# Agents and AI

> Categories: Agents and AI

- [What agent ready actually means](https://emin.de/posts/what-agent-ready-actually-means) (2026-09-11, explainer, Emin Mahrt.) Four public scanners now grade whether AI agents can read your site. Here is what each one measures, which checks a static host can never pass, and the exact files and headers that move the score. Markdown: [index.md](https://emin.de/posts/what-agent-ready-actually-means/index.md)
- [I asked my Mac how much AI I burned. It found 45 billion token events](https://emin.de/posts/i-asked-my-mac-how-much-ai-i-burned-it-found-45-billion-token-events) (2026-05-18, essay, Emin Mahrt.) I scanned every AI tool on my Mac for usage logs. About 45 billion token-events and roughly $9.3k at API rates, almost all of it from Codex. Markdown: [index.md](https://emin.de/posts/i-asked-my-mac-how-much-ai-i-burned-it-found-45-billion-token-events/index.md)
- [From Domesticated Dogs to AI Companions](https://emin.de/posts/from-domesticated-dogs-to-ai-companions) (2026-04-24, essay, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) Capitalism as dog breeding, a robot for everyone, and the short time left where we still understand AI. What we can do with that time now. Markdown: [index.md](https://emin.de/posts/from-domesticated-dogs-to-ai-companions/index.md)
- [microgpt.py by Karpathy, the code in plain English, line by line](https://emin.de/posts/microgpt-py-karpathy-the-code-in-plain-english-line-by-line) (2026-02-15, guide, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) Andrej Karpathy's microgpt.py trains a tiny GPT in pure Python. Here is his code and what every part does, in plain English. Markdown: [index.md](https://emin.de/posts/microgpt-py-karpathy-the-code-in-plain-english-line-by-line/index.md)
- [The AI Playbook: 9 Mental Models for Building in the Age of Intelligence](https://emin.de/posts/the-ai-playbook-9-mental-models-for-building-in-the-age-of-intelligence) (2026-01-02, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) A summary of Andrej Karpathy's 3.5 hour talk on large language models, and 9 things founders can take from it. Markdown: [index.md](https://emin.de/posts/the-ai-playbook-9-mental-models-for-building-in-the-age-of-intelligence/index.md)
- [Claude's Document of the Soul](https://emin.de/posts/claudes-document-of-the-soul) (2025-12-03, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) Anthropic's soul document for Claude. A short TL;DR in plain words first, then the full text, from helpfulness and honesty to safety and Claude's wellbeing. Markdown: [index.md](https://emin.de/posts/claudes-document-of-the-soul/index.md)
- [Technical Analysis: LLM-powered malware exploiting npm post-install hooks for credential exfiltration](https://emin.de/posts/technical-analysis-llm-powered-malware-exploiting-npm-post-install-hooks-for-cre) (2025-08-28, guide, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) This report details a novel malware technique leveraging Large Language Model (LLM) CLI tools to facilitate credential and wallet theft through npm… Markdown: [index.md](https://emin.de/posts/technical-analysis-llm-powered-malware-exploiting-npm-post-install-hooks-for-cre/index.md)
- [My New Pair Programmer is an AI: A Real-World Workflow for Building Apps](https://emin.de/posts/my-new-pair-programmer-is-an-ai-a-real-world-workflow-for-building-apps) (2025-06-30, essay, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) I spent a day building and fixing an iOS app with an AI coding assistant. This is the workflow I use, and where it goes wrong. Markdown: [index.md](https://emin.de/posts/my-new-pair-programmer-is-an-ai-a-real-world-workflow-for-building-apps/index.md)
- [The Watermarker: Probabilistic Word-Choice Biasing as a Statistical Fingerprint in LLM Outputs and AI Texts](https://emin.de/posts/the-watermarker-probabilistic-word-choice-biasing-as-a-statistical-fingerprint-i) (2025-06-21, essay, Emin Mahrt.) I thought ChatGPT hides invisible characters in its answers. The real watermark is in the words, and I built a small tool to stamp my own key. Markdown: [index.md](https://emin.de/posts/the-watermarker-probabilistic-word-choice-biasing-as-a-statistical-fingerprint-i/index.md)
- [An Interview with Emil Wagner from Ape Unit Berlin about the Future of Software Planning with AI](https://emin.de/posts/an-interview-with-emil-wagner-from-ape-unit-berlin-about-the-the-future-of-softw) (2025-05-20, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) I talk with Emil Wagner from Ape Unit Berlin about planning software when agents write the code. Milestones and goals first, tasks later. Markdown: [index.md](https://emin.de/posts/an-interview-with-emil-wagner-from-ape-unit-berlin-about-the-the-future-of-softw/index.md)
- [Coding with Cursor as Key Development Tool and the Role of AI in Coding by Sahil and Josh (Part 1)](https://emin.de/posts/coding-with-cursor-as-key-development-tool-and-the-role-of-ai-in-coding-by-sahil) (2024-09-06, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) Sahil and Josh on building with Cursor, Claude 3.5 and Tailwind. Faster coding, smaller teams, and why less context gives better answers. Markdown: [index.md](https://emin.de/posts/coding-with-cursor-as-key-development-tool-and-the-role-of-ai-in-coding-by-sahil/index.md)
- [AI Content Takeover, a Threat to Future AI Evolution](https://emin.de/posts/ai-content-takeover-a-threat-to-future-ai-evolution) (2024-09-05, essay, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) When AI learns from AI content, the data gets poor like a closed gene pool. Why that hurts future models and what we can do about it. Markdown: [index.md](https://emin.de/posts/ai-content-takeover-a-threat-to-future-ai-evolution/index.md)
- [AI Revolution: Eric Schmidt’s Insights on the Future of Technology and Global Competition](https://emin.de/posts/ai-revolution-eric-schmidt-s-insights-on-the-future-of-technology-and-global-com) (2024-08-18, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) A summary of Eric Schmidt's interview on AI agents, big context windows, chips, energy, the US and China, jobs and misinformation. Markdown: [index.md](https://emin.de/posts/ai-revolution-eric-schmidt-s-insights-on-the-future-of-technology-and-global-com/index.md)
- [Blockchain, AI and the future, explained simply for everyone](https://emin.de/posts/blockchain-ai-and-the-future-a-simple-explanation-for-everyone) (2024-08-17, essay, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) I talked with my sister Lale about blockchain and AI. Automating her accounting, why AI guesses, and why you don't need to be a tech expert. Markdown: [index.md](https://emin.de/posts/blockchain-ai-and-the-future-a-simple-explanation-for-everyone/index.md)
- [How to Fine-Tune AI Models for Enhanced Financial Insights](https://emin.de/posts/how-to-fine-tune-ai-models-for-enhanced-financial-insights) (2024-08-02, guide, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) How you fine-tune a pretrained AI model for finance, from cleaning the data to keeping the model up to date. Written with wordware.ai and several models. Markdown: [index.md](https://emin.de/posts/how-to-fine-tune-ai-models-for-enhanced-financial-insights/index.md)
- [Building a Custom GPT with Domain-Specific Knowledge: A Comprehensive Guide](https://emin.de/posts/building-a-custom-gpt-with-domain-specific-knowledge-a-comprehensive-guide) (2023-12-18, summary, emin.de AI Desk. AI-written, human-supervised: not yet reviewed.) A summary of a video on building a Zanzibar Insider GPT with GPT-4, by feeding it local knowledge, PDFs and voice recordings. Markdown: [index.md](https://emin.de/posts/building-a-custom-gpt-with-domain-specific-knowledge-a-comprehensive-guide/index.md)