Hacker Newsnew | past | comments | ask | show | jobs | submit | blacktechnology's commentslogin

A few days ago, someone posted on Reddit about “the guy who consumed $50,000 worth of Claude Code tokens on a $200 subscription plan.” That post ended up being about me. https://www.reddit.com/r/Anthropic/comments/1mqs5rf/this_guy...

Here’s the context: Claude Code currently offers a $200/month plan with very generous usage, including advanced features like Ultrathink, Opus, and (as of v1.0.71) background commands that let it run 24/7. I’ve been building real products on top of it, using these features exactly as intended. By leaning heavily on Claude Code (with some help from Cursor), I ended up being the top token consumer globally.

Some commenters on Reddit saw this as “abuse.” Personally, I don’t. Everything I did was within the rules Anthropic set. To me, it shows what’s possible when you treat Claude Code not just as a coding assistant, but as an actual development engine.

Here are the products I’ve built so far: 1.Raphael AI https://raphael.app – AI image generation with no login, no daily cap, and unlimited free usage. Paying users can unlock higher-fidelity models. 2.AnyVoice https://anyvoice.net – Ultra-fast voice cloning from a 3-second sample, producing voices that are nearly indistinguishable from the original. 3.Fast3D – https://fast3d.io A platform that turns text or images into 3D models in seconds, designed for speed and simplicity.

All three products are already profitable and sustaining a small company, and we have two more major launches planned before September.

The Reddit thread raised an interesting debate: - Where is the line between “power use” and “abuse” when a SaaS product offers generous terms? - Should companies celebrate heavy users who push their tools to the limit, or restrict them? - And more broadly, what new kinds of businesses become possible when development itself can be automated at this scale?

Curious to hear the HN community’s take.


I don't think it's abuse. I actually welcome the insight:

The only way the sleight of hand can be sustained is by burning ever increasing amounts of resources.

People are convinced they are seeing some "emergent" property. We are all just entranced by paraidolia.

It isn't very impressive to see a corporate data simulator that is fed all the data and runs on giant GPU farms. There is no gain, only a massive loss. They just hope to hook everyone on it before the investment money runs out and they just might.


Why do you care about what users on reddit think?


is it an official site? https://qwen-image.ai


Just go to chat.qwen.ai and click on "Image Generation" below the text input.


thank you


fixed


No, still doesn't work. Progress bar now stuck at 95% for dozens of seconds after initial progress to 95% takes only 5 seconds. Broken.


Still getting the same error I was (and the same error mentioned by parent commenter) when this was first posted.


Hey everyone! I recently created an AI voice cloning product that can clone any voice with just 3 seconds of original audio! I'd like to share it with you all. Here's how it started: A friend of mine always sends me voice messages, and I wanted to play a little prank on her - what if I could reply using her own voice? With AI technology being so advanced now, I made it happen, haha, and it's been so much fun. Try it out at: https://anyvoice.net/ai-voice-cloning What Makes AnyVoice Special?

Only needs 3 seconds of audio to clone a voice! Yes, it's that simple. Just record something like "The weather is beautiful today, let's go get dinner" on the webpage - say anything you want. Other solutions out there require dozens of sentences to record, who has time for that... Supports four languages - English, Chinese, Japanese, and Korean: In testing, the results are incredibly natural. When my friend heard her own voice, she was so shocked she thought she was getting a scam call!

Real Usage Scenarios:

Prank Your Friends: Reply to their voice messages using their own voice. Just record their voice (most voice messages are longer than 3 seconds, right?), then record it on the website, and it's instantly cloned.

Pro tip from my experience: Use two phones for easier operation. Phone A plays your friend's voice, Phone B opens https://anyvoice.net/ai-voice-cloning to record and clone the voice. After cloning, Phone B plays the cloned voice while Phone A records a voice message - your friend will be totally shocked!

Short Video Voiceovers: Content creators know the pain - you either need to find someone or do the voiceover yourself, which is tiring. If you want to use voices from TikTok or other platforms, they often charge fees. We're completely free - you can get any voice you want! Language Learning: Having your own voice read foreign language texts feels amazing and really boosts learning motivation. Pretend to Speak Foreign Languages: Record yourself in English, clone your voice, then input Chinese, Japanese, or Korean text - you'll get your voice speaking these languages!

How to Use: Super simple interface, just three steps:

Upload/record a voice sample (3-10 seconds) Choose the text you want to generate, up to 200 characters (it's a free service, please understand - we might increase this limit in a paid version in the future) Wait for generation (usually takes just seconds)

Try it here: https://anyvoice.net/ai-voice-cloning Finally: We're currently in public beta, completely free with no paid features! Welcome all to try it out and give feedback. If you find it useful, please give it a like Feel free to leave any suggestions or questions in the comments below, I'll respond to each one.


This is quite similar to https://screensharing.net. Works like magic. Thanks for sharing!


Is there any plan to add image/screenshot support to Windsurf?


For example, click here to have a try: https://domainsnap.net/?q=hackernews


live demo here: https://chat.vercel.ai/


I really love this game! Is there any plan to make it open source?


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: