Tools
A curated collection of tools I have explored and worked with over time.
22 items
Sort by
2026
Claude Code Skills Favorite
Reusable, project-scoped AI instructions that turn common multi-step tasks into one-command workflows. Define a skill once in Markdown, then invoke it with a slash command — like this post, which was drafted by one. Best for: Automating repetitive medium-sized tasks: content generation, audits, scaffolding, and any workflow that follows a consistent pattern across invocations.
Not for: One-off tasks or problems that require significant creative judgment each time — skills shine when the process is repeatable.
Notes: A genuine productivity multiplier. I built skills for generating new portfolio posts, running accessibility audits, and producing llms.txt — each takes seconds to invoke and produces consistent, schema-valid output. The key insight is that skills encode process knowledge, not just prompts.
(opens in new tab)
2026
PowerToys Command Palette Favorite
A fast, keyboard-driven launcher for Windows from Microsoft PowerToys. Quickly access commands, shortcuts, files, and even open VS Code projects — all without leaving the keyboard. Best for: Quick access to commands, keyboard shortcuts, and launching VS Code projects without touching the mouse.
Not for: macOS or Linux users — it's Windows-only. Spotlight or Albert cover similar ground on other platforms.
Notes: One of the first things I install on any Windows machine. The command palette alone justifies PowerToys, but the full suite has other useful utilities too.
(opens in new tab)
2024
uv Favorite
A lightning-fast Python package installer and project manager written in Rust. Combines pip, venv, and project management into a single tool with dramatically faster dependency resolution. Best for: Pure Python projects.
Not for: Projects with complex or multi-language dependencies.
Notes: Replaced pip, venv, and hatchling in my projects. Dependency installation is incredibly fast, and all configuration is centralized in pyproject.toml.
(opens in new tab)
2023
Tailscale Favorite
Personal VPN that’s simple, secure, and powerful. Best for: Securely accessing infrastructure and connecting multiple devices without traditional VPN complexity.
Not for: Users who don’t need secure cross-device connections.
Notes: Used for secure remote access to GPUs, deployment servers, and workstations. No traditional VPN setup required; very easy and secure.
(opens in new tab)
2026
Context7
An MCP server that fetches up-to-date library documentation on demand, injecting current docs directly into your AI coding assistant's context to reduce hallucinations about outdated or deprecated syntax. Best for: Keeping AI coding assistants accurate when working with fast-moving libraries where training data lags behind the current API.
Not for: Internal or private documentation, or stable APIs where the AI's training data is already reliable.
Notes: AI coding assistants often frustrate me with outdated syntax. `Context7` addresses this by inspecting your dependencies, fetching the correct version of the documentation on demand, and injecting it into the model’s context. This enables accurate answers even for fast-moving libraries. It effectively acts as a real-time, version-aware syntax reference. In theory, it should be faster than browser-based agents that fetch and parse HTML docs.
(opens in new tab)
2026
GitHub Agentic Workflows
AI agents as CI infrastructure — running on every push to enforce architectural constraints, validate accessibility compliance, detect stale documentation, and check API contracts. An adaptive test suite that goes beyond what unit tests can cover. Best for: Teams that want guardrails running automatically on every push — enforcing WCAG compliance, catching stale docs, and validating API contracts without anyone waiting or reviewing manually.
Not for: Tasks requiring judgment calls, architectural context, or nuanced decisions that go beyond what can be inferred from diffs and repo state alone.
Notes: The framing that made this click: the agent is infrastructure, not a chatbot. It behaves like an adaptive test suite — beyond traditional unit tests, it can enforce architectural constraints, validate accessibility compliance, detect stale documentation, flag risky migrations, and check API contracts. No one waits. It runs on every push and acts as an automated guardrail.
(opens in new tab)
2026
llms.txt
A proposed standard for making website content AI-discoverable. Sites serve a structured Markdown file at /llms.txt that summarizes what the site offers, so LLMs can quickly understand and reference the content without crawling entire pages. Best for: Small, structured cheatsheets and concise information summaries that LLMs can consume directly.
Not for: Complex, multi-layered information that benefits from interactive queries — consider MCP servers for that.
Notes: Implemented on this site at jasonlo.dev/llms.txt. A great first step toward AI discoverability before committing to a full MCP setup. Simple to maintain and surprisingly effective for giving LLMs context about your content.
(opens in new tab)
2025
GitHub Copilot Coding Agent
A cloud-based AI agent that autonomously works on GitHub issues and creates pull requests. Assign an issue, walk away, and come back to a ready-to-review PR. Best for: Addressing multiple issues in parallel without babysitting — the agent runs in the cloud, so you can focus on other work while PRs are being drafted.
Not for: Complex architectural changes or tasks requiring deep domain context that the agent can't infer from the codebase alone.
Notes: The biggest win is time efficiency. Unlike local coding agents where you watch them work, Copilot Agent runs on cloud infrastructure and delivers a PR. You can fire off several issues at once and review the results later. But the performance varies.
(opens in new tab)
2025
Cursor
An AI-first code editor built on VS Code. Best for: Interactive coding with AI assistance, rapid prototyping.
Not for: Developers who prefer strictly local, non-AI workflows or those with privacy concerns about code being sent to external AI services.
Notes: Most features have already been adopted by VSCode, making it less compelling.
(opens in new tab)
2025
Zed
A high-performance code editor built in Rust, designed for speed and minimal resource usage. Supports real-time collaboration out of the box. Best for: Fast, lightweight coding sessions where startup time and responsiveness matter more than plugin breadth.
Not for: Workflows that depend on a rich extension ecosystem — the plugin library is still limited compared to VS Code.
Notes: Tried it briefly. It definitely feels faster than VS Code, but the lack of many useful plugins I rely on kept me from switching.
(opens in new tab)
2025
Claude Code
One of the leading AI-powered coding assistants, designed to support software engineering tasks—from debugging and refactoring to implementing new features—by understanding the codebase deeply and enabling interactive development. What you see here was created by Claude Code in a single Saturday afternoon. Best for: Rapid feature implementation and prototyping for small to medium-sized problems.
Not for: Fully autonomous development, complex architectural decisions, or entirely novel research code.
Notes: Integrated into my development workflow for this portfolio and other projects. Excellent for understanding unfamiliar codebases, generating boilerplate, and working through implementation details in common problems.
(opens in new tab)
2025
Gemini 2.0
Google's AI model suite, offering top-tier performance and long context windows at the time of release. Best for: Complex reasoning and understanding long-form text.
Not for: Offline-first applications or fully private inference without cloud access.
Notes: Still one of the top-performing model families in early 2026, though adoption is lower than ChatGPT.
(opens in new tab)
2025
pixi
A cross-language package manager for Python, R, C++, and more, primarily using conda-forge. Built for reproducible environments and scientific computing workflows. Best for: Polyglot projects combining Python with C++, Rust, or R, complex scientific workflows, and ensuring reproducible environments across teams. Example: molecular dynamics simulations.
Not for: Simple, single-language Python projects.
Notes: Great for projects requiring more than Python dependencies. The lockfile-based system ensures reproducibility across machines and CI/CD pipelines.
(opens in new tab)
2025
Capacities
Feature rich, object-based note-taking and knowledge management tool, similar to Notion. Best for: Structured knowledge management, daily notes with objects, and building interconnected information systems.
Not for: Simple note-taking needs or users who prefer pure Markdown workflows.
Notes: Explored this as an alternative to Obsidian in 2025. It has many appealing features but is arguably too complex and lacks clear focus.
(opens in new tab)
2024
Milvus
An open-source vector database built for scalable similarity search and AI applications. Purpose-built for high-performance vector operations with production-ready tooling including VTS, horizontal scaling, and multi-replica deployments. Best for: RAG pipelines, similarity search, and recommendation systems where vector query performance matters at scale.
Not for: Workloads requiring complex relational joins — pair it with a traditional relational database for those queries.
Notes: My go-to choice when I need best-in-class vector database performance. Production-ready out of the box with solid operational tooling. Fast, scalable, and reliable for real workloads.
(opens in new tab)
2023
vLLM
A high-throughput, memory-efficient inference engine for large language models, optimized for concurrent requests and top-tier performance. Best for: High-performance LLM serving, offline batch inference, API deployment, and applications needing low-latency responses at scale.
Not for: Simple single-request testing or users who prefer easy setup over performance.
Notes: Used for LLM inference servers. Setup is more complex, but performance gains are significant.
(opens in new tab)
2023
Ollama
A simple model server designed for a smooth developer experience. Best for: Local AI experimentation, offline LLM inference, privacy-focused AI applications, and testing open-source models without API costs.
Not for: Production deployments requiring top performance.
Notes: Ideal for experimenting with open-source models and evaluating performance on your hardware. The CLI is straightforward, and the Dockerfile-like Modelfile is a clever touch. Recent shift toward cloud-based models reduces its original offline advantage.
(opens in new tab)
2022
.devcontainer
A standardized development container that ensures reproducible environments using Docker, fully integrated with VSCode and GitHub Codespaces. Best for: Team projects requiring consistent environments, fast onboarding of new developers, avoiding 'works on my machine' issues, and complex projects with multiple dependencies.
Not for: Simple Python projects that only need basic dependency management.
Notes: Essential for team projects—setting up a .devcontainer once saves hours across the team. With GitHub Codespaces, environment setup friction is eliminated. VS Code integration is seamless, with extensions and tools ready to use consistently. Initial setup can be slow with large Docker images and may be finicky at times.
(opens in new tab)
2022
Obsidian
A flexible, Markdown-based system for personal note-taking and knowledge management, with extension support, ideal for creating a well-organized knowledge base using the Zettelkasten method. Best for: Personal knowledge management, linking ideas, and long-term documentation.
Not for: Team collaboration or real-time collaborative editing.
Notes: Used for personal notes in 2022 for a year or so. Its local-first approach and Markdown-based files ensure full control over my data. However, sharing notes isn't seamless, particularly for non-technical users.
(opens in new tab)
2020
Notion
All-in-one collaborative workspace combining documents, databases, wikis, and project management with AI-powered features. Best for: Team collaboration, knowledge bases, project management, and teams needing a centralized workspace for multiple use cases.
Not for: Simple note-taking.
Notes: Used extensively from 2019–2021 for personal notes. Offers powerful, database-like features and high flexibility, but the block-based editor and reliance on the cloud led me to move personal notes to Obsidian. Still one of the best wiki/knowledge management tools for team collaboration, where its database capabilities truly shine.
(opens in new tab)
2020
Visual Studio Code
Primary code editor with a strong extension ecosystem and a large, active user community. Best for: General purpose coding.
Not for: Minimalist.
Notes: Been using it since 2020. The extension ecosystem keeps improving, though it can feel bulky and slower compared to newer editors like Zed.
(opens in new tab)
2018
Google Docs
Cloud-based word processor with real-time collaboration and easy sharing. Widely known and easy to use. Best for: Collaborative editing in real time, seamless document sharing, and cloud-based writing accessible from any device.
Not for: Offline-first workflows, advanced formatting needs, or users who prioritize privacy over cloud storage.
Notes: Essential for team collaboration. Very easy for anyone to use, with features that consistently work well.
(opens in new tab)