Shipyard

Up until recently, there was a very real tradeoff between delivery velocity and stability. Shipping fast meant monitoring new deployments really closely. Fast forward to 2026, and we have a new environment “stack” that developers and coding agents move with, not around: https://lnkd.in/e2a-rbqH

Now you can schedule your Claude Code tasks to repeat on a loop. Devs are finding this super valuable for PR babysitting: tell Claude to watch your PRs, react to build logs, and resolve any issues brought up in the comments. Here's how to use the /loop command: https://lnkd.in/e3mb_t_b

What does it mean when you hit your Claude Code token limit? Claude tokens == NLP tokens: the smallest units of text that LLMs use to process input/generate output (1 token might equal one word, a subword, single letter, depending on your model) Here's how to understand your token usage, and how Claude handles different types of tokens: https://lnkd.in/eVDHSd9Q

Most container registries are functionally similar, so there isn’t exactly a wrong choice, but there are a few things you’ll want to consider: - Do you like the one bundled with your cloud/git ecosystem? - OSS vs. proprietary: they'll cost you in different ways - Artifact support (OCI, Helm, AI model) - Security scanning/access management https://lnkd.in/eZCjvHpc

Engineering teams are improving their #ClaudeCode workflows every day, but making Claude iterate and test its own work is still a challenge. Agents rarely get a complex feature right after a one-shot, and they usually need human intervention to give them the feedback they need to iterate. A feature might take ~10 tweaks to get right, many of which are obvious after running simple smoke or UI tests. When you give agents access to real infra and give them rules to self-verify, you can get stronger outputs *and* your agent can work longer autonomously. Learn why ⬇️ https://lnkd.in/g-wnuGUP

Some tasks are too complex for Claude subagents. That's where multi-agent orchestrators come in, allowing a lead coding agent to delegate tasks and spawn agents as needed. In this post, we take a quick look at Gas Town and Multiclaude ⬇️ https://lnkd.in/eTeh3iDh

6 months in, #ClaudeCode and #GeminiCLI have become two of the most established CLI-based coding agents. They're both driven by SOTA models (Opus 4.5 and Gemini 3 Pro). Here's how they compare. https://lnkd.in/g8c9tETM

Claude Code has always stood out for its filesystem and ops capabilities. Anthropic has made it clear that CC is a CLI-based agent for anything, not just coding. Anthropic saw an opportunity here: they built a solid UI on top of Claude Code’s foundation to make a tool for file-related work, releasing it as Claude Cowork. https://lnkd.in/gtMuRjTF

Engineering teams spend so much time during the year talking about developer productivity. December is a difficult time to keep those metrics up anyway, so why not use the holiday stretch to catch up on your skills and learn new things? Here's a model that we like to follow: dedicated developer enrichment time. Find out why: https://lnkd.in/ebNtuMGE

The Linux Foundation's latest project, the Agentic AI Foundation, was launched after major project contributions from Anthropic (#MCP), Block (#goose), and OpenAI (AGENTS.md). It introduces a new open standard for #AI devtools/integrations, which could point to a more user-centric + collaborative model. Here's what this means for devs: https://lnkd.in/eM4RxUMJ