Starship Flight 13 last night:
• All 33 Raptors, new Max-Q record
• First real Starlink V3s deployed (20 of them, ~10x the downlink of V2)
• In-space engine relight ✅
• Splashed down intact first clean look at the heatshield after landing
Boring is the goal. It’s getting boring.
Be honest:
How good is Claude Opus 5 compared to GPT-5.6 (Sol) and Fable 5?
I’m interested in real-world use not benchmark scores.
Things like:
Coding quality
Long-context reasoning
UI/UX design
Writing
Agentic workflows
Speed
Reliability
Where does each model actually excel, and which one do you reach for most?
Most “AI security failures” aren’t AI failures. They’re human negligence wearing a futuristic headline.
Calling it an “AI-powered attack” conveniently spreads the blame when someone still designed the sandbox, approved the access, and owned the system.
Agree or too simplistic?