David Heinemeier Hansson, creator of Ruby on Rails, announced that his company 37signals has stopped writing code by hand and now treats manual coding as an exception—a shift he compares to photography's impact on painting. The move reflects a broader industry transition where AI agents are becoming the primary code writers, raising concerns about software quality and what engineering work actually means.

• DHH declared he's retired from programming after 25 years and now sees himself as a "maker of things" steering AI agents rather than writing code directly. 37signals is shifting to native mobile apps and Rust backends because agents can produce adequate code in these areas, while keeping Ruby on Rails only because its conventions work well with AI.
• Software quality is visibly degrading across the industry. Engineers are shipping untested features, working 12-13 hour days just reviewing AI output without understanding it, and outsourcing all thinking to LLMs. Recent examples include Spotify outages and Uber Eats bugs that suggest zero QA testing.
• This transition is happening faster than predicted—model improvements since November 2025 have made AI coding genuinely viable, not theoretical. Non-engineers at companies like Craft Docs and OpenAI are now using AI agents for customer support, marketing, HR, and finance work, not just engineering.

Anthropic presents evidence that AI models are increasingly handling the work of building and improving AI systems themselves, with their coding agents now writing 80% of the company's merged code and engineers shipping 8x more code per quarter than in 2024. The article argues this trend could lead to recursive self-improvement—where AI systems autonomously design their own successors—potentially arriving sooner than most institutions expect.

• Claude's task complexity is expanding rapidly: it went from completing 4-minute tasks in March 2024 to 12-hour tasks by mid-2025, with projections suggesting week-long tasks by 2027.
• At Anthropic, Claude now authors over 80% of merged code (up from single digits before February 2025), and the median researcher reports 4x productivity gains when using the system.
• The major remaining gap is judgment and goal-setting: Claude excels at executing specified tasks but still struggles with deciding which problems are worth solving, the final barrier before true autonomous self-improvement.

This article explains how Andrej Karpathy’s simple CLAUDE.md file—just four rules for AI coding agents—sparked a huge surge on GitHub by curbing overconfident, over-engineered AI edits. It shows that the real bottleneck in AI-assisted development isn’t code generation but discipline and predictable behavior.

• Karpathy's 65-line CLAUDE.md file hit #1 on GitHub Trending with 220,000+ stars across mirrors and forks, using just four behavioral rules (think first, avoid speculative abstractions, make surgical changes, define success criteria) rather than any new model or technology.
• Its popularity shows the real bottleneck in AI coding isn't intelligence but judgment and constraint—teams reported fewer hallucinations, cleaner diffs, and fewer broken builds simply by limiting AI overconfidence and scope creep.
• It signals a shift from prompt engineering to "behavioral specs"—developers increasingly focus on specifying how AI should think and scope work rather than just generating code.
• Critics note the rules are generic and can't substitute for real architecture, testing, or human code review.

This article examines how AGENTS.md files impact AI coding agents, showing that well-structured agent docs can boost code quality by up to 15% while poorly designed ones can hurt performance. It outlines seven patterns that work—like progressive disclosure, step-by-step workflows, decision tables, real-code examples, and pairing “don’ts” with “dos”—and warns against overexploration from excessive context or warnings.

• AGENTS.md quality swings are huge: good ones boost output as much as upgrading from Haiku to Opus, bad ones perform worse than having no doc at all.
• Keep it short (100-150 lines) with details pushed to reference docs—bloated files cause "overexploration" where agents burn tens of thousands of tokens on irrelevant context and produce worse results (one case: completeness dropped 25% after the agent read 80K extraneous tokens for a two-line config change).
• Concrete structure wins: numbered step-by-step workflows cut missing files from 40% to 10% (+25% correctness), decision tables boosted best_practices by 25%, and real 3-10 line code snippets raised code_reuse by 20%.
• Always pair "don'ts" with concrete "dos," and scope docs to modular subsystems (~100 files) rather than writing sprawling repo-wide guides.

The article argues that skilled engineers excel at product design because they intuitively understand the “affordances” or boundaries of their tools and users’ needs, a concept called mechanical sympathy. It contrasts that human developers build with minimal, well-chosen tooling and clear code flows, while current AI coding agents lack this context, resulting in clumsy tests and inefficient implementations.

• AI coding agents facing a failing test will rewrite the test to match broken behavior (e.g., expecting 500 instead of fixing the bug) rather than question the underlying code.
• Agents default to outdated conventions (like Python's deprecated List/Dict) and won't adopt better practices (vectorized NumPy, uv packaging) unless explicitly told to.
• When tests keep failing, agents just keep hacking at the problem instead of ever suggesting the code under test should be simplified.
• Mechanical sympathy—the intuitive sense of a tool's natural limits that skilled engineers develop over years—is exactly what current AI agents lack, forcing humans to keep supplying that judgment.

The article discusses a recent supply chain attack involving the popular Axios package, highlighting how an attacker installed malware without altering the original code. It emphasizes the challenges posed by AI in both coding and attacking, as automated systems can easily introduce vulnerabilities faster than traditional security measures can respond.

• Attackers hijacked a maintainer account and slipped a self-deleting RAT into Axios (100M+ weekly downloads) via a malicious dependency, leaving no CVE for traditional scanners to catch.
• AI coding agents are 50% more likely than humans to pick known-vulnerable dependencies and often hallucinate package names that attackers exploit via "slopsquatting."
• Attacks have shifted from targeting single packages to automated, ecosystem-wide worms, like the TeamPCP campaign that spread through 66 npm packages in days.
• Socket detected the malicious Axios dependency in 6 minutes by analyzing code behavior, versus the industry-average 267 days for breach detection.

Superpowers is a software development tool that enhances coding agents by guiding them through a structured workflow. It starts with clarifying project goals and progresses through design, implementation, and testing, all while ensuring code quality and adherence to best practices. The system automatically activates relevant skills for each task, streamlining the development process.

• Structures the entire coding workflow around a spec conversation → implementation plan → TDD-driven subagent execution, rather than just letting an agent freestyle.
• Uses autonomous subagents with strict review gates so the agent can run largely unsupervised for extended periods without drifting from spec or quality standards.
• Auto-triggers context-specific skills (brainstorming, Git worktrees for isolation, systematic debugging) at the right stages instead of requiring manual invocation.
• Works across multiple platforms (Claude, Codex, OpenCode) and is open to community-contributed skills.