Welcome to asklar.dev—my corner of the internet for longer-form thoughts on AI, agentic systems, and what it’s really like building the future at the operating system level.
Why This Site?
Over the past year, I’ve been sharing insights on LinkedIn about my work on Windows as an agentic operating system, the Model Context Protocol (MCP), and the challenges of integrating AI at the platform level. Those posts have been great for sparking conversations, but there’s only so much you can say in 250 words.
This blog is for the ideas that need more room to breathe:
- Deep technical explorations: When I want to walk through architectural decisions, implementation challenges, or design patterns that can’t fit in a LinkedIn post
- Longer-form analysis: Industry trends, framework comparisons, and broader implications that deserve more than bullet points
- Building in public: Documenting projects, experiments, and lessons learned along the way
- Synthesis and reflection: Connecting dots across different areas—from OS architecture to AI agent design to developer experience
What to Expect
I’m a Principal Software Engineer at Microsoft working on making Windows an agentic operating system. My work focuses on:
- Platform-level AI integration: How do you make an entire OS AI-native without compromising security, performance, or user trust?
- Model Context Protocol (MCP): Building the infrastructure that lets AI agents safely interact with applications and system resources
- Agentic workflows: The practical realities of agents that actually work at enterprise scale, not just demos
This blog will reflect those interests—expect technical deep dives, architectural discussions, and honest takes on what’s working (and what isn’t) in the world of agentic systems.
Building Frontier Firms
Microsoft’s Work Trend Index describes 2025 as the year the frontier firm is born—organizations that fundamentally reimagine how work gets done with AI at the core. This isn’t just about adopting AI tools; it’s about cultural transformation at every level.
For engineering and product teams, this transformation is both technical and organizational. We’re not just building features that use AI—we’re rethinking entire workflows, development processes, and product experiences. The infrastructure we build today determines whether our organizations can become frontier firms or get left behind.
My work on agentic Windows and MCP is directly in service of this transformation. By making the OS itself AI-native, we’re enabling every application and every team to leverage agentic capabilities without having to rebuild the foundational infrastructure themselves. This blog will explore both the technical challenges of that work and the broader implications for how teams can embrace this shift.
I want to be a propellant for this transformation—sharing what we’re learning, the patterns that work at scale, and the honest realities of what it takes to make the leap. The frontier firm isn’t built in a day, but it starts with teams willing to experiment, fail fast, and share their learnings openly.
The Format
Most posts here will expand on ideas I’ve shared on LinkedIn. When I publish here, I’ll share a condensed version on LinkedIn with a link back for those who want the full story. Think of LinkedIn as the teasers and this blog as the director’s cut.
I’ll also use this space for:
- Project documentation: Open source work like McpExtract and other tools I’m building
- Tutorial content: When there’s a gap in documentation or I find myself explaining the same thing repeatedly
- Research notes: Explorations that might not have a clear conclusion yet but are worth sharing
A Work in Progress
Like any good software project, this site will evolve. The focus areas might shift, the format might change, and I’ll probably experiment with different types of content along the way.
What won’t change is the commitment to writing about the practical realities of building systems that integrate AI—not just the theory or the marketing, but the actual trade-offs, constraints, and decisions that go into shipping production code.
Stay Connected
If you found your way here from LinkedIn, thanks for following along. If you’re new, welcome! Here’s where you can find me:
- LinkedIn: linkedin.com/in/asklar - Where I share shorter insights and industry commentary
- GitHub: @asklar - where the actual code lives
- This blog: For everything that needs more than a few paragraphs
I’m always interested in connecting with fellow engineers, researchers, and builders who are working on similar problems. Whether you’re implementing AI features in your own applications, building agent frameworks, or just trying to make sense of where this technology is headed—let’s talk.
Drop a comment, open an issue on one of my projects, or reach out on LinkedIn. The best insights come from conversations, not monologues.
What’s Next
I have a backlog of topics I’ve been wanting to write about:
- How we’re thinking about security for agentic Windows
- Deep dive into MCP server architecture and design patterns
- Lessons from building tools over almost two decades at Microsoft—what worked, what didn’t, and why
- The future of UI when agents can interact directly with applications
- Trade-offs in different approaches to on-device AI
Some of these will happen soon, others will take time to marinate. Subscribe (or check back periodically) if any of that sounds interesting.
Thanks for being here at the beginning. Let’s see where this goes.
—Alex