.Dev @AnExiledDev

I work with AI all day, usually Claude, always willing to share what I know, occasionally I share cool stuff I've built. I also run a Claude changelog site. changelogs.core-directive.com Joined June 2015

Tweets

1K
Followers

110
Following

80
Likes

142

.Dev @AnExiledDev

9 hours ago

@MarkGerrardAI My issue before wasn't what work could get done, it worked just fine, just considerably lower quality. The work I can effectively get done now is quite a bit lower. The quality is for sure higher, I can trust it more, but I can no longer work in parallel. I do less now than befor

0 0 0 0 0

View Details

.Dev @AnExiledDev

a day ago

Once the usage reduction goes live, assuming they don't stack an increase on top, I'll share what the new average is. Currently we're sitting at about $23-$25 of API equivalent usage per 1% of the 7d limit on 20x. [1/3]

2 0 4 119 0

View Details

.Dev @AnExiledDev

10 hours ago

@MarkGerrardAI Except OpenAI provides more usage than Anthropic for $200, even at the currently increased rate, which is still several times smaller than it was last year.

1 0 1 13 0

View Details

.Dev @AnExiledDev

a day ago

@RhysSullivan Yeah but openai gives way more usage than anthropic, that's the difference. Our current usage is also 3x lower than this time last year. Context is important.

0 0 0 25 0

View Details

.Dev @AnExiledDev

a day ago

If there's actual interest, I'll clean mine up into something properly distributable to share. I think sharing statistics like these is important. If you're new, remember, usage resets when you upgrade from Pro->5x->20x, abuse that. Multiple 5x accounts is the best value.

0 0 1 24 0

View Details

.Dev @AnExiledDev

a day ago

That comes out to roughly $2,400 per week on 20x. From people I know on the Pro plan, they are getting roughly $300 a week. So still not near 20x, not even half. If people would be interested in a dedicated tool for tracking these metrics, let me know. [2/3]

1 0 1 21 0

View Details

.Dev @AnExiledDev

a day ago

@TheAlbrechtCorp @ClaudeDevs People still pretending this myth is real?

0 0 0 50 0

View Details

.Dev @AnExiledDev

2 days ago

Soooooo 25% lower rates? Rates that are already 3x smaller than last years rates? Thanks, I guess?

0 0 0 21 0

View Details

.Dev @AnExiledDev

2 days ago

Sounds like Astra is slated for a Thursday release. Curious if Anthropic is going to continue matching release dates or release earlier. I think attempting to compete for the same day is a terrible idea personally. Astra looks amazing, granted I'm sure what we get will not be.

1 0 1 91 0

View Details

.Dev @AnExiledDev

2 days ago

@DanDr1s This happens every single time. Then when the model drops, the benchmarks are much worse. Enjoy it while you get it, because what the rest of us get won't be this good. It's just marketing, as always.

1 0 4 269 1

View Details

.Dev @AnExiledDev

2 days ago

@kloss_xyz @bot Wow you're shilling like you're paid to.

1 0 0 623 0

View Details

.Dev @AnExiledDev

2 days ago

@Ananth7e I can confirm not nerfed. I have tooling that monitors 7d usage percent to API spend. Still holding steady at $25 per 1%

0 0 1 30 0

View Details

.Dev @AnExiledDev

2 days ago

@DanDr1s Then. Fix. It. It's not hard, a few paragraphs gets you most of the way there. Just give it your expectations. If a model follows instructions well, nothing else matters, any issue you find is solved by giving it instructions. People just run it default which is laughable.

0 0 0 237 0

View Details

.Dev @AnExiledDev

3 days ago

@ClaudeCodeLog changelogs.core-directive.com/v/2.1.251

0 0 0 455 0

View Details

.Dev @AnExiledDev

3 days ago

Claude Code v2.1.250 changelogs.core-directive.com/v/2.1.250

0 0 2 236 0

View Details

.Dev @AnExiledDev

3 days ago

Claude Code v2.1.248 changelogs.core-directive.com/v/2.1.248

0 0 2 293 1

View Details

.Dev @AnExiledDev

3 days ago

Alright pipeline is running, looks like 2.1.250 also just landed, working them sequentially.

0 0 0 50 0

View Details

.Dev @AnExiledDev

4 days ago

New Claude Code version has appeared, but I'm delaying the pipeline run a little bit. I'm elbow deep in optimization work and I want that to land before I run the pipeline so I can get hard data on how effective my changes are. Plus it'll save me some tokens :D

0 0 1 362 1

View Details

.Dev @AnExiledDev

4 days ago

@ffuhyyyy @devinjameson No because you instructed it to "rewrite your last message" meaning it sends the rewrite, it hides the verbose one you can't easily read only. You see the stop hook and the rewrite when it's finished. I'm just offering solutions, take it or leave it. Bitchin' doesn't fix much.