Tag
A formally verified 3D mesh intersection implementation in Lean 4 that requires reviewing only 93 lines of specification, trusting the Lean checker rather than 1000+ lines of AI-generated code. It demonstrates a novel approach to reducing human review effort while ensuring correctness.
The article proposes using stacked branches (small, sequential pull requests) to make reviewing AI-generated code more manageable and effective, addressing the common problem of large, hard-to-review diffs.
JAXBench is a new benchmark suite of 50 JAX workloads for evaluating AI-generated kernel optimization on Google Cloud TPUs, with hand-tuned baselines and an agent evaluation harness. The paper finds that conditioning on curated TPU documentation significantly improves correctness and speedup, with Autocomp beam-search achieving up to 1.6x geomean speedup over XLA on hand-tuned kernels.
This essay examines how AI-generated code has broken the traditional link between the worker's credentials and their output, sparking debates over gatekeeping and quality. It proposes that new verification tools can rebundle contribution with a verifiable receipt, addressing both quality and access concerns.
The author recounts using Fable and Grok 4.5 to identify and remove poor AI-generated code from their codebase.
AI model Claude was used to write a FlashAttention forward kernel using the pyptx DSL, achieving near-parity performance with hand-tuned FlashAttention-4 on NVIDIA B200 hardware.
Slopfix offers a paid service to refactor and reduce AI-generated codebases, charging $10k per week with a fixed price based on reduction targets.
A developer used AI to build a PHP engine in Rust, passing 17% of PHP's official test suite and rendering WordPress, despite having no prior Rust or compiler experience.
The git-annex maintainer details efforts to remove dependencies containing LLM-generated code, expressing concerns about code quality and ethical implications.
Godot Engine announces updated contribution policies to address overwhelming PR volume and AI-generated contributions, including stricter rules on AI use and new contributor guidelines.
AI has made writing code cheaper but shifted the hard work to context-setting, review, and cleanup, requiring more skilled oversight. The article argues that teams often treat AI-generated code as finished work when it is actually a fast first draft.
This article explores how Java development teams are establishing guardrails and best practices to manage the quality, security, and reliability of AI-generated code.
Companies are banning developers from pushing vibe-coded (AI-generated) software to production, highlighting the risks of letting automatically generated code reach users without proper oversight.
A developer expresses concern about the increasing reliance on AI-generated code, fearing that their career will devolve into merely reviewing AI output rather than engaging in creative problem-solving and coding.
An old-school web-based sports sim developer reflects on how the rise of AI-generated 'vibe coded' games threatens his niche, contrasting modern AI-assisted development with the manual efforts of the past.
The article argues that increasing the amount of AI-generated code does not necessarily improve team speed and may even reduce efficiency.
Mutation testing is now generally available in the sydtest Haskell testing framework, enabling developers to automatically verify test suite quality by generating code mutations and checking that tests catch them. The author was motivated by the rise of AI-generated code (via Claude) and the need for an objective, automated measure of test coverage.
Explores the challenges of AI-generated software in organizations, including questions of ownership, understanding black-box code, and maintenance responsibilities.
SQLite has added an AGENTS.md file stating it does not accept agentic code but welcomes agentic bug reports with reproducible test cases, and has split AI-generated bug reports into a separate forum.
A team encounters a production bug caused by AI-generated code, sparking a discussion on who owns the error and how to redefine code ownership in AI-assisted development.