Hey HN! When I started looking into LLMs and agents for software development and introducing them at work, I quickly realised that a person new to the topic faces a real barrage: - all the hype (AGI, engineers getting replaced by AI etc.) - conflicting opinions in virtually every discussion—for every person saying they’ve 10x-ed their productivity, there is a comment decrying LLMs as an utter failure - a lot of jargon (MoE, MCP, RAG, distillation, quantisation etc. etc.) - a profusion of models, IDEs/IDE extensions, CLI agents, other tools etc.

Sorting through all of this can be quite tricky for somebody coming in fresh! HN users have been discussing LLM tools for a long time of course, but many programmers I know still haven’t _really_ tried anything other than an LLM web UI and haven't been following the progress of tools much.

So my goal for this project was to provide a balanced overview of the topic, point people to substantive resources on eg. context management and productivity effects, and cover the concerns and risks as well (from prompt injection to shady training data sourcing). I hope it's useful!