I help devs use AI tools to ship faster. Full-time iOS engineer & indie builder · SwiftUI enthusiast.
Istanbul
Joined May 2020
- Tweets1.4K
- Following137
- Followers43
- Likes311
I hope that Anthropic and OpenAI will go bankrupt because what is happening is not honest. They give us the REAL inteligence and than for economy they nerf models and cut usage limits. This is not serious for any type of real job.
Looks like Claude links parsing is totally broken on other languages than english, <cite…> everywhere in text @trq212
NIKITA STUPIN retweeted
they want us to hate them
We’ve written an FAQ to answer some of the questions we've received about watermarking.
In summary:
• We’re implementing watermarking to comply with the EU AI Act. Other major model developers have signed the same Code of Practice and will also be implementing watermarking;
• Our watermarking method doesn’t have any practical impact on the quality or content of Claude’s outputs;
• The difference between watermarked and un-watermarked text will not be distinguishable to readers;
• Nothing is added to the text and there are no hidden characters;
• Watermarking doesn’t require extra tokens, and will not be more expensive;
• Watermarks can’t be traced to a specific person, organization, or chat.
Read more: anthropic.com/news/claude-te…
NIKITA STUPIN retweeted
Don’t fall for the EU AI Act as Anthropic’s reason for watermarking their texts.
I don’t support the EU AI Act.
But in this case you only need to look at Amodei’s constant whining about Chinese labs training on Anthropic model distillation, his push to ban open-source models, and how he keeps begging for government regulation.
They may claim they did this to comply with the EU AI Act - it’s a convenient public excuse.
Personally I think it’s about slapping a watermark on people’s knowledge and claiming it as theirs.
Very Amodei.
We’ve written an FAQ to answer some of the questions we've received about watermarking.
In summary:
• We’re implementing watermarking to comply with the EU AI Act. Other major model developers have signed the same Code of Practice and will also be implementing watermarking;
• Our watermarking method doesn’t have any practical impact on the quality or content of Claude’s outputs;
• The difference between watermarked and un-watermarked text will not be distinguishable to readers;
• Nothing is added to the text and there are no hidden characters;
• Watermarking doesn’t require extra tokens, and will not be more expensive;
• Watermarks can’t be traced to a specific person, organization, or chat.
Read more: anthropic.com/news/claude-te…
NIKITA STUPIN retweeted
If the “watermark” is embedded in the statistical sequencing of text ie it can’t be stripped out without rewriting the text - doesn’t that necessarily make LLM text generation less useful?
The model is no longer performing purely for the user’s request. Some fraction of its freedom in choosing words, phrasing, and structure is now being spent on preserving the watermark.
You get an inherent tradeoff between watermark robustness and model flexibility/compliance. Ergo, the stronger & more robust the watermark, the greater the constraint.
In other words: LLMs will suck more and be harder to work with.
NIKITA STUPIN retweeted
Opus 5 is so random...
sometime it feels like shit
but sometimes it is truly great
even finding things that GPT 5.6 Sol Max missed
very weird model...
ChatGPT is unusable, I have 2 accounts with different chat history and I can't switch. Forced to use claude for now
NIKITA STUPIN retweeted
anthropic’s biggest aura moment was dropping opus 4.5
for a while, nobody cared about benchmarks anymore
Idk but Codex 5.6 is the most inefficient model currently. Hundreds of millions tokens for same tasks that before cost me less than a hundred
