Writing
Essays about systems I built and what real workloads changed in them, lab notes from tiyuvta, and talks. I don't turn release notes into thought leadership.
Essays
Edited editions of essays first published on DEV.
-
5 min read
How AI workloads changed the queue I was already building
Why streaming, spend limits, token-aware rate limits and human pauses moved into glide-mq instead of living in a second system around it.
-
5 min read
Your AI agent configs are probably broken (and you don't know it)
Agent configuration fails silently across skills, hooks, memory, MCP and tool-specific formats. The case for treating that layer like code.
-
tiyuvta.ai
I thought NVFP4 broke MTP
I checked whether NVFP4 broke Qwen3.5-9B's co-trained MTP head. The first result was a measurement trap; the clean comparison found no loss.
-
tiyuvta.ai
Masked MTP drafts
Acceptance went down and tokens per second went up. Trimming the draft head released compute on a compute-bound run.
Talks and elsewhere
The same systems, with room for the questions that do not fit in an article.
-
Inside Valkey GLIDE
Rust at the core of a multi-language client, zero-config ergonomics, socket security and community feedback.
-
Glide into resiliency
Practical Valkey GLIDE patterns: client reliability, Pub/Sub, vector data, sessions and real integration examples.
-
Stack Overflow answers
Redis and Valkey, Node.js, AWS, GitHub Actions, TypeScript, and the bugs that only show up in someone else's setup.
-
DEV Community
The original editions of the essays, with comments and discussion.