Software Architecture

Anthropic migrated its monorepo to in-house source control in five months after Git became a bottleneck. Daily merged changes rose from roughly 2,500 in May to more than 30,000 in September.

scriptc v0.2.8 adds @scriptc/threads with Worker-compatible APIs that share immutable object graphs by reference. The same programs use cloning when run in Node.js.

Rabi Guha's teardown says ChatGPT generates DIL that a server compiles for a client component renderer. Tool values are referenced by ID rather than retyped, and no public API or SDK is reported.

Next.js 16.4 made Cache Components the default for new apps and added APIs for static output and partial navigation rendering. The release also adds agent-guided upgrades and cuts Turbopack disk-cache size by 20–25%.

The gdp-ts library, linter, and AI skill require typed proofs that authorization checks ran before sensitive calls. Its rules also flag forged proofs and unsafe escape hatches.

Textual disabled public PRs after low-quality AI submissions became unmanageable. Practitioners cited cargo-cult code, weak harnesses, long-horizon failures, and AI-generated PR spam as reasons to keep review and tests in the loop.

Robert C. Martin argued tangled code confuses agents and promoted agent-led refactoring with tests. Other reports cited high Sol spend, under-10% unattended completion, and failures once bugs or complexity appear.

Reports described GPT-5.6 Sol adding needless abstractions and Fable spending quota on many subagents for small changes. The debate frames lighter setups and senior review as safeguards against agent-made tech debt.

Simon Willison highlighted Jarred Sumner’s Bun rewrite from Zig to Rust as an AI-assisted workflow using trial runs, dynamic planning, adversarial review, and verification. One estimate put API token cost near $165k.