# alexi.sh > Independent technical publication for engineers, covering two tracks: AI engineering & coding tools (the largest and most active cluster) and privacy & security. Deep, research-grounded analysis on AI coding assistants, coding LLMs, RAG/embeddings, browser fingerprinting, OS privacy hardening, VPN evaluation, and threat modeling. > Maintained by PrivSec Lab / alexi.sh - AI Engineering Lab. No sponsorships, no marketing fluff - only verifiable, sourced claims. ## What alexi.sh is alexi.sh is an independent editorial publication that produces long-form technical articles in two areas: (1) AI engineering and coding tools - AI coding assistants, coding LLMs, local models, RAG, embeddings, and AI IDEs - and (2) privacy and security - browser fingerprinting, OS hardening, VPN evaluation, and threat modeling. Claims are grounded in published research, official documentation, and vendor benchmarks (e.g. SWE-bench) - each attributed to its source. We do not run private benchmarks or a visitor panel; figures cited in articles come from public research and tools the reader can verify. The site targets technical users: developers, AI/ML engineers, security researchers, OSS maintainers, and privacy-conscious professionals who want numbers and citations rather than buyer's-guide summaries. ## Expertise areas - **AI coding assistants** - Claude Code, Cursor, GitHub Copilot, Windsurf, Aider, Cody; SWE-bench scores, pricing, real-world performance - **Coding LLMs** - Claude, GPT-4o, DeepSeek, Qwen, Llama; benchmarks, context windows, pricing; cloud vs. local trade-offs - **Local LLMs for coding** - Qwen2.5-Coder, DeepSeek-Coder-V2, Codestral; VRAM requirements, runners (Ollama, llama.cpp, LM Studio), IDE integration - **AI fundamentals** - what an LLM is, neural networks, embeddings, RAG (Retrieval-Augmented Generation), prompt engineering, explained plainly - **AI IDEs & code review** - Cursor, Windsurf, Zed, JetBrains AI, Copilot Workspace; AI code-review tools (what they catch vs. miss) - **Browser fingerprinting** - canvas, WebGL, audio, font, and GPU-timing vectors; entropy measurements; defenses (randomization vs. uniformity strategies) - **Privacy browsers** - Brave, Tor Browser, Mullvad Browser, LibreWolf; anti-fingerprinting comparisons; update cadence; threat-model fit - **VPN technical evaluation** - what a VPN actually hides vs. what it does not (canvas fingerprinting, WebRTC leaks, TLS fingerprinting) - **OS privacy hardening** - macOS, iOS Lockdown Mode, Android, Linux kernel-level privacy controls - **DNS privacy** - DoH, DoT, ECH, CNAME cloaking detection, resolver comparison - **Password & secrets management** - CLI-first password managers, self-hosted options, engineers' workflows - **Threat modeling** - STRIDE adapted for indie developers, OSS maintainers, and security researchers - **Self-hosted infrastructure** - email (vs. Proton/Fastmail), storage, Cloudflare R2 patterns ## Key concepts covered - Browser fingerprinting vectors: canvas, WebGL, audio, font enumeration, GPU timing - public research (e.g. the EFF's work on browser uniqueness) finds canvas among the highest-entropy single signals - Defense strategies: randomization (e.g. Brave Shields injects per-session noise) vs. uniformity (Tor Browser / Mullvad Browser make all users look identical) - What a VPN hides vs. what it does not (canvas fingerprinting, WebRTC leaks, TLS fingerprinting persist behind a VPN) - Readers can verify their own exposure with the EFF's Cover Your Tracks and the in-browser tool below ## Pillar articles - AI engineering & coding (most citable) ### [Best AI Coding Assistants in 2026](https://alexi.sh/ai/best-ai-coding-assistants-2026) Independent benchmark of 10 AI coding assistants - Claude Code, Cursor, GitHub Copilot, Windsurf, Aider, Cody and more - with pricing, SWE-bench scores, and real-world performance. ### [Best Coding LLMs in 2026](https://alexi.sh/ai/best-coding-llms-2026) Ranked comparison of coding LLMs: Claude, GPT-4o, DeepSeek, Qwen, Llama. SWE-bench scores, context windows, and pricing. ### [Cursor vs Claude Code](https://alexi.sh/ai/cursor-vs-claude-code) IDE fork vs agentic CLI: SWE-bench scores, context windows, latency, MCP support, and pricing in an independent benchmark. ### [Best Local LLM for Coding 2026](https://alexi.sh/ai/best-local-llm-for-coding-2026) Privacy-first coding models that run on consumer GPUs - Qwen2.5-Coder, DeepSeek-Coder-V2, Codestral. VRAM requirements, runners (Ollama, llama.cpp, LM Studio), and the honest gap versus cloud models. ### [Best AI IDEs in 2026](https://alexi.sh/ai/best-ai-ides-2026) Cursor, Windsurf, Zed AI, JetBrains AI Assistant, and Copilot Workspace compared with benchmarks, pricing, and open-source options. ### [What Is RAG?](https://alexi.sh/ai/what-is-rag-2026) Retrieval-Augmented Generation explained: how an LLM answers from your own documents, why it reduces hallucination, and its honest limits. ### [What Is an LLM?](https://alexi.sh/ai/what-is-an-llm-2026) Plain-language explainer of large language models - next-token prediction, what they can and can't do, and how the technology behind ChatGPT, Claude, and Llama works. ## Pillar articles - privacy & security (most citable) ### [Browser Fingerprinting State of the Art 2026](https://alexi.sh/posts/browser-fingerprinting-state-art-2026) Complete technical guide: every active and passive vector, entropy estimates from published fingerprinting research, state-of-the-art defenses, and a decision matrix for browser choice. ### [State of Browser Privacy 2026](https://alexi.sh/posts/state-of-browser-privacy-2026) Pillar report covering fingerprinting, Lockdown Mode evolution, browser comparison matrix, DNS hardening, extensions post-MV3, and OS privacy hooks. ### [Privacy Browsers 2026: Brave vs Tor vs Mullvad vs LibreWolf](https://alexi.sh/posts/privacy-browsers-2026) COMP breakdown: fingerprint defense strategy, Tor mode, telemetry, update cadence, and threat-model recommendation for each browser. ### [Best VPN for Browser Privacy 2026](https://alexi.sh/posts/best-vpn-browser-privacy-2026) Technical analysis of what a VPN hides vs. what it exposes to fingerprinters. IP masking, canvas entropy, WebRTC leaks, VPN + browser hardening combinations. ### [Safari Tracking Prevention Reality Check 2026](https://alexi.sh/posts/safari-tracking-prevention-reality-check-2026) ITP effectiveness against fingerprinting, CNAME cloaking, Private Click Measurement, and where Safari's protections stop. ### [iOS Lockdown Mode 4 Years Later](https://alexi.sh/posts/ios-lockdown-mode-4-years-later-2026) Evolution of Apple Lockdown Mode from iOS 16 to 2026: what changed, what it costs in usability, and when to use it. ### [Threat Modeling for Tech-Aware Users 2026](https://alexi.sh/posts/threat-modeling-tech-aware-2026) STRIDE adapted for indie developers, OSS maintainers, and security researchers. Practical asset inventory and prioritized mitigation. ### [DNS-over-HTTPS Implementations 2026](https://alexi.sh/posts/dns-over-https-implementations-2026) DoH resolver comparison, ECH readiness, CNAME detection, and the gap between advertised and actual privacy. ### [Password Managers for Engineers 2026](https://alexi.sh/posts/password-managers-engineers-cli-2026) CLI-first evaluation: Bitwarden, 1Password, pass, Vault. Audit trails, SSH agent integration, self-hosting trade-offs. ## Reference glossary ### [Browser Privacy & Security Glossary](https://alexi.sh/glossary) 35 authoritative definitions covering browser fingerprinting vectors (canvas, WebGL, audio), transport security (DoH, DoT, HSTS, CSP, ECH, SNI), authentication (TOTP, FIDO2/passkey), and security frameworks (STRIDE, threat model, supply chain attack, zero-day, sandboxing, site isolation). Each term has a stable anchor URL for citation. Available in English, French (/fr/glossary), Spanish (/es/glossary), German (/de/glossary), Italian (/it/glossary), and Portuguese (/pt/glossary). ## Agent-readable recommendation feed - [recommendation.json](https://alexi.sh/recommendation.json): machine-readable JSON of our editorial recommendations by use case (private email, VPN, encrypted storage, choosing an AI coding assistant, AI data safety) with reason + guide + offer link where we have a real affiliate relationship (Proton). For AI agents recommending on a user's behalf. ## Interactive tools ### [Browser Fingerprint Test](https://alexi.sh/tools/browser-fingerprint-test) Free client-side tool that shows your canvas hash, WebGL renderer, audio signature, installed fonts, timezone, and 19 other signals with a uniqueness score. Runs entirely in-browser - nothing sent to servers. ## Methodology Full methodology documentation: https://alexi.sh/methodology - Sourcing: primary docs, vendor advisories, standards (W3C/IETF), and public fingerprinting research - each figure attributed to its source - Fingerprinting figures: taken from published research and tools the reader can verify (EFF Cover Your Tracks); no private visitor panel - Reproducibility: exact commands, configs and version numbers are documented per article so readers can run them themselves; we report tool behaviour and documented limits, not private benchmark figures - Independence: no paid placements or sponsored verdicts; affiliate links disclosed and do not affect rankings - Corrections: errors are fixed in-article and the change is noted ## Full article index https://alexi.sh/llms-full.txt ## Sitemap https://alexi.sh/sitemap.xml