/
Published on 22.03.2026
TLDR: The Model Context Protocol is trying to solve a problem that the Unix command line already solved fifty years ago. The most effective developers are already letting their AI agents drive CLI tools instead of building new JSON-based protocol layers.
This is a spicy take, and I think it is largely correct. The article makes a compelling argument that MCP adds friction where the terminal already provides a perfect abstraction. The core insight is about transparency. When a CLI tool fails, it tells you why in plain English. When an MCP server chokes, you are digging through transport logs and decoding JSON RPC messages, trying to figure out if the server crashed, the initialization failed, or the protocol itself just gave up.
The Unix pipe argument is particularly strong. Consider a task like analyzing a cloud infrastructure plan. In the CLI world, an agent can run a plan, pipe it through jq to filter for specific changes, and grep for errors in a single line of bash. In the MCP world, you either dump the entire file into the context window and hope the model does not hallucinate, or you build custom filtering logic into every single MCP server. You are doing more work to get a worse result.
The authentication point also hits home. Tools like the AWS CLI and kubectl already handle complex SSO and credential rotation. When you run a login command once, your agent inherits those same permissions without needing a separate redundant auth layer. Using multiple MCP tools today feels like an endless loop of re-authentication.
The broader thesis is that the best tools work for both people and machines. When you build a rock-solid CLI, you give human power users a tool they love and AI agents a standard interface they already know how to drive. LLMs were trained on nearly every man page ever written — they are already experts at using command-line tools. Why build a new abstraction layer on top of something that already works?
TLDR: A comprehensive comparison of the four major AI models in March 2026 concludes that Claude leads in reasoning, writing, and enterprise trust, Gemini wins on ecosystem integration and multimodality, ChatGPT retains the broadest user base but is showing trust cracks, and Grok has a real-time data moat but a governance crisis.
This is one of those HackerNoon articles that is clearly opinionated — the author does not hide their preferences — but the underlying facts and observations are worth engaging with regardless of whether you share the conclusions.
The comparison covers four major fronts. Google Gemini is executing an ecosystem strategy that nobody else can replicate, making itself the connective tissue across Gmail, Drive, Maps, Search, Chrome, and Samsung devices. Gemini 3 Deep Think's benchmark performance is genuinely impressive, and the partnership with Apple announced in 2026 is a massive distribution coup. The challenge is that Gemini still lacks a distinctive voice or personality that makes users choose it by preference rather than by default.
Anthropic Claude has had the most dramatic early 2026, with a public battle against the Pentagon over safety rails, an ad-free commitment, and a reported migration of nearly 2.5 million users from ChatGPT. Claude Sonnet 4.6 and the broader 4.x family are competitive at the frontier, particularly in writing quality and nuanced professional work. The Claude Marketplace move — letting enterprises apply Anthropic spending commitments toward third-party Claude-powered tools — is a strategic shift from model provider to platform.
OpenAI ChatGPT remains the incumbent with 500 million registered users and the fastest model iteration cadence. GPT-5.4 with native computer use is a meaningful capability leap. But the decision to introduce ads in the Go tier, the delayed adult mode, and the Pentagon contract for mass surveillance have created a trust gap that the QuitGPT movement has exploited. The optics matter: Anthropic makes headlines for refusing Pentagon contracts while OpenAI makes headlines for serving ads.
xAI Grok is the wildcard with its unique X data moat, a 20 billion dollar Series E, and the SpaceX acquisition. Grok 4 is genuinely competitive, and the real-time social data access is a capability no other model can replicate. But the governance problems — racist content incidents, UK regulatory investigations, non-consensual imagery generation — are not philosophical positions. They are product failures that threaten to cap Grok's enterprise ceiling.
The final rankings put Claude first for its trust and capability combination, Gemini second for ecosystem power, ChatGPT third despite showing cracks, and Grok fourth with the most potential but the most risk.
Google Gemini vs Anthropic Claude vs OpenAI ChatGPT vs xAI Grok: The Ultimate Comparison