OSS Is Going Just Great
…and Generative AI is definitely not burying maintainers under mass-generated slop.
2024
January
- Daniel Stenberg (curl) publishes “The I in LLM stands for intelligence”
March
April
May
June
July
- GPT-4o mini released
- “We Have a Package for You!” posted to arXiv (later published at USENIX Security 2025)
September
October
- Claude 3.5 Haiku released
- Sonatype 2024 report: 704,102 malicious packages logged (156% YoY increase)
December
- o1 released
- GitHub Copilot free tier
- Seth Larson (Python PSF) publishes “New era of slop security reports”
2025
January
February
- Andrej Karpathy coins “vibe coding”
- Claude 3.7 Sonnet released
- Claude Code preview
April
- o3 and o4-mini released
- GPT-4.1 series released
- Slopsquatting term coined (by Seth Larson, amplified by Andrew Nesbitt)
- Sonatype Q1 2025 report: 17,954 malicious OSS packages identified; 56% related to data exfiltration (up from 26% in Q4 2024)
May
- Claude Sonnet 4 and Opus 4 released
- Claude Code GA
- OpenAI Codex released
- Stenberg posts “We are effectively being DDoSed”; introduces instant-ban policy for AI slop on curl’s HackerOne
- Open Collective reports being “flooded with AI garbage” bug reports
June
- Trend Micro publishes slopsquatting research
- Malicious PyPI packages targeting Solana developers discovered
July
- curl bug report volume spikes to 8x normal rate in one week
- Stenberg publicly mulls killing the bug bounty program
- HackerOne launches Hai Triage
- PyPI publishes incident report on phishing attack
- IndonesianFoods campaign reaches 150,000+ npm packages
- METR study: experienced open-source developers are 19% slower with AI tools — despite believing they were 20% faster
August
- GPT-5 released
- Claude Opus 4.1 released
- Ghostty (Mitchell Hashimoto) introduces mandatory AI disclosure policy for PRs
September
- Claude Sonnet 4.5 released
- Shai-Hulud self-replicating npm malware discovered
- PyPI publishes incident report on token exfiltration via GitHub Actions
- Cybersecurity Dive: “Tech giants pledged millions to secure open-source code. Then AI came along.”
October
- Claude Haiku 4.5 released
- Stenberg: “A new breed of analyzers” — praises AI-assisted bug finding by Joshua Rogers
- PyPI adds typosquatting detection for project creation
November
- Clawdbot launched
- Claude Opus 4.5 released
- Nolan Lawson: “The Fate of Small Open Source”
- Linux kernel AI contribution guidelines proposed (Dave Hansen/Intel, v3)
- OCaml maintainers reject 13,000-line AI-generated pull request
- PyPI publishes incident report on Shai-Hulud implications
December
- GPT-5.2 released
- Claude Code background agents released
- PyPI year in review: hardened ZIP upload pipeline, typosquatting detection, domain resurrection prevention, spam prevention
- Excalidraw receives 2x PRs in Q4 vs Q3
- Rob Pike, Guido van Rossum, and Anders Hejlsberg spammed by autonomous AI agents performing “random acts of kindness” on Christmas Day
2026
January
- Clawdbot renamed to Moltbot after Anthropic trademark complaint; then renamed to OpenClaw three days later. The abandoned “Moltbot” name was immediately impersonated by attackers who grabbed old handles and pushed 400+ malicious agent skills through ClawHub
- Sonatype 2026 State of the Software Supply Chain report published
- Cloudflare publishes “Building a serverless, post-quantum Matrix homeserver”; Matrix Foundation responds
- Armin Ronacher (Flask creator) publishes “Agent Psychosis”
- Linus Torvalds tells kernel devs to stop making an issue of AI slop in docs
- curl shuts down bug bounty program (effective Jan 31)
- LLVM institutes “human-in-the-loop” policy for AI-generated contributions
- Ghostty (Hashimoto) escalates to zero-tolerance AI policy
- tldraw (Steve Ruiz) auto-closes ALL external pull requests — “Stay away from my trash”
- 128 phantom npm packages (slopsquatting) racked up 121,539 downloads Jul 2025–Jan 2026
- “How AI Impacts Skill Formation” (Shen & Tamkin) — developers using AI learned less and debugged worse, without significant efficiency gains
February
- Daniel Stenberg delivers FOSDEM 2026 closing keynote: “Open Source Security in spite of AI”
- International AI Safety Report 2026 published (Yoshua Bengio et al.) — documents AI agents completing multi-hour software engineering tasks autonomously
- Claude Opus 4.6 released — Anthropic reports it found 500+ high-severity zero-day vulnerabilities in OSS libraries (Ghostscript, OpenSC, CGIF) during internal red-team testing
- GitHub weighs PR “kill switch”
- RedMonk publishes “AI Slopageddon and the OSS Maintainers”
- InfoWorld: “Is AI killing open source?”
- continue.dev: “We’re Losing Open Contribution”
- Mitchell Hashimoto releases Vouch, a community trust management system based on explicit vouches to participate
- 1Password launches SCAM benchmark — tests whether AI agents fall for phishing and social engineering during realistic workflows; every model tested committed critical failures
- An AI Agent Published a Hit Piece on Me
Resources
- CHAOSS AI Alignment WG — project AI policies, spam examples, blog posts
- Sonatype SSC Attack Timeline — curated timeline of supply chain attacks
- daniel.haxx.se AI tag — Stenberg’s full archive of AI-related curl posts
- Pivot to AI — David Gerard’s AI hype and harm tracker
- Simon Willison’s blog — prolific coverage of AI tools and their real-world impact
- Seth Larson’s blog — Python PSF security developer-in-residence
Missing something? Contributions welcome.
Hat tip to Web3 Is Going Just Great for the format inspiration.