First, the comparison of running an Anthropic model in Bedrock vs. Claude Code is not apples-to-apples. With Bedrock pay API pricing (or whatever negotiated pricing AWS and the vendor agreed to), just like the other providers in Bedrock. You know what else is more expensive/token than a Claude Code subscription? API access for pretty much all of the other big providers. Whether we like it or not, Anthropic sees subscrition usage as different than API usage. This is not at all an AWS thing.

Second, regarding Lambda: the cold start criticism is fair but for heavy workloads usually isn't an issue and for light workloads there are some workarounds. But, if you have a discreet function that can exist outside of your app, Lambda still may be a better option here. As far as the "MASSIVE development complexity" I wish the author cited specifics. I feel like this can be true of any technology and depends on how you go about setting it up and implementing it. I've used Lambda for all sorts of use cases, including a full web application (I'd advise staying away from this one!) and there definitely is a sweet spot where it can work well.

Lastly, the whole bit about using AWS WorkMail would also suggest they never really "left" AWS. But maybe I misunderstood that part as I'm not very familiar with that service.

Off topic: what’s the history behind the naming of Pull Request (PR) vs. Merge Request (MR)? I understand why both can be considered “correct”, but I’m curious why, say, GitHub uses PR and Gitlab uses MR.

Plenty of engineers have loose (or no!) standards and practices over how they write coee. Similarly, plenty of engineering teams have weak and loose standards over how code gets pushed to production. This concept isn't new, it's just a lot easier for individuals and teams who have never really adhered to any sort of standards in their SDLC to produce a lot more code and flesh out ideas.

Note: I am a CloudFlare customer, but on a very low plan and probably have no use for this.

Recently, I asked Claude to log into my local grocery store chain's website and add all of the items from my shopping list to a cart. It was hilariously slow, but it did get the job done.

Unless I missed it, the article doesn't explictly mention speed in the copy, but the results do show a 17 minute (!!!) total time for the vision agent vs. 0.5s - 2.8s for the API approach.

A big part of the challenge with vision is that to manipulate the DOM, you first have to be sure the entire (current) DOM is loaded. In my experience this ends up in adding a lot of artificial waits for certain elements to exist on the page.

Comments URL: https://news.ycombinator.com/item?id=48027474

Points: 2

# Comments: 0

There’s nothing stopping Google Chrome from doing something similar except, I suspect, Google knows or feels it will result in many fewer installs of its bloatware.

Comments URL: https://news.ycombinator.com/item?id=47922728

Points: 7

# Comments: 1

If a junior or new employee made this mistake, it would be because you, as the founder, and your engineering team, didn’t have protections in place from editing/destroying production data for this particular scenario.

Using best practices and least privilege principles is more important now than it ever has been. For those of us with our hands close to button, we should be always mindful of this now more than ever.

Comments URL: https://news.ycombinator.com/item?id=47883034

Points: 3

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=47870409

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=47862241

Points: 12

# Comments: 7

I thought most modern laptops have dedicated video decode hardware that is fairly easy on battery. At only 250nit though...that seems dim by today's standards. I'm happy to be wrong though!