The CEO of Anthropic called for an AI slowdown. I kept building anyway.

Dario Amodei said this week that frontier AI development might be moving too fast.

My response: I shipped a new language pair in Genie 007.

Here is my honest take on why these conversations frustrate me.

There is a gap nobody mentions. Frontier AI (training trillion-parameter models, safety research at the edge of capability) is a completely different activity from applied AI (building products that use those models to solve real problems).

A call to slow frontier research has nothing to do with whether you should ship your feature this week. These are not the same thing.

The companies calling loudest for slowdowns tend to be the ones who already trained the frontier models. A slowdown now is a moat later. I am not saying that is bad faith, just that it is how competitive markets work.

What I actually care about:

I am building Genie 007 (genie007.com), a voice-to-action tool. You say something rough, it delivers the finished action. Works everywhere: Windows, Mac, Android, iOS, Chrome extension. No frontier training. No alignment research. I am using the APIs these companies ship to help people do their work faster.

If anything, a genuine slowdown in capability would hurt tools like mine more than the frontier labs. We depend on the APIs improving.

This week: added support for 8 more language pairs in real-time translation. We are at 147 now. The users who care about that update are not reading AI safety papers. They are trying to communicate clearly in a second language on a client call tomorrow morning.

That is the actual implementation gap. The distance between an AI headline and the person it was supposed to help.

Does the slowdown framing feel like a genuine safety concern to you, or a competitive strategy? I have my view, but I am curious what people building in this space actually think.