Laravel 13.34, weekly updates and tip

Laravel 13.34

This week brings us to Laravel 13.34. Here are the highlights:

  • Add --queue option to queue:flush in #61630
  • Allow jobs to count worker crashes towards maxExceptions in #61737
  • Allow assertDatabaseCount to accept an array in #61727
  • Support LazyCollection values in between clauses in #61678
  • Support named S3 credential providers and Cloud IAM disks in #61676
  • Add Schema\Builder::getColumn() in #61759
  • Include duration in JobProcessed event in #61672

You may review the full branch diff on GitHub for a complete list of changes.

Weekly Journal

Last week I did four livestreams converting more first-party Laravel packages to Double. Nearly all of them have been merged. Given the adoption of Double by the Laravel ecosystem and that I have yet to run into any issues converting a test suite, I went ahead and tagged v1.0.0. If nothing else, to signify Double is ready for wider adoption.

With the time I had outside of the streams, I continued my quest to improve the Laravel framework test suite. While this is good in general, it may also ease the potential conversion to Double. Unfortunately, this review burned through all my tokens, tripping my weekly rate limit. So Saturday I decided to use my only reset.

With my limits reset, I switched gears to increasing the AI integrations for Shift. This has turned into a multi-part effort, including:

  • Finalizing the Shift MCP server.
  • Adding "AI Review" to all Laravel Shifts going back to 5.0.
  • Reviewing every PR comment for potential copyable prompts.
  • Redesigning the home page to emphasize "Shift + AI".

Weekly Tip

As I was burning through my tokens, I discovered /statusline in Claude Code. This allows me to customize the line under my prompt. Much like my Bash prompt, I had an agent add the current working directory and Git status. But I also had it add the "raw token count" of my current session.

This let me immediately notice that even though I ran /clear, the new session consumed 80k tokens. After some prompting, I realized this was the "startup cost". Basically all your system and project guidelines, skills, MCP server resources, etc. All those MD files read into context.

I didn't need half of them. I was able to lower my startup cost to 36k tokens. So, the tip is, check what gets loaded for each one of your sessions.

View Archives →