Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Nate Berkopec Profile picture
@nateberkopec
Mar 9 • 1 tweets • 1 min read • Read on X
Serving your CSS/JS from a different subdomain (assets.mydomain...) than your document (www.mydomain...) is costing you 100+ milliseconds on your LCP for no good reason. Here's how to use DevTools to inspect and repro this, and why you need to use chrome://net-internals:

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Nate Berkopec

Nate Berkopec Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @nateberkopec

Nate Berkopec Profile picture
Jul 15
I'm sold. Agentic coding is the future of web application development. There is no going back. Close the editor. Open Claude.

Your job is now to manage, review, corral and improve a relentless junior dev who is working on 6+ PRs in parallel.
If your mental model of LLM-mechanized coding is just tab-autocomplete in Cursor, you should try closing your editor for a week and only work out of Claude Code w/a parallel workflow using git worktrees (many tools now do this). Work on improving the agent if output is bad.
I'm less sold on agentic coding in the following contexts:

1. Libraries
2. Extremely large open-source codebases with long history (i.e. the METR study)
3. Concurrent/distributed systems
4. Probably more I can't think of. It's magical autocomplete, not a panacea.
Read 11 tweets
Nate Berkopec Profile picture
Feb 14, 2022
ActiveRecord performance puzzle: "SELECT <long-list-of-columns>" sometimes takes ~10-40x longer than "SELECT *", randomly. Issue is not at the database, where query is consistently <1ms. Issue reproduces in Rails console.
Basically run it 10x in the console and you'll see 1ms, 1ms, 1ms, 40ms, 1ms, 1ms, 40ms, 1ms, 1ms.
More data: it isn't GC, it doesn't appear to be logging, and it does not occur in production
Read 5 tweets
Nate Berkopec Profile picture
May 7, 2021
In a production environment, the optimal Puma configuration for 80% of apps is probably 4 workers, 5 threads on a 4vCPU machine with ~8GB of memory.
and yes, in k8s, this would be per-pod, horizontal pod autoscaler running on top of that with request queue time as the metric
8GB for 4 workers is probably more memory overhead then you're ever gonna need but that 1vcpu:2GB ratio is the lowest that most cloud vendors sell so lower memory usage doesn't get you much.
Read 4 tweets
Nate Berkopec Profile picture
Mar 12, 2021
I put in a big PR to speed up my client's most important background job by 10-30%. This was really important because it generates 90%+ of their load. Let me walk you through it, commit-by-commit. (1/6)
First, we write some simple scripts to benchmark and profile the code in question. I used Benchmark from stdlib here because running this code once takes about 40 seconds. Note the warmup and use of transactions. github.com/WikiEducationF… (2/6)
Next, we fix an N+1. This is an example of what I call an "Enumerable" refactoring for an N+1. Rather than use the DB to do things like sum a column, we do it in Ruby. This is faster here b/c we were already going to load all the data anyway. github.com/WikiEducationF… (3/6)
Read 6 tweets
Nate Berkopec Profile picture
Apr 12, 2020
All of these qualities are not ineffable. We can (and should) quantify them: LOC, dependency LOC, hours spent fixing bugs and dealing w/incidents, and server costs. Your customer has requirements for these. You don't need to magically guess them on their behalf: you have to ask.
Craftsperson attitude is about delivering the highest possible software quality regardless of customer requirement. This is wrong. You should deliver what the customer requires: no more and no less.
And of course, you are the expert: so it's your role to help shape these requirements. Customers can be misguided about what scalability they require (usually they overestimate), or what maintenance they'll need to do (usually they underestimate).
Read 7 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

Send Email!

:(