Graham Dumpleton, creator of mod_wsgi, has launched Muru, an AI-assisted Rust server aimed at providing the best tooling for running ASGI and WSGI Python applications in both development and production. Continuing the approach he used with wrapture, he is publishing an unusually detailed build-in-public account of the project as he develops it.
AI-Assisted Coding
AI 辅助编程
Development paradigm using large models for code generation, completion and refactoring; its real delivery gains are the focus o
Curated Intelligence
A developer poses a counterintuitive question: if AI could truly replace programmers, bug backlogs would be empty and feature requests caught up, yet neither is true. The thread shifts the debate from whether AI can write code to why higher coding throughput has not translated into delivery output.
M-VAVE open-sourced the firmware, dev kit and browser emulator for its ~$70 FM-1 pocket synth under GPLv3 in October, after a community wave of AI-written custom firmware ranging from virtual analog synths to a Doom port. The author argues replaceable, safely reversible firmware will become a key criterion for agent-recommended hardware.
Former Meta engineer Evan X. Merz argues that teams heavily reliant on AI-generated code are shipping brittle codebases riddled with subtle technical debt, with even senior engineers overlooking glaring issues. He contends AI did not deliver order-of-magnitude productivity gains and that high-quality software will still depend on human developers.
Taobao's TEKO introduces a 'Write-Verify-Fix' autonomous loop to tackle the lag in validation capabilities for AI-generated code. The framework addresses two multiplicative bottlenecks causing a mere 6.4% executable rate for AI test cases: 'criteria gap' (lack of structured correctness definitions) and 'signal gap' (inability to perceive real-time runtime feedback). By establishing a four-dimensional feedback mechanism covering environment, data, execution, and automation, TEKO aims to shift engineering practices from 'generation as completion' to full-chain collaborative verification.
Tim Cook, Apple's Executive Chairman, highlighted AI's dual impact on software development and hardware manufacturing during his first China visit post-CEO role. He noted that AI-assisted coding lowers barriers for non-programmers, while machine learning has transformed supply chain quality control, enabling highly automated and precise color consistency checks in iPhone mass production.
Former Databricks engineer Yuchen Jin highlights that new hires leveraging internal AI dev tools can land impactful pull requests by day two. This contrasts sharply with the traditional onboarding process, which typically required two weeks of reading design docs and codebases before submitting even a minor patch in week three.
Linus Torvalds has submitted his first series of kernel patches developed with the assistance of an LLM (specifically Claude Code), targeting scalability bottlenecks in page reference counting. The experimental branch boosts concurrent file read throughput by approximately 3x, rising from ~10M to over 30M reads per second, while drastically reducing CPU cycles per operation. Additionally, Torvalds leveraged AI tools to identify and fix three subtle bugs that had persisted in the Linux codebase for 13 years, signaling a significant shift toward deep human-AI collaboration in core infrastructure development.
Anthropic's team built the internal source control system Scion and code review tool Antique from scratch using Claude in just five months, successfully migrating their monorepo away from Git. This demonstrates AI-assisted programming's potential to refactor complex infrastructure, signaling a shift toward AI-native developer toolchains.
Anthropic's internal data reveals a 12x surge in daily code merges from 2,500 in May to over 30,000 in September. This extreme throughput increase is attributed to deep integration of AI coding tools but raises concerns about GitHub infrastructure stability.