# The Closing Window > Essays and AI insights by Pablo Oliva — exploring the intersection of artificial intelligence, software engineering, and the human experience in an AI-transformed world. ## Blog Posts - [Securing an Always-On AI Agent at Home: A Journey Through Dead Ends](https://pablooliva.de/the-closing-window/securing-an-always-on-ai-agent-at-home/): The honest record of securing an always-on AI agent with shell access at home: why firewalling it in place failed, how network segmentation and an ngrok tunnel worked instead, and a frank list of what it still doesn't protect against. - [Hermes vs. OpenClaw: A Practical Comparison](https://pablooliva.de/the-closing-window/hermes-vs-openclaw/): OpenClaw and Hermes are both self-hosted personal AI agents. A factual comparison of their features, and why Hermes' isolation model and security posture make it the safer bet to run on your own hardware. - [Our AI Rollout Started With People, Not Processes](https://pablooliva.de/the-closing-window/our-ai-rollout-started-with-people-not-processes/): AI adoption runs on two tracks — top-down process automation and individual everyday work. We started with people: one shared, self-hosted AI tool for a sub-500-person company across several EU countries (LibreChat on Azure OpenAI in an EU region, with Redakt PII detection in the stream). Here's why we sequenced it that way, and what's in the setup. - [Give Your Coding Agent More Autonomy](https://pablooliva.de/the-closing-window/give-your-coding-agent-more-autonomy/): I run a development pipeline with permissions off, then trace every session into LangSmith. The plain case for why: an objective record of what the agent did unattended, evals that catch model regressions before they cost me, and a loop where the agent reviews its own behavior and patches my setup. - [Which AI Should You Actually Use? I Built an AI for That.](https://pablooliva.de/the-closing-window/which-ai-should-you-actually-use/): Most people use a fraction of what AI tools can do. Sherpa is an open-source agent that interviews someone about their job and recommends which AI features actually fit their work. - [Claude Made an 18-Minute Tutorial of My App](https://pablooliva.de/the-closing-window/claude-made-an-18-minute-tutorial-of-my-app/): I pointed Claude Code at our live internal product and had it film and narrate an 18-minute tutorial. The pipeline, the parts still done by hand, and how to reproduce it. - [Opus 4.7 is Terse](https://pablooliva.de/the-closing-window/opus-4-7-is-terse/): Opus 4.7's terser default style breaks down on complex projects. The fix is a custom output style that restores thorough explanations, plus evals to catch this kind of drift on future model releases. - [Thin Slices Stop /sdd-flow from Overflowing My Capacity](https://pablooliva.de/the-closing-window/thin-slices-stop-sdd-flow-from-overflowing-my-capacity/): One-shotting features with /sdd-flow breaks down on complicated projects run in parallel. The fix: thinner vertical slices that keep oversight tight. New per-slice mode is live in the plugin. - [From One Agent to Many](https://pablooliva.de/the-closing-window/from-one-agent-to-many/): Agent coordination tools now make it easy to run multiple AI agents in parallel. A look at what's available and what Trail of Bits achieved by systematizing it across an organization. - [Close Call, a Voice AI Sales Coach](https://pablooliva.de/the-closing-window/close-call-a-voice-ai-sales-coach/): I built an open-source voice agent for practicing sales calls against an AI customer. Runs on Pipecat + Gemini 2.5 Flash, costs $0.14 per call. - [Everything is Energy, Everything in Berlin is Beautiful](https://pablooliva.de/the-closing-window/everything-is-energy-everything-in-berlin-is-beautiful/): On a perfect afternoon in Berlin, everyone looked happy and oblivious. I see where the train is heading, and the accident is coming. But all crises present opportunity. - [Spec Driven Development Fans, You'll Love sdd-flow](https://pablooliva.de/the-closing-window/spec-driven-development-fans-youll-love-sdd-flow/): A Claude Code skill that orchestrates spec-driven development through subagents: research, adversarial review, specification, implementation, code review, and critical review from a single command. Built Redakt with it in two days. - [Meet Redakt: Practical GDPR Compliance for AI Teams](https://pablooliva.de/the-closing-window/meet-redakt-practical-gdpr-compliance-for-ai-teams/): Redakt is an open-source PII anonymizer built on Microsoft Presidio. It sits between employees and AI tools, replacing personal data with placeholders before prompts leave your network. No data stored, no DPA needed, fully self-hosted. - [Meet Sift: A Knowledge Base for Everything That Isn't a Note](https://pablooliva.de/the-closing-window/introducing-sift/): Sift is an open source personal knowledge base that ingests any document type and makes it searchable by meaning. The story of why it exists, what went right, and what to expect. - [Agent: Do You Understand the Words Coming Out of My Mouth?](https://pablooliva.de/the-closing-window/ai-agent-discoverability/): A practical checklist for making your website discoverable, readable, and citable by AI agents and answer engines — with a ready-to-paste prompt to implement it all. - [Finding Meaning in Your Notes with CK Search](https://pablooliva.de/the-closing-window/finding-meaning-in-your-notes-with-ck-search/): CK Search adds local semantic search to any text directory, with built-in MCP server support for AI agents. A practical walkthrough of setup, configuration, and where semantic search helps — and where it doesn't. - [From Folders to Knowledge Base: How I Made My Notes Work for Me](https://pablooliva.de/the-closing-window/from-folders-to-knowledge-base/): A personal journey from rigid folder hierarchies to an AI-powered knowledge base — through tags, semantic search, and knowledge graphs — and why investing in how you organize information pays off in ways you can't predict. - [Feed the AI - Digitize Everything](https://pablooliva.de/the-closing-window/feed-the-ai-digitize-everything/): AI agents are only as useful as the context you give them. Learn what it takes to build an external brain that lets AI act on your behalf — without asking every time. - [FolderTether: Obsidian Breaks Out of the Vault](https://pablooliva.de/the-closing-window/foldertether-obsidian-breaks-out-of-the-vault/): A bidirectional link between your Obsidian notes and the folders where your actual work lives. FolderTether keeps your vault and your filesystem connected without sync, cloud, or complexity. - [The Notes Setup That Actually Works with AI](https://pablooliva.de/the-closing-window/obsidian-and-markdown-in-the-ai-agent-era/): Why Obsidian and Markdown files turned out to be the right bet for the AI era — not for productivity reasons, but because plaintext with embedded metadata is exactly what AI agents need. - [Shadow AI and the Compliance Gap That Won't Close Itself](https://pablooliva.de/the-closing-window/shadow-ai-and-the-compliance-gap-that-wont-close-itself/): Shadow AI — employees using unapproved AI tools with company data — is creating a growing GDPR liability across Europe. Two overlapping laws apply to every AI prompt containing personal data, and most companies don't know it. - [The Workplace Blind Spot](https://pablooliva.de/the-closing-window/the-workplace-blind-spot/): Most knowledge workers assume their role's complexity will protect them from AI. I held the same belief — until I was proven wrong, twice. The models don't need to improve; they just need to be better harnessed. - [Close Enough to See](https://pablooliva.de/the-closing-window/close-enough-to-see/): From the dot-com bubble to mortgage fraud to Nord Stream to a livestreamed genocide, I've spent my life inside the systems that exploit people — across continents. This is what it looks like when the pattern finally clicks. - [I Am No Longer Needed](https://pablooliva.de/the-closing-window/i-am-no-longer-needed/): How I automated my own role as AI lead — piece by piece, meeting by meeting — until I realized the tools I championed were doing my job better than I was. A honest look at what that actually feels like. - [Which AI Should You Use in February 2026?](https://pablooliva.de/the-closing-window/which-ai-to-use-in-2026/): A practical, non-technical guide explaining which AI tool (ChatGPT, Claude, or Gemini) to use in 2026, how to pick the right model, and how AI has shifted from simple chatbots to agents that can actually do work for you. - [AI's Social Trap](https://pablooliva.de/the-closing-window/ais-social-trap/): The article argues that society is trapped in reactive, experience-based thinking that worked for linear innovation but is fatally inadequate for the exponential disruption of AI, and that — like Kodak and Nokia before us — our social structures risk becoming obsolete unless leaders start making proactive, reason-based decisions before the damage is done. - [The Software Factory: AI-Driven Development Without Human Code Review](https://pablooliva.de/the-closing-window/the-software-factory-ai-driven-development-without-human-code-review/): StrongDM's AI team eliminates human code writing and review entirely, instead using coding agents validated by holdout scenario tests and agent-built clones of third-party APIs to prove software works at scale. - [You Spent Your Whole Life Getting Good at the Wrong Thing](https://pablooliva.de/the-closing-window/you-spent-your-whole-life-getting-good-at-the-wrong-thing/): AI agents are collapsing execution into intention, shifting the bottleneck from "how" to "what" — making taste, judgment, agency, and curiosity the new core skills, though your existing domain expertise remains essential as the quality filter for AI output.