All articles
36774 articles
Are 90% of AI-Written Tests Pseudo-Logic? Three Blind Spots and Probe Tactics
Why are 90% of AI-generated unit tests pseudo-logic? This article breaks down three blind spots—fake coverage, hollow logic, and missing side effects—with hands-on mutation testing, boundary probes, and side-effect hooks to turn AI tests from decoration into a real quality line.
AI Writing Too Much Code? Ponytail Plugin Cuts Output by 54%
Ponytail is an open-source plugin that makes AI coding assistants write less code, using a six-tier code ladder to prioritize reusing existing solutions. In tests across twelve tasks, lines of code dropped by an average of 54%, with the date picker task down 94%. Works best with Claude Code.
Ponytail: 176K Stars — Teaching AI Coding Assistants to Be Lazy and Halve the Code
Ponytail is an open-source project that teaches AI coding assistants to be lazy with a seven-rung ladder strategy: find existing solutions first, then write minimal code. Official tests show 53% less code, 41% less time, and high-risk logic test coverage rising from 68% to 98%. It has 176K stars on GitHub.
Can AI-Written Specs Generate Code Directly? A Multi-Agent Review Workflow Explained
Can AI-written Specs generate code directly? Bilibili creator Lao Dao Qian Li shares the SDFlow multi-agent review workflow: fresh contexts, parallel reviews from multiple angles, citation verification, then item-by-item adjudication. 50 issues trimmed to 18 adopted; humans only make key decisions. 17 minutes and 15 million tokens buy reduced uncertainty and early error correction.
Codex Freelancing in Practice: From Zero to Your First Earnings with AI Programming
A Bilibili creator shares how they monetized AI programming gigs with Codex from zero, covering tiered freelance platforms, collaboration models, and the opportunities and risks of AI coding freelancing.
Why Is Writing AGENTS.md First the Wrong Move When Handing Legacy Code to AI?
Handing legacy code to AI for refactoring? Writing AGENTS.md first is wrong. The real prerequisite is runnable tests as an exit condition—let AI draw the map itself.
AI API Keys Stolen: Single Company Loses Nearly $1 Million
A Hacker News discussion reveals a company incurred nearly $1M in charges after its AI API keys were stolen. This article analyzes common leak routes, the risk amplification of usage-based AI API billing, and protections such as key rotation and usage alerts.
OpenAI Appoints Dan Driscoll as Senior…
OpenAI Appoints Dan Driscoll as Senior Counselor for National Security Modernization
OpenAI appoints Dan Driscoll as Senior Counselor for National Security Modernization, advising on responsible AI in national security, cyber resilience, and allied collaboration. This article analyzes the role's scope, the trends behind it, and the controversies.
AI Keeps Reworking Your Code? Plan First, Code Later to Slash Rework
AI code always needs rework? Learn plan mode in AI coding: for big changes, have AI propose a plan before coding. Focus on three things—which files to change, how to verify, where to confirm—plus two copy-paste prompts to cut rework.
Mathematician Explains Optimal Stopping Strategy: The Wisdom of Dating and Algorithmic Decision-Making
A mathematician on Lex Fridman's podcast explains optimal stopping theory and the secretary problem, exploring online algorithms, the limits of worst-case analysis, and how to make decisions under uncertainty, including mathematical insights for dating strategies.
LLMs Are Not Superhuman: The Ability to Read and Write Neural Network Weights Is Overestimated
Hacker News discussion shows LLMs lack superhuman ability to read and write neural network weights. This article analyzes the reasons, experimental findings, and implications for AI research.
Vercel Project Configuration Adds vercel.toml Support
Vercel adds vercel.toml support for project configuration, with properties matching vercel.json, plus comments and a flatter structure. Learn the one-of-three constraint with vercel.json and vercel.ts, plus Services configuration examples and migration tips.
vLLM Releases v0.40.14-rc0: Fixes Cloud Web Search Response Decoding
vLLM releases v0.40.14-rc0 pre-release; core commit decodes cloud web search responses (#19476), fixing response parsing in the search tool calling pipeline. Version info and upgrade advice inside.
LLM 0.37 Released: New Models and Lessons from AI-Assisted PRs
LLM 0.37 adds gpt-6.1-sol and gpt-6.1-terra models, the llm openai endpoint add command, a cheaper default model gpt-6-luna, and multimodal embeddings.
YC Head Garry Tan Spends $200K a Year on AI Tokens, Calls Himself 'Cyber Psychotic'
Y Combinator CEO Garry Tan revealed on a podcast that he spends up to $200,000 a year on AI tokens and has joked about having 'cyber psychosis.' Here's what his heavy AI agent usage reveals.
Letterpress-1 Released: A Coding Agent Built for Growth Engineering
The Letter Company releases Letterpress-1 beta, a coding agent for growth engineering that reads git sites and market info to auto-open PRs for SEO, GEO, and conversion.
Hinge Tests Lower-Priced Subscription: A Breakout Attempt Amid Dating App Payment Fatigue
Hinge is testing a cheaper subscription plan in response to growing payment fatigue among dating app users. This article analyzes the motivations behind the low-price strategy, its target users, and the direction of adjustments to the dating app business model.
SparsePR: Training-Free Sparse Attention to Accelerate Video Generation
SparsePR is a training-free sparse attention method for video generation and world models that aims to cut attention computation without retraining. This article reviews its approach, use cases, and trade-offs to consider.
Vercel AI SDK's Z.ai Adapter Releases Patch Update
AI SDK's Z.ai adapter releases patch 3.0.46, only syncing provider-utils and openai-compatible dependencies, no API changes, low upgrade risk. This article analyzes the changes and upgrade recommendations.
@ai-sdk/xai 5.0.38 Patch Release: The Ecosystem Signals Behind a Dependency Update
@ai-sdk/xai releases patch 5.0.38, upgrading @ai-sdk/provider-utils to 5.0.74. We analyze what this dependency update means for the AI SDK ecosystem, multi-provider consistency, and developer maintenance.