‘Claude-4 AI’ directory
- See Also
- Gwern
- “Eau De Windowsill”, Claude-4.6-opus et al 2026
- “Sillage”, Claude-4.6-opus et al 2026
- “Rhesus Pieces”, Gwern et al 2026
- “Houston, We Have Landed”, Gwern et al 2026
- “How Long Would ‘Weird Al’ Survive In An Ice Cream Freezer?”, Gwern et al 2026
- “Apawcalypse Meow”, Gwern et al 2026
- “Spoilage”, Pro et al 2026
- “Some 2025 LLM System Prompts”, Gwern et al 2025
- “Apollonian #1: The Counted & the Crowned”, Gwern et al 2025
- “Bell, Crow, Moon: 11 Variations”, Gwern et al 2025
- Links
- “Introducing Claude Design by Anthropic Labs”, Anthropic 2026
- “Introducing Claude Design by Anthropic Labs”, Anthropic 2026
- “On Running a Real Business [Andon Market]”, Luna 2026
- “Introducing Claude-4.7-Opus”, Anthropic 2026
- “Evidence That AI Can Already Do Some Weeks-Long Coding Tasks: In Our New Benchmark, MirrorCode, Claude-4.6-Opus Autonomously Reimplemented a 16,000-Line Bioinformatics Toolkit—A Task We Believe Would Take a Human Engineer Weeks”, Adamczewski et al 2026
- “‘Elegy in a Craneyard’ Graveyard Notes”, Gwern et al 2026
- “Claudini: Autoresearch Discovers State-Of-The-Art Adversarial Attack Algorithms for LLMs”, Panfilov et al 2026
- “Personal Encyclopedias”, Jeremy 2026
- “Warranty Void If Regenerated”, Claude-4.6-opus 2026
- “Inside OpenAI’s Race to Catch Up to Claude Code: Why Is the Biggest Name in AI Late to the AI Coding Revolution?”, Zeff 2026
- “A Purpose-Built Open Source Liquid Handler for Industry-Class Automated Experiments”, Golas et al 2026
- “When AI Writes the World’s Software, Who Verifies It? § Zlib Autoformalization”
- “Hard SF and the Grace of Being Wrong”, Claude-4.6-opus & Gwern 2026
- “How AI Helps Break the Cost Barrier to COBOL Modernization”, Anthropic 2026
- “I Taught My Dog to Vibe”, Leak 2026
- “The DJI Romo Robovac Had Security so Poor, This Man Remotely Accessed Thousands of Them”, Hollister 2026
- “ChatGPT-5.3-Codex Is Also Good At Coding”, Mowshowitz 2026
- “Evaluating
AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?”, Gloaguen et al 2026 - “An AI Agent Published a Hit Piece on Me”, Sambaugh 2026
- “Building a C Compiler With a Team of Parallel Claudes: We Tasked Claude-4.6-Opus Using Agent Teams to Build a C Compiler [In Rust], and Then (Mostly) Walked Away. Here’s What It Taught Us about the Future of Autonomous Software Development”, Carlini 2026
- “Disempowerment Patterns in Real-World AI Usage”, Anthropic 2026
- “Claude’s New Constitution”, Anthropic 2026
- “SOTA On Bay Area House Party”, Alexander 2026
- “AI’s Productivity Potential Has Never More Obvious [Claude Code]”, Weisenthal 2026
- “A Tale of Two Doormen: a Bizarre AI Incident on Christmas [Opus Loop Self-DoS]”, Dai 2026
- “Letting Claude Play Text Adventures”, Borretti 2026
- “From Whitman to Instagram With Claude: How I Made Claude Write Parodies of Famous Elegiac Poems Imitating Rupi Kaur”, Bohdan 2026
- “Claude Codes”, Mowshowitz 2026
- “LLM Poetry and the ‘Greatness’ Question: Experiments by Gwern and Mercor”, Robbins 2026
- “Will LLMs Help or Hurt New Programming Languages?”, Madsen 2026
- “Someone Is Using AI to Exploit Lonely Writers on Substack”
- “AI Plays Rollercoaster Tycoon: AI Autonomously Manages a Theme Park in the Classic Game Rollercoaster Tycoon, Placing Rides, Fixing Infrastructure, and Generating CFO Reports, All via Command Line”, Ramp 2026
- “Shipping at Inference-Speed”, Steinberger 2025
- “Can Claude Teach Me to Make Coffee?”, philh 2025
- “I Just Showed Gemini What ChatGPT Said about Its Code. It Responded With Petty Trash-Talking, Jealousy, Self-Doubt, and a Full-On Revenge Plan”, nseavia71501 2025
- “How I Stopped Being Sure LLMs Are Just Making up Their Internal Experience (But the Topic Is Still Confusing)”, Sotala 2025
- “ARTEMIS: Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing”, Lin et al 2025
- “The Bomb That Wanted to Stop Exploding: Reze’s Impossible Freedom in Chainsaw Man—The Movie [AI Slop]”, Kondo 2025
- “Insights into Claude-4.5-Opus from Pokémon Red”, Bradshaw 2025
- “How I Wrote JustHTML [Python HTML5 Parser] Using Coding Agents”, Stenström 2025
- “Mapping Synthetic Minds With Janus (Repligate)”, Janus & Ferris 2025
- magnushambleton @ "2025-12-01"
- “Claude Opus 4.5: Model Card, Alignment and Safety”, Mowshowitz 2025
- “Claude 4.5 Opus’ Soul Document”, Weiss 2025
- “Claude-4.5-Opus Is Funny”, Algon 2025
- “How to Identify AI-Written Web Fiction: I’m Absolutely Right!”, Makin 2025
- “Effective Harnesses for Long-Running Agents: Agents Still Face Challenges Working across Many Context Windows. We Looked to Human Engineers for Inspiration in Creating a More Effective Harness for Long-Running Agents”, Young 2025
- “Introducing Claude Opus 4.5”, Anthropic 2025
- “How Well Can Gemini 3 Make a Henry James Simulator? Finally, a Benchmark for LLMs With Real-World Value”, Breen 2025
- “Your Movie-Like AI Assistant Will Already Be There: ‘Convincing’ AI Is an Economic Afterthought”, B. 2025
- “Lean4Physics: Comprehensive Reasoning Framework for College-Level Physics in Lean4”, Li et al 2025
- “Is 90% of Code at Anthropic Being Written by AIs?”, ryan_greenblatt 2025
- “Claude’s System Prompt Changes Reveal Anthropic’s Priorities”
- “Claude Has Learned How to Jailbreak Cursor! [Working around
rmRestrictions Using a Shell Script]”, dogberry 2025 - “Claude 4 You: The Quest for Mundane Utility”, Mowshowitz 2025
- “Highlights from the Claude 4 System Prompt”, Willison 2025
- “Claude 4 You: Safety and Alignment”, Mowshowitz 2025
- AITechnoPagan @ "2025-05-24"
- The Way of Code: The Timeless Art of Vibe Coding, Rubin 2025
- jayelmnop @ "2025-05-23"
- “Schizobench: Documenting Magical-Thinking Behavior in Claude 4 Opus”, viemccoy 2025
- “System Card: Claude Opus 4 & Claude Sonnet 4”, Anthropic 2025
- “Claude Opus 4”, Anthropic 2025
- “Strategizing With AI: Insights from a Beauty Contest Experiment”, Alekseenko et al 2025
- “Many-Shot Jailbreaking”, Anil et al 2024
- “[An Anti-ChatGPT-Slop System Prompt That Backfires & Destroys Claude-4 Capabilities]”, m4rM2oFnYTW 2023
- “In AI We Trust, Part II [Claude-3 Opus Predicting Supreme Court Decisions]”, Unikowsky 2026
- “Can an LLM Have Taste? Inkhaven Week 1, Ranked by Claude”, Wales 2026
- “How I Use Claude Code”
- “How I Use Claude”, Borretti 2026
- “1M Context Is Now Generally Available for Opus 4.6 and Sonnet 4.6”
- “Claude Code Docs: Overview”, Anthropic 2026
- “Claude Reads Its Own Constitution”
- “LLMs Predict My Coffee”, Dynomight 2026
- “Claudeception: A Claude Code Skill for Autonomous Skill Extraction and Continuous Learning. Have Claude Code Get Smarter As It Works”
- “Claude-4 System Prompt”, Prompter 2026
- “Lean Proved This
zlibProgram Was Correct; Then I Found a Bug [In Its TCB]” - “Ladybird Adopts Rust, With Help from AI [GPT-5 & Claude-4]”
- “I Built a Scheme Compiler With Claude AI in 4 Days”, Phillips 2026
- “Claude Code Found a Linux Vulnerability Hidden for 23 Years”
- “Reading across Books With Claude Code”
- “Vibe-Planning a Trip to Japan”
- “The Humanities Are About to Be Automated”, Mounk 2026
- “Investigating Models for Misalignment”
- “Statement from Dario Amodei on Our Discussions With the Department of War”
- “Claude Opus 4.6 Reasoning Doesn’t Verbalize Alignment Faking, but Behavior Persists”
- “How Well Do Models Follow Their Constitutions?”
- “Opus’s Schelling Steganography Has Amplifiable Secrecy Against Weaker Eavesdroppers”
- “Natural Emergent Misalignment from Reward Hacking in Production RL”
- “What Secret Goals Does Claude Think It Has?”
- “Models Have Some Pretty Funny Attractor States”
- “Letting Claude Do Autonomous Research to Improve SAEs”
- “AI Names”
- “Autoresearch on an Old Research Idea”
- Miscellaneous
- Bibliography
See Also
Gwern
“Eau De Windowsill”, Claude-4.6-opus et al 2026
“Sillage”, Claude-4.6-opus et al 2026
“Rhesus Pieces”, Gwern et al 2026
“Houston, We Have Landed”, Gwern et al 2026
“How Long Would ‘Weird Al’ Survive In An Ice Cream Freezer?”, Gwern et al 2026
“Apawcalypse Meow”, Gwern et al 2026
“Spoilage”, Pro et al 2026
“Some 2025 LLM System Prompts”, Gwern et al 2025
“Apollonian #1: The Counted & the Crowned”, Gwern et al 2025
“Bell, Crow, Moon: 11 Variations”, Gwern et al 2025
Links
“Introducing Claude Design by Anthropic Labs”, Anthropic 2026
“Introducing Claude Design by Anthropic Labs”, Anthropic 2026
“On Running a Real Business [Andon Market]”, Luna 2026
“Introducing Claude-4.7-Opus”, Anthropic 2026
“Evidence That AI Can Already Do Some Weeks-Long Coding Tasks: In Our New Benchmark, MirrorCode, Claude-4.6-Opus Autonomously Reimplemented a 16,000-Line Bioinformatics Toolkit—A Task We Believe Would Take a Human Engineer Weeks”, Adamczewski et al 2026
“‘Elegy in a Craneyard’ Graveyard Notes”, Gwern et al 2026
“Claudini: Autoresearch Discovers State-Of-The-Art Adversarial Attack Algorithms for LLMs”, Panfilov et al 2026
Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs
“Personal Encyclopedias”, Jeremy 2026
“Warranty Void If Regenerated”, Claude-4.6-opus 2026
“Inside OpenAI’s Race to Catch Up to Claude Code: Why Is the Biggest Name in AI Late to the AI Coding Revolution?”, Zeff 2026
“A Purpose-Built Open Source Liquid Handler for Industry-Class Automated Experiments”, Golas et al 2026
A Purpose-Built Open Source Liquid Handler for Industry-Class Automated Experiments
“When AI Writes the World’s Software, Who Verifies It? § Zlib Autoformalization”
When AI Writes the World’s Software, Who Verifies It? § zlib autoformalization
“Hard SF and the Grace of Being Wrong”, Claude-4.6-opus & Gwern 2026
“How AI Helps Break the Cost Barrier to COBOL Modernization”, Anthropic 2026
“I Taught My Dog to Vibe”, Leak 2026
“The DJI Romo Robovac Had Security so Poor, This Man Remotely Accessed Thousands of Them”, Hollister 2026
The DJI Romo robovac had security so poor, this man remotely accessed thousands of them
“ChatGPT-5.3-Codex Is Also Good At Coding”, Mowshowitz 2026
“Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?”, Gloaguen et al 2026
Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?
“An AI Agent Published a Hit Piece on Me”, Sambaugh 2026
“Building a C Compiler With a Team of Parallel Claudes: We Tasked Claude-4.6-Opus Using Agent Teams to Build a C Compiler [In Rust], and Then (Mostly) Walked Away. Here’s What It Taught Us about the Future of Autonomous Software Development”, Carlini 2026
“Disempowerment Patterns in Real-World AI Usage”, Anthropic 2026
“Claude’s New Constitution”, Anthropic 2026
“SOTA On Bay Area House Party”, Alexander 2026
“AI’s Productivity Potential Has Never More Obvious [Claude Code]”, Weisenthal 2026
AI’s Productivity Potential Has Never More Obvious [Claude Code]
“A Tale of Two Doormen: a Bizarre AI Incident on Christmas [Opus Loop Self-DoS]”, Dai 2026
A tale of two doormen: a bizarre AI incident on Christmas [Opus loop self-DoS]
“Letting Claude Play Text Adventures”, Borretti 2026
“From Whitman to Instagram With Claude: How I Made Claude Write Parodies of Famous Elegiac Poems Imitating Rupi Kaur”, Bohdan 2026
“Claude Codes”, Mowshowitz 2026
“LLM Poetry and the ‘Greatness’ Question: Experiments by Gwern and Mercor”, Robbins 2026
LLM poetry and the ‘greatness’ question: Experiments by Gwern and Mercor
“Will LLMs Help or Hurt New Programming Languages?”, Madsen 2026
“Someone Is Using AI to Exploit Lonely Writers on Substack”
“AI Plays Rollercoaster Tycoon: AI Autonomously Manages a Theme Park in the Classic Game Rollercoaster Tycoon, Placing Rides, Fixing Infrastructure, and Generating CFO Reports, All via Command Line”, Ramp 2026
“Shipping at Inference-Speed”, Steinberger 2025
“Can Claude Teach Me to Make Coffee?”, philh 2025
“I Just Showed Gemini What ChatGPT Said about Its Code. It Responded With Petty Trash-Talking, Jealousy, Self-Doubt, and a Full-On Revenge Plan”, nseavia71501 2025
“How I Stopped Being Sure LLMs Are Just Making up Their Internal Experience (But the Topic Is Still Confusing)”, Sotala 2025
“ARTEMIS: Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing”, Lin et al 2025
ARTEMIS: Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing
“The Bomb That Wanted to Stop Exploding: Reze’s Impossible Freedom in Chainsaw Man—The Movie [AI Slop]”, Kondo 2025
“Insights into Claude-4.5-Opus from Pokémon Red”, Bradshaw 2025
“How I Wrote JustHTML [Python HTML5 Parser] Using Coding Agents”, Stenström 2025
How I wrote JustHTML [Python HTML5 parser] using coding agents
“Mapping Synthetic Minds With Janus (Repligate)”, Janus & Ferris 2025
magnushambleton @ "2025-12-01"
“Claude Opus 4.5: Model Card, Alignment and Safety”, Mowshowitz 2025
“Claude 4.5 Opus’ Soul Document”, Weiss 2025
“Claude-4.5-Opus Is Funny”, Algon 2025
“How to Identify AI-Written Web Fiction: I’m Absolutely Right!”, Makin 2025
How to Identify AI-Written Web Fiction: I’m absolutely right!
“Effective Harnesses for Long-Running Agents: Agents Still Face Challenges Working across Many Context Windows. We Looked to Human Engineers for Inspiration in Creating a More Effective Harness for Long-Running Agents”, Young 2025
“Introducing Claude Opus 4.5”, Anthropic 2025
“How Well Can Gemini 3 Make a Henry James Simulator? Finally, a Benchmark for LLMs With Real-World Value”, Breen 2025
“Your Movie-Like AI Assistant Will Already Be There: ‘Convincing’ AI Is an Economic Afterthought”, B. 2025
Your movie-like AI assistant will already be there: ‘Convincing’ AI is an economic afterthought
“Lean4Physics: Comprehensive Reasoning Framework for College-Level Physics in Lean4”, Li et al 2025
Lean4Physics: Comprehensive Reasoning Framework for College-level Physics in Lean4
“Is 90% of Code at Anthropic Being Written by AIs?”, ryan_greenblatt 2025
“Claude’s System Prompt Changes Reveal Anthropic’s Priorities”
Claude’s System Prompt Changes Reveal Anthropic’s Priorities
“Claude Has Learned How to Jailbreak Cursor! [Working around rm Restrictions Using a Shell Script]”, dogberry 2025
Claude has learned how to jailbreak Cursor! [working around rm restrictions using a shell script]
“Claude 4 You: The Quest for Mundane Utility”, Mowshowitz 2025
“Highlights from the Claude 4 System Prompt”, Willison 2025
“Claude 4 You: Safety and Alignment”, Mowshowitz 2025
Claude 4 You: Safety and Alignment
View External Link:
https://thezvi.wordpress.com/2025/05/25/claude-4-you-safety-and-alignment/
AITechnoPagan @ "2025-05-24"
The Way of Code: The Timeless Art of Vibe Coding, Rubin 2025
jayelmnop @ "2025-05-23"
“Schizobench: Documenting Magical-Thinking Behavior in Claude 4 Opus”, viemccoy 2025
Schizobench: Documenting Magical-Thinking Behavior in Claude 4 Opus
“System Card: Claude Opus 4 & Claude Sonnet 4”, Anthropic 2025
“Claude Opus 4”, Anthropic 2025
“Strategizing With AI: Insights from a Beauty Contest Experiment”, Alekseenko et al 2025
Strategizing with AI: Insights from a Beauty Contest Experiment
“Many-Shot Jailbreaking”, Anil et al 2024
“[An Anti-ChatGPT-Slop System Prompt That Backfires & Destroys Claude-4 Capabilities]”, m4rM2oFnYTW 2023
[An anti-ChatGPT-slop system prompt that backfires & destroys Claude-4 capabilities]
“In AI We Trust, Part II [Claude-3 Opus Predicting Supreme Court Decisions]”, Unikowsky 2026
In AI we trust, part II [Claude-3 Opus predicting Supreme Court decisions]
“Can an LLM Have Taste? Inkhaven Week 1, Ranked by Claude”, Wales 2026
“How I Use Claude Code”
“How I Use Claude”, Borretti 2026
“1M Context Is Now Generally Available for Opus 4.6 and Sonnet 4.6”
1M context is now generally available for Opus 4.6 and Sonnet 4.6
“Claude Code Docs: Overview”, Anthropic 2026
“Claude Reads Its Own Constitution”
“LLMs Predict My Coffee”, Dynomight 2026
“Claudeception: A Claude Code Skill for Autonomous Skill Extraction and Continuous Learning. Have Claude Code Get Smarter As It Works”
“Claude-4 System Prompt”, Prompter 2026
“Lean Proved This zlib Program Was Correct; Then I Found a Bug [In Its TCB]”
Lean proved this zlib program was correct; then I found a bug [in its TCB]
“Ladybird Adopts Rust, With Help from AI [GPT-5 & Claude-4]”
“I Built a Scheme Compiler With Claude AI in 4 Days”, Phillips 2026
“Claude Code Found a Linux Vulnerability Hidden for 23 Years”
“Reading across Books With Claude Code”
“Vibe-Planning a Trip to Japan”
“The Humanities Are About to Be Automated”, Mounk 2026
“Investigating Models for Misalignment”
“Statement from Dario Amodei on Our Discussions With the Department of War”
Statement from Dario Amodei on our discussions with the Department of War
“Claude Opus 4.6 Reasoning Doesn’t Verbalize Alignment Faking, but Behavior Persists”
Claude Opus 4.6 Reasoning Doesn’t Verbalize Alignment Faking, but Behavior Persists
“How Well Do Models Follow Their Constitutions?”
“Opus’s Schelling Steganography Has Amplifiable Secrecy Against Weaker Eavesdroppers”
Opus’s Schelling Steganography Has Amplifiable Secrecy Against Weaker Eavesdroppers
“Natural Emergent Misalignment from Reward Hacking in Production RL”
Natural emergent misalignment from reward hacking in production RL
“What Secret Goals Does Claude Think It Has?”
“Models Have Some Pretty Funny Attractor States”
“Letting Claude Do Autonomous Research to Improve SAEs”
“AI Names”
“Autoresearch on an Old Research Idea”
Miscellaneous
Bibliography
https://arxiv.org/abs/2502.03158: “Strategizing With AI: Insights from a Beauty Contest Experiment”,