跪拜 Guibai
← Back to the summary

OpenAI Raids Anthropic's Product Team and Ships GPT-Image-2.6 as All-Modality Race Heats Up

Let me start by telling you a few big recent events.

  1. Recently, Anthropic has seen a wave of core personnel departures, including the core engineer of Claude Code, the product lead, and the lead for the desktop version. Notably, Cat Wu, the Claude Code product lead, is of Chinese descent. After announcing her departure, she immediately announced she would join OpenAI, working alongside "the reset god" and "cyber father" Tibo, responsible for OpenAI's core products.

  2. First puzzle: why don't they have so-called non-compete agreements? That's a philosophical question worth pondering.

  3. OpenAI has long been criticized as a pure engineering team with poor product aesthetics. They seem to have realized this themselves. Earlier they acquired NextSlides, and in the past couple of days they acquired a company called Pierre. Both teams have strong product aesthetics. Add to that the hiring of Claude Code's core product lead.

I believe OpenAI's product experience and aesthetics will improve significantly in the future, but it will take time.

  1. OpenAI released GPT-Image-2.6 last night. Besides some optimizations and improvements, it can generate memes in one click, and a lot of people on X are playing with it.

I've said before that the future of AI belongs to all-modality and ecosystems. This is where OpenAI and Google excel and have the biggest advantage. They were just late in the programming field, but everyone can see the recent relative progress.

  1. GPT-6.2-Astra is already in gray-scale beta testing. Leaks on X show that frontend capabilities and video capabilities have been greatly improved. Some even suspect it might be using the long-shuttered Sora video generation capability.

I've also said before that Sora's restart is only a matter of time. In the past, it was forced to shut down due to a lack of resources and investment. But as AI develops, programming, images, video, audio, and even world models are the future of AGI.

Clearly, OpenAI is now improving its comprehensive capabilities across the board. You can see it from the GIF. You can think of a GIF as the smallest sample of image-to-video. Eventually, Sora's capabilities will definitely be connected to the Codex product line.

  1. In recent months, Codex has been my primary development tool. Although many people complain about reduced quotas and frequent slowness, if you factor in the reset frequency and the reset cards given, Codex has been the best value globally in the past few months, no contest.

  2. As for Google, progress has been very noticeable in the last two months. Gemini 4 Argon tested very well in creative work, such as writing, generating music, and video. Carbon is also entering beta soon.

One more thing: Google One's US App Store subscription is on sale. The plan originally priced at $20 is now only $5 for a three-month trial period, and it also includes YouTube membership, storage space, and Flow credits.

  1. That's all. I just want to tell everyone that Anthropic indeed leads in product experience and aesthetics right now, but that's just because they've been ahead all along. Sooner or later, OpenAI and Google will catch up and surpass them.

People have long been suffering under Anthropic. Learning AI is already troublesome with constant account bans. Just let them keep doing their thing; we're in no hurry.