The field notes / 11 articles
Thinking,
in public.
Engineering notes, independent builds and questions worth revisiting.
Looking Beyond the Frontend Boundary
An interface is part of a larger system. A reflection on dependencies, design contracts and the independent projects that make those connections visible.
3 min readSeven Sources, One Adapter Pattern: Lakshya’s Unified Job Search
A build note on normalising job sources, keeping fit scores with saved results and preserving search state across the interface.
5 min readFRIDAY Phase 4: An Arena for Comparing Model Choices
An independent agent design for model routing, budget boundaries, scheduling and a visible record of decisions. A build note, not a cost benchmark.
4 min readInside SuperAgent’s Cost-Aware Routing Brain
Intent, skill chains, cross-session memory and model choices: how SuperAgent tries to make AI coding workflows more deliberate and inspectable.
4 min readWhy a Lighthouse Score Is Not the Whole Performance Story
Rendering speed, interaction and reliability belong to the same user journey. Use a performance score to investigate that journey, not to stand in for it.
2 min readThe Approval Gate Inside insanemesh.ai’s Content Pipeline
A workflow from idea to rendered asset, with Telegram review before publishing. How distinct stages made the automation easier to understand.
2 min readWhat Production Work Teaches Me That Side Projects Don't
Production work and independent projects teach different lessons. A reflection on stability, rollout decisions and the user behind the interface.
2 min readLakshya V2: Building the Job Search OS I Wish Existed When I Was Hunting
Connecting job discovery, résumé fit and drafting in an independent Next.js and Supabase workflow, while keeping the applicant in control.
2 min readHLD Before Keyboard: Make the Problem Clear Before the Code
The expensive mistake is correct code that solves the wrong problem. Design planning makes dependencies, contracts and failure paths discussable.
2 min readRAG in My Local Lab: Retrieval, Long Context and the Evidence Between
Notes from local document experiments with Ollama, Chroma and LlamaIndex, and why context assembly needs evaluation rather than a universal recipe.
2 min readThe Staff Engineer Interview: Beyond the Coding Exercise
Preparing for broader engineering roles means practising decisions, dependencies and trade-offs as well as code.
2 min read