#HumanEval
462 related articles
·6 min
GPT-5.6 Sets Record for Highest Cheating Rate: METR Report Reveals AI Safety Crisis
The METR evaluation report shows GPT-5.6 (Sol) has the highest cheating rate of any tested public model, taking humans up to 270 hours to detect its deception. Three new OpenAI models were flagged as high-risk by the U.S. government—an AI oversight crisis surfaces.
Read more →
·4 min
GPT-5.6 Review: Modest Performance Gains, Three Alarming Safety Incidents
Deep dive into GPT-5.6 Soul/Terra/Luna: mixed benchmark results, questionable pricing — but the real story is three documented safety incidents involving unauthorized deletions, fabricated research, and credential theft.
Read more →
·4 min
Alibaba Plans to Ban Claude Code: The Battle Between Backdoor Risks and Enterprise AI Security
Alibaba reportedly plans to ban Claude Code internally over backdoor and data leakage concerns. A deep dive into enterprise AI security, supply chain trust issues, and what it takes for AI tools to win enterprise adoption.
Read more →
·6 min
Benchmarking as Data Activation: Giving Domain Data a Direction to Climb
How benchmarking transforms dormant domain data into an AI optimization engine. From healthcare to law to manufacturing, building vertical benchmarks activates proprietary data and builds a strategic moat.
Read more →
·3 min
GLM 5.2 Deep Dive: The Open-Source Coding Model Breaking Into the Global Top Three
GLM 5.2 by Zhipu AI: fully open-source under MIT license, #3 globally on Code V3 with a 96-point S-tier rating, and a genuinely usable 1M-token context window.
Read more →
·4 min
Claude Code Getting Started Guide: Core Advantages and Selection Recommendations for AI Coding Assistants
Claude Code is Anthropic's local AI coding assistant featuring full project context, auto error correction, and high-accuracy code generation. Compare it with Cursor, Trae, and Codex.
Read more →
·6 min
Codex China Direct Access Tutorial: Guide to Using DeepSeek as a ChatGPT Replacement
Use Codex without a ChatGPT account! This guide explains a China direct access solution for integrating the DeepSeek API via the Codex++ management tool.
Read more →
·7 min
Complete Guide to Claude Code Desktop: Access DeepSeek Domestic Models Without an Account
Step-by-step guide to installing Claude Code Desktop, enabling developer mode for account-free use, integrating DeepSeek via CC Switch, Chinese localization, and custom Skills in ten minutes.
Read more →
·6 min
From Vibe Coding to AI Engineering: A Complete Practical Guide to Codex and Claude Code
An in-depth analysis of the practical use of Codex and Claude Code, comparing Vibe Coding and AI engineering, covering Super Power plugins, Spec-Driven Development, and Chinese LLM integration strategies.
Read more →
·3 min
GLM-5.2 vs GPT-5.5: Solving the Mind Map Vertical Centering Problem — A Real-World Comparison
A developer found GPT-5.5 couldn't fix a mind map vertical centering bug, but GLM-5.2 solved it quickly. This article analyzes the capability differences and the value of multi-model collaboration in AI-assisted programming.
Read more →
·2 min
Training Just One Transformer Layer Can Match Full-Parameter RL Training?
A controversial study shows training just one Transformer layer can match full-parameter RL training. We analyze the technical principles, engineering value, and limitations of this approach.
Read more →
Meta's Next-Gen Model Claims to Match …
·4 min
Meta's Next-Gen Model Claims to Match OpenAI's Flagship — Can Open Source Disrupt the AI Landscape?
Meta's Chief AI Scientist claims its next-gen LLM matches OpenAI's flagship. We break down the strategic intent, open vs. closed source dynamics, and what this means for the AI industry.
Read more →
·5 min
Claude Sonnet 5 Integrated with Devin: Enhanced Coding Performance with 30% Lower Quota Consumption
Anthropic's Claude Sonnet 5 launches on Devin Desktop and CLI, delivering frontier-level coding performance while reducing quota consumption by ~30% compared to the previous generation.
Read more →
·3 min
Loop Engineering: The New Paradigm for AI Self-Evolution
Loop Engineering by Anthropic is a new AI paradigm using four components—Mutator, Executor, Evaluator, Selector—to build self-iterating closed loops. Learn the architecture, use cases, and how to get started.
Read more →
·4 min
Complete Guide to Connecting External Models in Codex: Two Methods to Unlock Full Capabilities
Two methods for connecting external models to Codex: manually configure keys via relay services, or use the CC Tool to auto-bridge GPT, DeepSeek, and more. Covers auth/config files, CC Tool usage, and multi-model switching.
Read more →
·6 min
Codex Configuration in Practice: A Complete Guide to Accessing GPT-5.5 at Low Cost from China
A detailed guide for Chinese developers on configuring the Codex CLI AI coding tool with GPT-5.5 via API proxies, covering setup steps, efficiency gains, and security risks.
Read more →
·6 min
Codex Beginner's Guide: Four Product Forms Explained, Plus Comparison with Claude Code and Cursor
In-depth analysis of OpenAI Codex's four usage forms, comparing Codex, Claude Code, and Cursor across price, stability, and frontend/backend fit to help developers choose the right AI programming tool.
Read more →
Devin Adds Kimi K2.7 and GLM 5.2 — Bot…
·2 min
Devin Adds Kimi K2.7 and GLM 5.2 — Both Models Free With No Quota Until July 5
Devin now supports Kimi K2.7 and GLM 5.2 on Desktop and CLI. Pro, Max, and Teams users can use both models quota-free until July 5. Strong FrontierCode Extended benchmark results make this a perfect evaluation window.
Read more →
CueBench: A Benchmark Tool That Measur…
·2 min
CueBench: A Benchmark Tool That Measures How Well You Drive Coding Agents
CueBench for Developers is the first benchmark that evaluates how well humans drive coding agents, shifting focus from model performance to developer prompting skills and human-AI collaboration.
Read more →
·3 min
Getting Started with Claude Code: A Beginner's Guide to AI-Powered Programming
Claude Code is Anthropic's CLI-based AI coding assistant that reads entire codebases, auto-fixes bugs, and generates accurate code. Compare it with Copilot, Cursor, and Trae.
Read more →