Notes from the team.
Product updates, engineering write-ups, and research from the Reclit team.
· Muhammad Haroon · 1 min read
Virtual Column Reordering: How We Move Millions of Cells in O(1) Time
Moving a column in a huge spreadsheet shouldn't mean mutating thousands of rows in memory. In Reclit, we reorder virtual column IDs, keeping physical cell data untouched.
Read the post· Muhammad Haroon · 3 min read
Building Spreadsheets on Canvas: How Reclit Renders Thousands of Rows
Building a spreadsheet that can handle thousands of rows seamlessly requires moving away from the DOM. In Reclit, we leverage HTML5 Canvas and 2D arrays to create a highly performant, AI-navigable grid.
Read the post· Muhammad Haroon · 3 min read
Context Engineering: The Real Challenge Behind Reliable AI
We invented the phrase “context engineering” only months ago. Turning it from prompt hacking into something closer to actual engineering means better models of what the context is doing—and serious investment in the outer loop of data.
Read the post· Muhammad Haroon · 2 min read
Solving Hard Problems in Complex Codebases
Complex codebases need more than better prompts. Research, planning, context management, and human thinking help coding agents solve harder problems with less slop.
Read the post· Muhammad Haroon · 3 min read
Reducing AI Hallucinations: From Prompting to Verification
Reducing AI hallucinations requires more than better prompts. Grounding, tool use, evidence, and deterministic verification provide stronger reliability.
Read the post· Muhammad Haroon · 3 min read
Harness Engineering: Turning AI Intelligence into Reliable Systems
AI models are becoming more capable, but intelligence alone does not guarantee reliable results. Harness engineering adds the tools, guardrails, context, and verification needed to turn AI reasoning into dependable action.
Read the post