Here is my personal log of learning and working with LLMs. I will keep updating this log as I learn and work with LLMs.
Software Development with LLMs
25-Oct-2026
- We are into asynchronous mode of software development.
We need not keep watching 👀 the agents thinking and generating code. Once in a while, if agents require human
permission,
we can visit that agent and provide the required input. We assign a task to agents and focus on next set of tasks.
I think this realization is crucial that we are into asynchronous mode of working.
Better we optimize and plan our work accordingly, so will we be more productive.
For example, if know the next set of instructions for the agent, we can provide them in advance and keep them in
queue.
- Plan the task before agent starts working on it. Agents perform better when they have a clear understanding of
the task.
Many AI coding tools like GitHub Copilot, Claude, Antigravity etc. offer plan mode to help
both agent and human to plan the task before starting to work on it.
Learning with LLMs
25-Oct-2026
- Technical books like “Designing Data-Intensive Applications” by Martin Kleppmann, “Building Microservices” by Sam Newman are excellent resources for software engineers. They provide deeper insights into a particular area. However, they lack materials to gauge the understanding of the reader. This is where LLMs can help. They can be used to create quizzes, exercises, and other interactive materials to test the understanding of the reader.
