惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
罗磊的独立博客
雷峰网
雷峰网
量子位
V
Visual Studio Blog
Vercel News
Vercel News
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
The Cloudflare Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
宝玉的分享
宝玉的分享
月光博客
月光博客
Martin Fowler
Martin Fowler
aimingoo的专栏
aimingoo的专栏
H
Hackread – Cybersecurity News, Data Breaches, AI and More
Microsoft Security Blog
Microsoft Security Blog
博客园 - 叶小钗
腾讯CDC
Engineering at Meta
Engineering at Meta
博客园 - Franky
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Y
Y Combinator Blog
Recent Announcements
Recent Announcements
Jina AI
Jina AI
A
About on SuperTechFans

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
MAREF — GitHub Is Becoming a Giant AI Code Dump
MAREF Engineering · 2026-06-24 · via Hacker News - Newest: "AI"

GitHub has 630 million repos. Nearly half of all new code is written by AI.

Sounds like a productivity explosion, right?

But there's another number in that same report: developer trust in AI code dropped from 77% to 60%. More people are using it. Fewer people believe in it.

And it gets worse.

Half the projects you see are fake

Carnegie Mellon University found 6 million fake stars on GitHub. Security firm Socket uncovered 370,000 "fix stars" directly tied to scams. That vibe coding project with thousands of stars? Probably half are fabricated.

That's not even the scary part.

CodeRabbit scanned 470 PRs. AI-written code had 1.7x more critical issues than human-written code. 45% of AI code ships with OWASP Top 10 vulnerabilities. 63% of developers say fixing AI code takes longer than writing it from scratch.

But here's the most mind-bending finding.

You're getting slower, but you have no idea

Meter ran a randomized controlled trial. The result: AI users were actually 19% slower, but they thought they were 20% faster.

When they showed the participants the data, they still insisted they were faster.

That's the real horror. You're getting slower, and you have no idea.

"This is metacognitive dissonance — the tool gives you the illusion of speed, but reality gives you a slap."

Open source is drowning in garbage PRs

Look at what major projects are doing:

  • curl — shut down its 7-year bug bounty program. AI-generated bug reports: only 5% were real. The rest was noise.
  • Ghost — outright banned AI-submitted code.
  • Tailscale — went further: closed all external PRs. AI or not, doesn't matter anymore.
  • GitHub itself — building a "PR kill switch" so maintainers can one-click disable external submissions.

Open source isn't being destroyed by hackers. It's being drowned to death by garbage AI-generated PRs from vibe coders.

Garbage in, garbage out

Here's the most ironic part of the whole chain:

You ask AI to write code → AI searches GitHub for references → It finds code written by other AIs → Garbage in, garbage out. What do you think comes out?

GitHub used to be a code repository. It's turning into a giant AI code dump.

The problem isn't AI. It's governance.

AI writing code isn't the problem. The problem is: nobody is auditing what AI does.

The core premise of vibe coding is "accept everything AI generates — don't review, don't modify." That works for a TODO app. For production code, it's planting time bombs.

But asking humans to line-by-line review AI code isn't realistic either. If 63% of devs say fixing is slower than writing, reviewing is even slower.

The real solution isn't "stop using AI" or "review everything manually." It's automated governance between AI code and production:

  • Every tool call is audited — every file change, every API call, cryptographically signed and recorded.
  • Dangerous actions are blocked — delete database? modify production? First pass through a 4-level decision tree. 97% auto-resolved, 3% escalated to human review.
  • Security policies evolve — every false positive, every bypass attempt feeds back into the governance engine.
  • Formal verification — not "we hope this is safe." Mathematically provable convergence toward safety.

That's MAREF. The open-source agent governance operating system. Brakes, seatbelts, dashcams, and a tireless security reviewer — for your AI agents.

Governance isn't a speed bump. It's what lets you go fast safely.

Un-governed AI code doesn't go faster. It just crashes faster.

The projects banning AI submissions — curl, Ghost, Tailscale — aren't anti-AI. They're saying the same thing: "We welcome AI assistance. We won't accept garbage without quality control."

MAREF is that quality control layer. Not between you and AI. Between garbage code and your production environment.

You can still vibe code. You can still have AI write your code. But before it deletes your database, someone will stop it.

You're getting slower? No. You're finally seeing what you're actually shipping.


📊 Sources: GitHub Octoverse Report, Carnegie Mellon fake star study, Socket security report, CodeRabbit 470 PR analysis, Meter RCT trial, curl/BT/Ghost/Tailscale official announcements. MAREF is an open-source agent governance operating system. Get started in 5 minutes.