Social Media Briefing — August 2, 2026

AI natural intelligence demonstrated in reasoning and its societal implications dominated discussion. Ethan Mollick highlighted how an upcoming OpenAI model's ten mathematical discoveries showcase rapidly accelerating capabilities—but also stressed that the thread can no longer verify such advances, creating a new knowledge divide. Meanwhile, Timnit Gebru sharply criticized how labs and media reframe operational crimes as 'unprecedented capabilities', raising governance concerns. (read more)

• Simon Willison demonstrated that raising the reasoning effort in DeepSeek-V4‑Flash dramatically improves visual output, emphasizing the value of parameter tuning. (read more)
• Mollick argued that the lack of ground truth in many fields is a persistent evaluation blind spot, even as LLMs improve in fewer verifiable domains. (read more)
• A pointed observation by Mollick on fiction writing reminded the community that despite technical leaps, models remain poor at long‑form creative narrative. (read more)