5 Quick Links for Devs: Week 13, 2026

Announcing TypeScript 6.0

This is a big one. TS 6.0 is explicitly a bridge release to prepare us for 7.0 (the native Go port), and it comes loaded with opinionated default changes: strict is now true by default, target defaults to the current-year ES spec, es5 is deprecated, and types defaults to [] — which alone can improve build times by 20-50%. If you've been lazy about your tsconfig, this release is going to have opinions about that.

Slowing Down in the Age of Coding Agents

While everyone else is trying to parallelise five Codex tasks at once, this dev went the opposite direction: e-ink tablet, handwritten annotations, coffee shop mornings reading design docs with a pen. The core argument is that the bottleneck in agent-assisted development is no longer writing code — it's the quality of thinking you feed into the context window, and that work can't be parallelised.

Auto mode for Claude Code

Anthropic just dropped an "Auto" mode for Claude Code, which basically lets the CLI loop through its own terminal outputs and file changes until the task is complete. Auto mode uses a classifier to let safe actions proceed automatically while blocking potentially destructive ones, which means you can actually kick off a task and walk away without worrying about your file system getting nuked.

Engineers do get promoted for writing simple code

A nice counterpoint to the cynical "nobody gets promoted for simplicity" take. The argument is straightforward: engineers who write simple code ship faster, accumulate a longer list of successful projects, and build a reputation for delivering with minimal fuss — and non-technical managers notice all of that, even if they can't judge the code itself.

A Coherent Vision for the Future of Version Control

Bram Cohen (yes, the BitTorrent guy) releases Manyana, a ~470 line Python demo that makes the case for CRDT-based version control. The key insight is that merges never fail by definition — conflicts become informative annotations rather than blocking operations — and rebase can preserve full history instead of creating fictional commit timelines. It's a proof of concept, not a product, but the conflict presentation alone is a compelling improvement over git's opaque blobs.