Get the latest tech news

The current state of LLM-driven development


I spent the past ~4 weeks trying out all the new and fancy AI tools for software development.

Jules is a catastrophe and it has not once landed a proper feature in neither Python nor Flutter repos, even with heavy configuration and effort on my side Gemini Code is supposed to be free for 1000 requests, except you hit the limits in 10 minutes. Working with a new stack LLMs are very good at processing large amounts of documentation and implementing features based on it, but also passing that information back to you with recaps in markdown files Crafting small, private, replaceable pieces of software I made a CLI logs viewers and querier for my job, which is very useful but would have taken me a few days to write (~3k LoC)

Get the Android app

Or read this on Hacker News

Read more on:

Photo of LLM

LLM

Photo of current state

current state

Photo of driven development

driven development

Related news:

News photo

Apple taught an LLM to predict tokens up to 5x faster in math and coding tasks

News photo

An LLM does not need to understand MCP

News photo

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality