I trapped your soul in a trading card (with client-side AI)
My most elaborate side project yet!
RAG in Coding Agents: Making Smarter Programming Assistants
By combining retrieval and generation, it enables coding assistants to understand tasks, write solutions, and debug errors—all while adapting to your needs.
RAG is the way about retrieval, agent, and grounding truth
I have been working on RAG for a while, and I think I have a few things to share.
Automating Persona Enriched Prompting
Persona Enriched Prompting is technique to instruct an LLM to generate knowledge about the inferred domain and then generate UX personas for that domain contextual to the problem. Once this context is generated, you use that for final answer generation.
Disabling Intel Graphics Security Mitigations Can Boost GPU Compute Performance By 20%
While not talked about as much as the Intel CPU security mitigations, Intel graphics security mitigations have added up over time that if disabling Intel graphics security mitigations for their GPU compute stack for OpenCL and Level Zero can yield a 20% perfor…
Serialization From NodeJS
2025-06-22
My most frequently used Jujutsu VCS commands
A cheat-sheet of jj commands you're most likely to need day-to-day using jj as a git replacement.
Baby boss
Think like a kid. Be a kid. Kids are very successful at making friends. Kids are very successful at making discoveries. Literally, every new day in a kid’s life is a new beginning, a new discovery....
Forking dangerous
TL;DR Python 3.14+ now duplicates processes with spawn over fork spawn duplicates the whole process space fork clones the process but doesn’t duplicate the thread space The following deadlocks (on linux, <py3.14) import threading import time fro...
Building a language server · Bull no one
The beginningStory ahead, skip to the next section for actual technical writing.The uxn ecosystem is my current rabbi...
Follow basics
When I started my career, I was eager to make a difference. I tried to be everywhere—attending every meeting, volunteering for every opportunity, and using all my free time to contribute beyond my...
How much slower is random access, really?
You may know that, because your computer has different caches (L1, L2, L3...), and memory operations operate on cache lines of about 64 bytes each, you should write programs that exhibit locality to get maximum performan…
How Long Contexts Fail
Taking care of your context is the key to building successful agents. Just because there’s a 1 million token context window doesn’t mean you should fill it.
The Paradox of Infinite Context
Why infinite context makes retrieval more important, not less.
How text diffusion works
None
rkayg
my personal site
Take a Break — haykerman
Vibe Learning to Fearlessly Explore Unfamiliar Tech | Kaveh's Blog
I vibe learned and vibe coded a Speech2Text GNOME extension in a weekend. It was an eye-opening experience.
Interesting Bits of Postgres Grammar
Lessons from building a parser
GitHub - astral-sh/uv: An extremely fast Python package and project manager, written in Rust.
An extremely fast Python package and project manager, written in Rust. - astral-sh/uv
Backyard Coffee And Jazz In Kyoto, Japan
A "magical" experience, and what we get when we let people pursue their passions
You can't use switch expressions in Blazor
Did you ever try to have a switch expression in Blazor? Yes, and did it go well? Probably not. Let's see why!
Why 51% of Engineering Leaders Believe AI Is Impacting the Industry Negatively
Many companies are experiencing FOMO when it comes to AI and it has a negative effect on engineering leaders. This is why!