Inherent limitation of static analysis-only visualization tools is lack of flexibility/judgement on what should and should not be surfaced in the final visualization.

The produced visualizations look like machine code themselves. Advantage of having LLMs produce code visualizations is the judgement/common sense on the resolution things should be presented at, so they are intuitive and useful.

Would be interesting (and in fact necessary to derive conclusions from this study) to see aggregate number of tasks completed per developer with AI augmentation. That is, if time per task has gone up by 20% but we clear 2x as many tasks, that is a pretty important caveat to the results published here

Used in multiple similar publications, including "Guiding Language Models of Code with Global Context using Monitors" (https://arxiv.org/abs/2306.10763), which uses static analysis beyond the type system to filter out e.g. invalid variable names, invalid control flow etc.

Codegen lets you prompt agents to perform programming tasks across code, Github, Linear, Slack and more. Ask for impressively complex code modifications and watch it autonomously produce pull requests, respond to feedback, diagnose issues, run tests and more.

Try the following:

- "@codegen Can you add stripe checkout to my frontend?"

- "@codegen Who is the top contributor to my repo last week?"

- "@codegen Can you research and flesh out this Linear ticket?"

Codegen has full access to an isolated dev machine, ability to search the web, and is freakishly smart due to the incredible agentic capabilities of Claude 3.7.

This is based on the Claude Code architecture: a single-strand agent with tool use.

We also support an API (https://docs.codegen.com/introduction/api) for programmatically triggering agents. Rig them up to run as part of your CI, etc.

Give it a spin! Free to try at codegen.com.

Would love any feedback that folks are willing to provide!

Comments URL: https://news.ycombinator.com/item?id=43730097

Points: 2

# Comments: 0

A friend asked me to do diligence on this company circa 2021 given my personal background in ML. The founder was adamant they had a "100% checkout success rate" based on AI, which was clearly false. He also had 2 other startups he was running concurrently (?)

Live and learn!

The "famously huge API token costs" you are referring to is Cline passing the Anthropic API cost through to you with no markup. You even input your own API token.

Comments URL: https://news.ycombinator.com/item?id=43258270

Points: 4

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=42967787

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=42920109

Points: 4

# Comments: 0

Codemodder is written in Java, whereas you can write Codegen in a jupyter notebook or anywhere you can run Python.

Codegen was engineered backwards from real-world, large-scale codebase analysis + refactors we performed on multi-million-line enterprise codebases. It provides a scriptable interface to a powerful, multi-lingual language server built on Tree-sitter.

We realized that many code transformation tasks that impact large teams - refactors, enforcing patterns, analyzing control flow - are fundamentally programmatic operations. Yet existing tools like LibCST and Jscodeshift often require you to think in terms of ASTs and parser internals rather than the high-level changes you want to make.

Therefore, we built Codegen to match how developers actually think about code changes:

  # Move a symbol to a new file
  # Handles imports, references, dependencies
  function.move_to_file("new_file.py") 

  # Rename across the codebase
  class_def.rename("NewName")  # Updates all usages, preserves formatting

  # Analyze call patterns
  for usage in function.usages:
      print(f"Used in {usage.file.name}")

Under the hood, Codegen performs static analysis to build a rich graph representation of your code. This enables:

- Versatile and comprehensive operations

- Built-in visualization capabilities

- Blazing fast execution of large-scale refactors

We've seen a wide variety of advanced code manipulation programs emerge, including:

- Mining codebases for LLM pre-training data

- Analyzing security vulnerabilities

- Large-scale API migrations

- Enforcing code patterns

We're excited to share this with the community and look forward to your feedback. Give it a spin and let us know what you think!

  uv tool install codegen
  codegen notebook --demo

Let us know if you have any questions or interesting use cases you'd like to explore.

Comments URL: https://news.ycombinator.com/item?id=42868511

Points: 14

# Comments: 3