adam brinkman

Backend @kognic_ Gothenburg. XC ski when I can.

[adam brinkman @brinkman_adam]

Very excited to try the concise output style!

[ClaudeDevs @ClaudeDevs]

You can now set Claude Code's output style to Concise.

Claude leads with the result, keeps responses short, and still gives full detail when you ask.

Turn it on in /config → Output style, or set "outputStyle": "Concise" in settings.json.

[adam brinkman @brinkman_adam]

How to keep Claude Code running:

lid open -> caffeinate -is claude
lid closed + external display + power -> clamshell, nothing needed
lid closed, no display -> requires disabling sleep

[adam brinkman @brinkman_adam]

It's funny how Claude estimates work like it's not a superhuman coding agent.

"Total scope is bigger than I implied earlier — sorry. It's now ~30 lines of new query-client code + 2 v5 migrations. Probably half a day, not a couple hours."

Then finishes before I finish my coffee.

[adam brinkman @brinkman_adam]

Been doing this manually a couple of times without realizing it should just be a skill

[Matt Pocock @mattpocockuk]

/handoff might be my new favourite skill

[adam brinkman @brinkman_adam]

My Claude Code setup includes auto-memory + global rules + per-project rules.

A favourite global rule is one about how I want Claude to use git. I want short, clear commits, and I want to use worktrees for new features.

Interested in hearing what other rules people use 👀

[adam brinkman @brinkman_adam]

Claude makes it really easy to make small experiments like this. I spent a few hours this weekend building a simple environment and training a self-driving model. Super fun!

[adam brinkman @brinkman_adam]

@utzle Det ska sägas att jag inte skriver så mycket mail, men att söka i slack och draft:a svar hanterar Claude Code utmärkt!

[adam brinkman @brinkman_adam]

@utzle Jag kör bara Claude Code och använder det till allt. Har skills för att itererat på setup:en också. Anthropic har också släppt en skill relaterat till detta, /fewer-permission-prompts. Har dock inte hunnit testa den så mycket än.

[adam brinkman @brinkman_adam]

A team working on software isn’t automatically more productive than the sum of its parts.

Something I keep coming back to:

1/ Make sure there's a shared goal. Ideally a small, focused one

2/ Figure out what you can do right now to move toward it

We do this every morning

[adam brinkman @brinkman_adam]

Lots of repos at Kognic, so I built a tiny menu bar app to track my PRs using SwiftBar. Polls GitHub every 5 min, shows the status of my latest PR, and lists all open PRs when opened. I use it! Idea -> Claude -> done in ~15 min.

Can recommend!

[adam brinkman @brinkman_adam]

inspired by @karpathy’s framing

[adam brinkman @brinkman_adam]

How do you claim leverage from agents without compromising the quality of the software? Claude is probabilistic.

CI is one part. Should you generate as many tests as possible?

[adam brinkman @brinkman_adam]

I merge significantly more PR:s now compared to pre-Claude. About 2-3x as many. Curious to see what that does over time. Quantity does matter for quality, but I used to force myself to think harder about prioritization

[adam brinkman @brinkman_adam]

I think a lot about how to close the gap between building things and getting real-world impact. "Iterate fast" sounds easy. It would be, if verifying that what you build is good was easy. It's not. How do you make a small iteration of a grand vision that proves its usefulness?

[adam brinkman @brinkman_adam]

@AnandButani Indeed. What's your go-to use case?

[adam brinkman @brinkman_adam]

Three things I love Claude Code for at work:

1/ /triage skill with Grafana MCP (for logs) + repo access to fix bugs. Alert in, PR out

2/ Code reviews by pointing it to a PR link

3/ Cross-service migrations spanning multiple repos and languages

[adam brinkman @brinkman_adam]

I was aware of the word limit, but never realised how hard it would be to actually compress anything useful into a text that short

[adam brinkman @brinkman_adam]

First post on X 🤯

Too late? I don’t know. And I don’t think there’s a way to find out without trying. It's an experiment. I will stay as long as I find it valuable.

I work 100% remote from Östersund. Excited to see if I find interacting on X valuable to me. Hello people!

[DeepLearningThesis @deep_thesis]

First official day of the thesis at @chalmersuniv! Follow updates on our blog https://t.co/9tucPHFmK4, first post out now! #DeepLearningThesis #MasterThesis #chalmersftw #ai #ConvNet #inspiration