@YxzRainy: Rumor has it that Claude's parameters have already stacked up to 25T. Is Anthropic planning to keep scaling all the way to 100T? The bigger the parameters get, the higher the cost, while the performance gains keep diminishing. Can you really scale your way to AGI just by stacking parameters?
Summary
Rumors claim that Claude's parameter count has reached 25 trillion, sparking discussion about whether Anthropic will continue scaling to 100T, and doubts about whether simply scaling up parameters can lead to AGI.
View Cached Full Text
Cached at: 10/21/26, 12:29 PM
Rumor has it that Claude’s parameters have been stacked up to 25T 😱😱😱
Is Anthropic planning to keep scaling all the way to 100T?
The more parameters you pile on, the higher the cost gets — yet the performance gains keep shrinking.
Can brute-forcing parameters alone really get us to AGI? 🤔
Similar Articles
@FinanceYF5: Anthropic internal data shows Claude is accelerating AI development — potentially a path to recursive self-improvement, or a path where AI autonomously builds a more powerful successor. This process is unfolding faster than expected, and its implications deserve more attention.
Anthropic internal data shows Claude is accelerating AI development, potentially leading to a path of recursive self-improvement. The process of AI autonomously building a more powerful successor is happening faster than anticipated.
@huxlab: I met a friend who does R&D at Anthropic, with an annual salary of 1.5 million bucks. I asked her how she tuned Claude …
一名用户分享了据称来自 Anthropic 内部的 Claude 调优技巧，总结出 5 条提示词工程核心原则，包括用 XML 隔离语义块、清理旧模型的防御性约束、用工具调用处理复杂计算、提供双向权衡以及用生成-评估-修复闭环替代巨型提示词。
@xiaohu: Anthropic launches Claude Science, an AI workbench for scientists with over 60 research skills built in. It is an application installed on your own computer or server: you ask an AI scientific questions in plain language, and it mobilizes dozens of specialized tools to query data, run analyses, draw charts, and draft manuscripts…
Anthropic has launched Claude Science, an AI workbench for scientists with over 60 built-in research skills. It supports local deployment and HPC clusters, and can autonomously draft computing tasks and review results.
What’s at the center of Claude’s mind?
Anthropic's research has discovered a structure inside the Claude model called 'J-space,' similar to human working memory, used for internal reasoning and thought. This finding can help monitor potential misbehavior in models and deepen understanding of AI's internal mechanisms.
@FinanceYF5: Which Parts Anthropic Refuses to Outsource. Anthropic is the company behind the AI assistant Claude, and Claude Code is its coding tool. This year it built more than a dozen components for its agent in-house, even rewriting its code repository from scratch. There was only one thing it willingly handed off to others: the sandbox for running code...
This article breaks down the dozen-plus key components Anthropic built in-house for its AI agent, and identifies its outsourcing principle: components that age alongside model updates are built in-house; sandboxes that merely provide compute can be outsourced.