As usual, big words with nothing to back them up

I enjoy watching the hype balloon get inflated, because the collision with reality always looks the same — nothing left but a bad smell and a bitter taste.
It’s painful to watch you squander all the good PR you’ve built around yourselves. The situation reminds me of what CD Projekt went through with Cyberpunk 😄
Here are my main concerns:
1. Tests taken from the GCC repo — this isn’t independent verification, it’s a self-fulfilling prophecy.
2. Model trained on GCC — so calling this “from scratch” work is misleading at best. This isn’t a compiler written from the ground up, it’s reprocessed knowledge from decades of GCC community effort.
3. Lack of transparency — it looks to me like the validation tests were used without being published in the repo. I’d be very curious to see how much GCC code would actually show up in your compiler if someone did a proper audit.
The overall message I’m getting: you spat in the face of the entire community that worked hard on GCC for years, then announced to the world that the same work can be done with Claude “from scratch.” Complete bullshit.
But to be fair — there’s no question you have a very good model, and the agent team concept is an interesting solution and a step in the right direction. It’s just a shame you can’t do PR properly, and instead of honestly showcasing what the tool can do, you chose a narrative that undermines trust.