Notes on building with LLMs and coding agents
- I had Codex and Pi build the same app in Orca, then compared them side by side
Two coding-agent teams, one chess coach, matched conditions. Every review passed both; side by side, the gaps were large.
- Can Open Source Coding Agents Build Comparable Software for Less?
Coding agents such as Claude Code and Codex can handle much of the work from planning through testing, but heavy model usage can be expensive.
- Rethinking Software Design Principles for Agent Product Development
Classical principles still hold, but their meaning shifts when the reasoning path is controlled by the model, not the code.
- Disrupted or Defensible: Business Models in the LLM Era
While much of the public conversation has focused on workforce disruption, an equally important question has received less attention:
- Model Context Protocol (MCP): Revolutionizing Software Development with LLMs — A Practical Demo on Search
The integration of Large Language Models (LLMs) with traditional software systems creates both opportunities and challenges. As AI adoption accelerates, developers need…
- Unlocking the Full Potential of RAG Systems: Avoiding Missteps and Embracing Best Practices
Retrieval-Augmented Generation (RAG) and semantic search are revolutionizing how we harness AI to retrieve and generate knowledge-rich responses. As these transformative…
- Practical Guidance for Evaluating Large Language Model (LLM) Products
In today’s data-centric landscape, machine learning (ML) models are crucial for driving decisions, automating processes, and enhancing user experiences across various…
- An LLM finetuning use case comparing Gemma and Llama2
This article guides you through setting up a local GPU machine for fine-tuning large language models (LLMs), Gemma and Llama2. It presents a case study comparing the two open…
- Prototyping and comparing Milvus and Elasticsearch in standalone mode
The tech industry has been exploring vector databases primarily due to the rise and increasing importance of machine learning applications. These technologies rely heavily on…
- Large Language Model Finetuning Practice
This article demonstrates how to finetune two of the most popular Large Language Models(LLM), OpenAI GPT and Meta Llama2. The main purpose is to demonstrate the step-by-step…
- An example of generating Q&A training/evaluation dataset from Large Language Models
The first way is to leverage LangChain’s API. Deeplearning.ai has a great class on LangChain providing a demo generating Q&A on top of OpenAI GPT-3.5. You can get more…