This code was written in pure vibe-coding style — mostly for fun. I've got about 10 years of experience in IT, and even I fully agree: a 1000-line main file like this one probably deserves to be locked away in a secure facility.

But hey — if it's stupid and it works, it ain't stupid.

It was cool until 2022. Then LLMs started injecting these emoji everywhere and it became the chief marker for code/doc smell.

It never caught on but I always liked setting the jumbo header images on Notion docs to creepy, unsettling pictures from Unsplash.

Need to write a document about converting a Rust project to Typescript? A picture of an abandoned warehouse full of expressionless baby doll heads fits perfectly.

Looks so unprofessional, lol, says the guy wanting to use a free app, this isn't a microsoft made app lol it's a guy making a github app for free the audacity people have these days to shit on peoples project for 0 reason

AI coding where the human stays in control and reads and confirms code is totally different from vibe coding where you don't read code and just prompt until it sort of works.

Yeah I noticed that too, it’s a bit crazy to me that stuff like this is getting upvotes and traction. It feels like it was vibe-coded in one-shot style without perhaps even reading any bit of the code. A bunch of hardcoded values, a `sleep(3)`, bunch of other antipatterns.

Up until recently I tended to “trust” github repos a bit more, now I feel like I need to have my guard up so I don’t fall into a trap of using something like this. Funnily enough a good first metric for me now is # of emojis in the readme - the more emojis the more likely you should stay away from it

I've been wondering this too. They said on a podcast that the average usage is around $6 a day in credits but I really question that. If that's the case though and they do have a lot of Pro and higher tier subscribers that might be making up the difference.

Huge margins on API lol, there's a reason the guys from deepseek told everyone that its bullshit that the big boys aren't making money, research is expensive but the inference isn't especially with the comments that even Sam Altman mentioned regarding pricing per response for the chat clients.

I have not ever actually hit the rate limit as far as I'm aware. I have gotten the "approaching Opus usage limit" a couple times but have not hit it.

Has to be wrong. I'm on that subscription as I wanted to reinforce my opinion that it's still shit for devs that actually have experience, like it was a few months ago.

While my plan didn't pan out, cuz it was way too effective, I can confidently say that I'm going through 3-6k tokens per prompt on average, and usually get around 3 hours of usage before I'm hitting the rate limit.

The limit is probably closer to 300k then <10k

Also the chat interface doesn't have a separate limit, once you hit it via Claude code, you cannot use the website either anymore.

Maybe it's a 7k limit per prompt? Dunno if I exceeded that before

I tried it with Roo Code (with 3.7 Sonnet, not Code). For agentic use you will probably hit the limit from your first prompt/assignment if it does some browser tool use.

actually reading more about how you ingest ~/.claude/projects//.jsonl the wrapper method is probably over kill..

how about making it a tool that claude can use directly?

You can get the un-redacted prompts via the OTel event logger[0], but unfortunately it won't give you responses. You could open up a GitHub issue to request that addition!

Disclaimer - I work at Anthropic but not on Claude Code, the team is responsive via GH issues though!

[0] https://docs.anthropic.com/en/docs/claude-code/monitoring-us...

Wow, I had no idea Claude Code had this kind of verbose logging turned on by default. Looking around I also found a cool tool for converting these logs to HTML format for easy viewing: https://github.com/daaain/claude-code-log

It's definitely interesting. Probably, one could also see how much of the context is unnecessary to optimise usage. I wonder if one could also detect whether cache was used.

Is there anyway to ensure these logs are output in a current project folder? Would like to store a copy in git for reference

This tool is for plans with a fixed cost — like Pro Max x5 or x10 — not for API usage.

Let's start with npm and node_modules first - I think it is responsible for at least half of the global warming.

We all were low carbon developers once

I don’t understand the concern of this. Let’s say it says you’ve used 1.521kg of CO2 today. How is that actionable?

A single flight from JFK to LAX produces around 20,000kg of CO2. Using the 8.3g value means a flight is equivalent to 2.41 billion tokens.

I just hit my usage limit this morning, and the article shows how to get these numbers so I can answer this myself -- I'd be curious to see what other people are doing token wise day-to-day

ccusage says I had 1k input tokens, 12k output and 1.2m cache create.

I'm not sure if that is 18.3g, 138.3g or 1213 * 8.3g.

At the highest number that's 10kg or 25% of average US daily per capita emission or 1 gallon of gas.

By any chance do you remember the source for that? Googling "1000 tokens emits 8.3g CO2" only gets me this thread.

Soon the AI companies will have different tiers of plan based on carbon output.

For $200 per month they’ll only use carbon-free power sources for your prompts. At $100 they’ll use nuclear, then $20 per month for coal, and then there will be a free tier where your prompts are powered by throwing baby seals into a furnace.

I used to think the same thing, and I still want a login/password alternative for me. But after seeing normies use online accounts and the trouble they have with password managers, I realized one of my friends had the most secure process: she would create extremely high-entropy passwords everywhere, but not remember them. Once she's logged out of her sessions after a couple of months, she uses the password forgotten link to generate another password, and so on. So her passwords are never stored anywhere, she's immune to many login stealing phishing attempts through genuine-looking fake websites as she can't enter the password, she doesn't have to deal with syncing the passwords between all her devices, and she doesn't have passwords on a post-it on her workstation. And she also doesn't get those annoying emails saying "your password is 6-months old, please change it or else!".

The email auth flow is a simplified and more efficient way to achieve the same outcome.

Another reason to ditch email-to-login: It decreases active user count because login is such a chore. I only use those websites if I absolutely have to and there are no alternatives.

Yeah, it's the reason I eventually tuned out of Slack too (obviously I don't use it day to day). I used it for some things, some community stuff, every now and then.

So when I open it up on a new machine or after months, you have to go through that magic link bs multiple times! For all your accounts/channels! I did that 2 or 3 times and then I just stopped using it.

I don't even remember what I signed up for over the years, I know some of it was nice (like an LoRa IOT channel, an "AWS professionals" channel, I set something up for the LoRaWan network in the previous city I lived in... All just tuned out because of the login bs.

Oh, thanks! You didn't go digging that up just for me, did you?

I'll try it out right now.

No.

https://console.anthropic.com/settings/billing

A lot of it was from the default behavior of Claude re-reviewing a PR on every commit.

It also just psychologically just saps the fun out of something when there's a fee attached per invocation. And the UX is that I buy $50 of API credits, time passes, and then it breaks because I have to refill it again.

If it were literally a "fix issue for 50 cents" button that charges you 50 cents and that's that, then it would be different. But instead your API credits drain and you can't evaluate if it was worth it.

it is on my todo list