llm Tag Posts Index

考虑设计个Hive Power Voter 智能量化点赞系统

昨天忽然想到，每天手动点赞其实是一件挺耗时间的事情。既然完全可以通过程序来完成，那为什么还要一直重复人工操作呢？ 不过，如果只是做一个傻瓜式的自动点赞，我觉得意义并不大，甚至还不如自己手动点。我的想法是，在多种策略的基础上，再引入 AI 智能决策，让整个点赞过程既能提高效率，又尽可能贴近真实的人工行为，这样才更有价值。

越来越先进，也越来越黑箱的大模型

早上一打开 Claude Code，发现左下角弹出了一个更新提醒，于是点了一下，重启后出现了下面这个界面。 标题的意思很明确： Fable 5 现在可以用了！ 再看下面的介绍，大意是说，在 7 月 7 日之前，可以体验 Fable 5，但最多只能使用套餐每周额度的 50%。如果超过了这个额度，就需要另外购买 Usage

No todo es ChatGPT: los mejores LLMs chinos en 2026 y para qué sirve cada uno 🇪🇸 ထ 🇺🇸 Not Just ChatGPT: The Best Chinese LLMs in 2026 and What Each One Is Good For

En conversaciones sobre inteligencia artificial, suele mencionarse a ChatGPT de forma automática, como si no existiesen otros LLMs. Y no es que ChatGPT sea malo,

Fable 5 and Mythos 5: the real reason emerged — and it's not what was in the script

Until now, the story revolved around a single word: jailbreak. A narrow, contested gap that the company swore was small. That framing ended this weekend. What surfaced

Fable 5 and Mythos 5: the crisis now has an address — and it's in Seoul

With every update to this case, it changes size. It started as a narrow technical dispute, became a confrontation between Silicon Valley and the US government, and

Fable 5 and Mythos 5: the negotiation just got more complicated

After the initial shock, the case has moved past the panic phase and into the negotiating-table phase — and, by Washington standards, it's moving fast. Except that,

Fable 5 and Mythos 5: where the negotiation stands now

After the initial shock, the case has moved past the panic phase and into the negotiating-table phase — and, by Washington standards, it's moving fast. The main

Three days of life: what the Fable 5 and Mythos 5 case tells us about the future of AI

Some weeks the tech industry moves at its usual pace. And some weeks an AI model is born on Tuesday and shut down on Friday. Anthropic's was the second kind. On

Megatrends and the use of artificial intelligence

Artificial intelligence has stopped being science fiction to become critical infrastructure. We are no longer discussing whether AI will transform business — it

Thinking budgets: the control developers needed

In 2025, the big news was extended reasoning — models that think before answering, exploring multiple paths before reaching a conclusion. In 2026, the news is subtler

AI in 2026: US$ 300 billion and the end of the hype

The artificial intelligence market is expected to surpass US$ 300 billion in 2026. It is a striking number — but what it represents in practice is more interesting

Mixture of experts: the architecture redefining efficiency in LLMs

When Z.ai's GLM-5.1 took first place on SWE-Bench Pro in April 2026 with 744 billion total parameters but only 40 billion active per inference, many people looked

Agentic AI: when the model stops answering and starts working

There is a fundamental difference between a model that answers questions and a model that executes tasks. That difference has a name: agentic AI. And in 2026, it

Real multimodality: who processes audio and video natively in 2026

"Multimodal" has become a marketing term. Almost every model released in 2026 describes itself as multimodal. But there is a fundamental technical difference

Small models and fine-tuning in 2026: the advantage of the specialized ones

There is a dominant narrative in LLMs: bigger is better. The headlines are about trillion-parameter models, 10-million-token context windows, frontier benchmarks.

Qwen 3.5 and 3.6: Alibaba and the ambition to cover every use case

Alibaba's strategy with the Qwen series is different from that of the other manufacturers. While Meta releases three models, DeepSeek releases two, and Mistral releases

Open source reached the frontier: what changed in 2026

For years, the narrative was the same: open source models lag 6 to 12 months behind proprietary ones. They were good for prototyping, adequate for simple use cases,

GPT-5.5 and Claude Opus 4.7: the new bar for autonomous models

In March 2026, GPT-5.4 and Claude Opus 4.6 defined the state of the art. In April, both companies released updates that were not fine-tuning increments — they were

Llama 4: Meta redefines open source with native MoE and multimodality

Meta has released relevant open models before — Llama 2 and Llama 3 had real impact on the community. But Llama 4, released in April 2026, represents a shift in

DeepSeek V4: the open source model that reached the level of closed ones

For a long time, the divide was clear: proprietary models like GPT and Claude delivered the best performance, while open source ones stayed a step or two behind.