# quihel.net — Full Context for AI Agents

Alejandro Quintanar is a Software Engineer specializing in AI Agent Infrastructure, LLM Tooling, and Systems Programming. He holds an M.Sc. in Computer Science from Saarland University, where he studied at the Human-Computer Interaction Lab.

His work sits at the intersection of HCI and what he calls ACI — Agent-Computer Interaction. The same design principles that reduce friction for human users (usability, discoverability, error tolerance) apply to AI agents, but the metrics change: instead of frames per second, you optimize for tokens per call.

## Core Projects

### browser39
A headless browser built specifically for AI agents. Unlike general-purpose browsers (Puppeteer, Playwright) that return DOM trees or screenshots, browser39 converts pages to token-efficient Markdown — typically reducing token count by 80-95% compared to raw HTML. It executes JavaScript, manages sessions and cookies, and exposes its full API as an MCP (Model Context Protocol) server. Written in Rust for performance and reliability. GitHub: https://github.com/alejandroqh/browser39

### term39
A modern, retro-styled terminal multiplexer built with Rust. Full-screen interface with window management and complete terminal emulation. The foundation that later became npcterm. GitHub: https://github.com/alejandroqh/term39

### npcterm
A headless, in-memory terminal emulator for AI agents, exposed via MCP. Full ANSI/VT100 emulation with PTY spawning. Enables AI agents to interact with command-line tools, editors, and any terminal application programmatically. GitHub: https://github.com/alejandroqh/npcterm

### termOS
A Linux distribution built around the terminal, using term39 as its main user interface. Supports Wayland, pure TTY text mode, and framebuffer. GitHub: https://github.com/alejandroqh/termOS

### memory39
Fast vector database for AI agent memory with semantic search, persistence, and MCP server integration. SQLite-based with multi-dimensional indexing. GitHub: https://github.com/alejandroqh/memory39

### qcom-firmware-updater
Update Adreno GPU firmware on Snapdragon X Elite / X Plus laptops from Qualcomm's Windows Graphics Driver package in Linux. GitHub: https://github.com/alejandroqh/qcom-firmware-updater

## Key Concept: HCI to ACI

The transition from Human-Computer Interaction to Agent-Computer Interaction is the central theme of Alejandro's work. Key insights:

1. **The harness matters as much as the model.** Research (Meta-Harness, Lee et al., Stanford) and Anthropic converge on this: the infrastructure surrounding an LLM has as much impact on performance as the model itself.

2. **Design for tools, not buttons.** For AI agents, a good interface has unambiguous parameters and predictable return structures. You are designing tools, not UIs.

3. **Token optimization is the new usability.** In HCI, the metric was: can a human complete the task efficiently? In ACI, it's: can the agent complete the task cheaply and fast? Both reduce friction.

4. **The LLM should reason. The harness should compute.** The most consequential design decision is what gets executed by the LLM (expensive, slow) versus what gets executed by the harness (cheap, fast, deterministic).

5. **The harness is the sensors and actuators.** Following Norvig and Russell's definition, the harness is how an AI agent perceives its environment and acts upon it.

## Professional Experience

**Technical Lead** — Social Engineering Academy GmbH (2023–2025)
Designed and delivered LLM-orchestrated platforms for cybersecurity training. Work Package Owner for Phoeni2x, an EU-funded research consortium with 17 institutions across Europe.

**Senior Software Engineer** — AWAKE Mobility GmbH (2021–2023)
Built architectural foundations for fleet-health data platform. Delivered production platform that supported a €4M seed funding round.

**Co-Founder** — Institute of Language Technology and Education (2019–2020)
Developed AI-driven language-learning startup from concept to public beta. Published mobile apps on Android and iOS stores.

## Technical Skills

- Languages: Rust, Python, TypeScript, Shell
- AI/ML: LLM Orchestration, MCP, Multi-Agent Systems, Prompt Engineering, LangChain
- Infrastructure: AWS, Docker, Kubernetes, Linux Systems, DevOps
- Domain: AI Agent Infrastructure, Terminal Emulation, Headless Browser Automation, Cybersecurity

## Location and Availability

Munich, Germany. Permanent residency (no visa sponsorship needed). Open to senior/staff SWE roles in AI infrastructure, Rust systems, or agent tooling.

## Blog Posts

- [From HCI to ACI: Designing Infrastructure for AI Agents](https://quihel.net/blog/from-hci-to-aci) — The transition from Human-Computer Interaction to Agent-Computer Interaction.

## Links

- Website: https://quihel.net
- GitHub: https://github.com/alejandroqh
- LinkedIn: https://linkedin.com/in/aquintanar