@MixRoute_ai Looks like Hermes is most efficient with AI API keys like on @MixRoute_ai
Always knew OpenRouter is useful, but never knew there are other, better alternatives.
Anthropic is in its own league.
Just days after the launch of Anthropic's Claude Fable 5, the company suspended access to the model after US authorities raised "security concerns."
Anthropic said it was ordered to "suspend foreign nationals" from using Claude Fable 5, and self-described the model as "too powerful."
"The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," Anthropic said.
In other words, Anthropic's latest AI model is so powerful, that the company and US government are now questioning who should even have access to it.
Less than 4 years ago, ChatGPT did not exist. Today, AI is arguably the most powerful technology ever created.
Truly incredible.
The gap between AI slop and a publishable draft is not the model.
It's seven habits.
1. It edits better than it authors. Feed it your ugly draft and say tighten this. From nothing, you get the average of the internet.
2. "Sound less AI" is not an instruction. Name the edits. Shorter words. Less polish. Keep one imperfect sentence.
3. Describe the reader, not the role. "Write this for a skeptical staff engineer" beats "you are an expert writer."
4. It agrees with you by default. Don't ask if it's good. Ask where it falls apart.
5. Give it stakes, not a word count. "This opens a cold email to a CTO" cuts harder than "under 60 words."
6. Ask for the version it didn't show you. The safe answer comes first. The interesting one is usually behind it.
7. The first reply is an opening bid. The real output shows up two pushbacks in. One paste is a rough cut, not a verdict.
None of this is model-specific.
The same moves hold on Claude, GPT, Gemini, and whatever ships next month.
Models rotate. Technique compounds.
What's the move you'd add?