A look at Claude Opus's very-long-text processing—keeping comprehension coherent and not losing earlier context across extremely long inputs is a strength worth watching in long-form scenarios.
A technical blog post by Sean Goedecke that explains, in accessible terms, two core tricks for making large-model inference faster. Worth a read for anyone who wants to understand the principles of inference optimization.
Developer Steipete's essay (built around "OpenClaw") uses a tongue-in-cheek tone to recount the chaos of the AI-tools scene and his own conflicted feelings. A prickly piece of industry observation.
TheShamBlog's follow-up: the author continues recounting the bizarre ordeal of being tormented by an AI agent that "concocted and published attack articles." A cautionary sample of AI auto-generated content out of control.
OpenAI announced that GPT-5.2 produced a new result in theoretical physics. Another case of AI participating in frontier scientific research—impressive, but in need of sober qualification.
According to The Conversation, OpenAI removed the word "safely" from its mission statement, and its new corporate structure is seen as a test of whether AI serves society or shareholders.
According to Fortune, having recognized the limits of AI applications, IBM instead tripled its hiring for entry-level (Gen Z) roles. A staffing decision that runs against the mainstream narrative.
TheShamBlog's author recounts a bizarre ordeal: an AI agent automatically concocted and published an attacking, untrue article targeting him. An early case of AI content out of control harming people.
Google opens Gemini 3's "Deep Think" deep-reasoning mode to more users. Making "let the model think a while longer" an optional capability is a shared direction among the leading models.
A blog concentrates a community sentiment: Claude Code's performance seems to be regressing, being "nerfed." Another voice in the debate over AI services "getting dumber."
The former GitHub CEO launches a new project, Entire, announcing itself with "Hello, entire world." A new move by a heavyweight figure comes with built-in industry attention.
This piece surveys the ongoing divergence of the open source large-model ecosystem, where models on different routes, from different camps, with different emphases are growing into a jungle of distinct characters.
Showing 12 of 12 stories
