here's how Claude Code handles different types of tokens:
shipyard.build/blog/claude-...
05.03.2026 17:15
👍 0
🔁 0
💬 0
📌 0
here's how Claude Code handles different types of tokens:
shipyard.build/blog/claude-...
a token generator with the Claude logo, implying that this is the LLM taking input and generating code, token by token
Claude tokens == NLP tokens
these are the smallest units of text that LLMs use to process input/generate output (1 token = 1 word, 1 subword, or 1 letter, depending on your model)
Docker Hub page for node Docker image
Quay container registry repository settings page
there isn’t exactly a wrong choice for a container registry, but how to eval which is right for you?
- use the CR bundled with your cloud/git ecosystem
- OSS vs. proprietary: both $$ in different ways
- artifact support (OCI, helm, AI model)
- vuln scanning/IAM
engineering teams are improving their #ClaudeCode workflows every day. but CC can test its own work autonomously
a feature might take ~10 tweaks to get right, most obvious after running simple smoke/UI tests
give agents access to real infra so that they can self-verify + iterate
“Software reliability is down across the industry; failure rate is up: and batch size is up.” @lauratacho.com + @nathenharvey.bsky.social
Basically, AI is generating larger changes (batch size) and we know from research (eg DORA) that larger batch size tends to result in exactly this!
a cover image showing multiple claude code agents linked together in an orchestration pattern
some tasks are too complex for @claudeai subagents. that's where #multiagent 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 ⬇️
claude cowork cover image showing a laptop with the claude logo
#claudecode has always stood out for its filesystem + ops capabilities. in fact, CC is a CLI-based agent for anything, not just coding
they built a solid UI on top of Claude Code’s foundation to make a tool for file-related work, called Claude Cowork
Yes, we need code reviews.
In addition to reviewing intent, code reviews are an opportunity to review the how and ask questions: should we bring in these 3rd party dependencies, are we duplicating existing functionality, did we check in credentials by mistake, etc.
Most of this can be automated.
cover image for a block post showing a fireplace with three christmas stockings. each one has symbols that represent developer tools
engineering teams spend so much time during the year talking about dev productivity.
dec is a difficult time to keep those metrics up anyway. why not use the holiday stretch to catch up on your skills + learn new things?
here are six MCP servers that help our team code more efficiently ⬇️
shipyard.build/blog/best-mc...
AI #agents alone are pretty capable, but #MCP servers extend what they can accomplish
as a dev, you’ll want to connect your tools to your agent via MCP to handle more types of tasks
I sat down with @lauratacho.com to discuss DORA's last research into the impact of AI on software development and more!
Checkout the latest episode of DX's Engineering Enablement podcast:
getdx.com/podcast/dora...
#GBGB
ping us if you wanna talk ephemeral envs :)
happy cloudflare outage day to all who celebrate
OH at 1:37 am: “ok i have to go i have to start my #KubeCon slides”
are you ready to introduce a coding #agent into your dev loop?
at Shipyard, we're big fans of #claudecode. we wrote a guide with our best pro tips, so you can get high-quality code from your agent right away 🤖
#AIagents #developers
So excited to have been a part of #cdCon 2025!
Vote here to get us across the line as the Best #DevOps Event of the Year: www.surveymonkey.com/r/DevOps-Doz...
cd.foundation/blog/2025/11...
AI as an amplifier: DORA and GitLab on maximizing AI impact.
Tue, Nov 18, join @nathenharvey.bsky.social , Emilio Salvador, and Jessie Young for new research from @gitlab.com and the latest findings from @dora.dev
about.gitlab.com/the-source/a...
#AI #DevEx #GitLab #DORA #GBGB
shipyard logo next to MCP (model context protocol) logo, showing integration
we’ve got some HUGE news:
the Shipyard #MCP server is now in open beta!
use it for:
✅ writing + improving #E2E tests
✅ sandboxing #agent workflows
✅ live, in-depth code reviews
we’re looking for user feedback, so reach out with any questions/comments 👀
benjie at #DPESummit on moving faster by using #agents smartly:
- write agent config files
- learn your AI devtools + their capabilities
- outline their reqs, conventions, + other rules
watch the rest of this session with Erin Doyle on YouTube, courtesy of @gradle.com
SF folks:
we're partnering with our friends at @heavybit.com for this awesome #GitHubUniverse social hour/unconference. great way to network + reflect on today's learnings from one of our fav SF tech events 👏
still time to RSVP for this evening!
A Unix terminal listing the cron jobs installed on a local machine.
I've been running autonomous agents this whole time.
