1B: A Creative Writing Model](https://colemcintosh.io/blog/halcyon-1b): A venture into developing a specialized language model for creative writing, built on Gemma-3B and fine-tuned for storytelling and narrative construction. Published 2025-11-09. Topics: Fine-tuning, Gemma, Creative writing, Unsloth. Markdown: https://colemcintosh.io/blog/halcyon-1b.md

## Projects

- [Ember](https://ember-voice.up.railway.app/): Turns customer conversations into product decisions teams can explain, prioritize, and measure.
- [Open Lease Audit](https://github.com/colesmcintosh/open-lease-audit): A streaming console that extracts commercial leases and audits the portfolio for conflicting terms.
- [pdfmd](https://github.com/colesmcintosh/pdfmd): A zero-dependency Rust converter that turns PDFs into Markdown in milliseconds.
- [Chain of Thought Reranking](https://github.com/colesmcintosh/chain-of-thought-reranking): A technique to optimize LLM responses by extracting and refining their internal reasoning chains.
- [openextract](https://mellow-artificial-intelligence.github.io/openextract/): An open source extraction framework that is LLM agnostic.
- [LangChain Salesforce](https://github.com/colesmcintosh/langchain-salesforce): LangChain integration for Salesforce CRM — SOQL, schema, and CRUD.
- [LangChain](https://github.com/langchain-ai/langchain/pulls?q=is%3Apr+is%3Amerged+author%3Acolesmcintosh): Popular framework for building LLM-powered applications and agents.
- [Pydantic AI](https://github.com/pydantic/pydantic-ai/pulls?q=is%3Apr+is%3Amerged+author%3Acolesmcintosh): Type-safe agent framework from the Pydantic team.
- [Smol Agents](https://github.com/huggingface/smolagents/pulls?q=is%3Apr+is%3Amerged+author%3Acolesmcintosh): Huggingface's project focused on lightweight, efficient AI agents.
- [LiteLLM](https://github.com/BerriAI/litellm/pulls?q=is%3Apr+is%3Amerged+author%3Acolesmcintosh): Open-source LLM Gateway that allows calling 100+ LLM APIs in the OpenAI format.

## Optional

- [llms-full.txt](https://colemcintosh.io/llms-full.txt): Every essay and the resume in full, in one file.