Franco

Attending the PM x AI Product Leadership Conference in Hangzhou, an intuitive feeling is that everyone knows the AI trend is unstoppable, but everyone is looking for where the opportunities really are.
Currently, there are a few relatively certain areas: one is organizational restructuring, another is product reshaping or new development, and also product marketing approaches.
I don't know how fast the changes are happening, but it seems to be accelerating, so there are all kinds of opportunities.

Some time ago, I switched my Chinese language processing model to DeepSeek. After using it for a while, I found that it is not verbose, works efficiently, and most importantly, it is very affordable.
Why did this change happen? My wife said that DeepSeek's content is of higher quality than Doubao and Qianwen. I paid special attention to it afterward and even asked GPT about it. Indeed, in terms of Chinese corpus and context understanding, DeepSeek's capabilities are very strong.
After using it for a period, I really think it’s quite good.
So why are there very few articles about DeepSeek

Many people say they can't use Claude Code and often get banned.
Actually, the core key point here is that everyone is thinking about using Max20x, and they feel they need to maximize traffic and benefits.
But in fact, I personally think that the current Pro version and Max5X version are actually sufficient for many non-heavy users.
The official risk control for these two products is not very strict.
So I suggest everyone consider this approach more.
All tutorials online are focused on preventing bans for Max20x, but little do they realize that using Pro and Max5x involves much less

I finally started using the native Claude code, which has three modes: chat, cowork, and CLI. I first purchased the Pro membership, and once things stabilize, I’ll upgrade to Max.
Currently, major model providers all lack computing power, which is a significant bottleneck and also a huge market opportunity. Just like when electricity was first invented—power was limited, but after technological upgrades, breakthroughs gradually happened.
When will AI models become freely accessible?

Claude Web now supports Managed Agents.
Official documentation:
Claude Cowork is now available to all paid users.
ChatGPT Pro $100 USD subscription is now launched:
- 5 times the Codex usage of Plus subscription users
- Supports all Pro features
Until May 31, a limited-time offer of 10 times the Codex usage of Plus subscription users.
Pro $200 USD subscription users will also receive doubled Codex quotas before May 31, and their current quota has been reset.
GPT is very generous, resetting quotas at will—how many times has this happened now?

Feeling great! Today, I integrated Codex into Claude Code, making it a more excellent and cost-effective top-tier Agent.
One method is the MCP invocation approach, where it functions as a plugin within Claude Code programming tasks. When executing or checking code, it can automatically call Codex to perform, improve, review, and optimize. Additionally, it can synchronize with CC for checks and provide comparisons, making it easier for developers to choose the best solution.
The second method is to encapsulate it as a skill, where specific tasks can explicitly invoke Codex's capabilities wh

Is there any expert who can clearly explain the differences between OpenClaw and Claude Code, such as application scenarios, model capabilities, and user experience?

Large models are still competing over parameters, but what truly begins to hit the industry ceiling is another issue: context storage.
As inference length, agent call chains, and long-term memory all grow longer, the real factors that determine experience and cost are not just computing power, but whether "context can be read, written, scheduled, and reused efficiently."
This is also why the market has recently started shifting focus toward infrastructure such as Context Memory, KV Cache, and layered inference storage.
In the next phase of AI competition, it may not be about who can gene

OpenRouter data shows that Qwen3.6-Plus was released only 1 day ago, and its daily call volume has already exceeded 1.4 trillion tokens.
Recently, the usage data for domestic large models has become increasingly impressive, and seedance2.0 has also opened public testing. There are two ways to access: one involves appearing characters and requires a prepayment of at least 14k RMB; the other does not involve characters and can be used with an enterprise account.
Looking forward to breakthroughs in domestic models for coding.
#AI #Large Models #OpenRouter #Domestic

Several stages of development using AI:
1. First stage: Discovering a New World, feeling like you’re all-powerful
Thinking that as long as you have an idea, you can fully realize it. Every day, there are countless ideas, plus all kinds of workflows you want to try—full of energy, with endless ideas;
2. Second stage: Starting to feel discouraged, feeling like you’re just going around in circles
Always circling—one skill after another, refining and refining. It feels like it works, and it’s not too bad, but you always feel it’s not good enough. It’s still far away from your original ideas. You k

Starting today, I am trying to use Codex as my main development tool, and I also purchased the ChatGPT Pro version. The overall usage quota is very high, and I don't have to worry about account bans or traffic congestion issues.
My previous development work was often intermittent. Although OPUS and Sonnet are very powerful, interruptions in development greatly affect my learning and development rhythm, and they also have a significant impact on the continuity of the project.
Do you have a similar situation?

Which one is more useful when everyone is using voice input methods? Input method, Doubao Input Method, or Lightning Speech? Which one do more people use?

Feishu's CLI is really easy to use.
Currently, API calls are free for up to 1 million times.
This can basically cover all message notifications.
Previously, using automation bots would consume free usage quotas.
Free accounts only have 200 calls per month, which is completely insufficient.
With the CLI, you can develop fully automated solutions using custom-built applications.
The same effect can save at least tens of thousands of dollars annually.

For most people, building skills within their professional field is one of the best practices for leveraging AI.
For content creators, AI can become a huge boost to their creativity, making it easier to quickly realize ideas. For example, writing articles by outlining the framework based on thoughts and pairing them with beautiful images; or creating a Xiaohongshu (Little Red Book) matrix, generating copy based on different roles, and producing marketing soft articles.
The first two examples are tasks that AI can accomplish quickly. For more complex tasks, programming thinking or skills ar