Using OpenClaw for a few days, I am on the brink of bankruptcy.

Direct feeling: On the brink of bankruptcy.

1. Employees really have their worth

At first, I used Claude Sonnet 4.5, and I was spending money (Tokens) like water, a dollar for a sentence, and I ran out of Tokens in no time.

It's hard to go from luxury to frugality; later, I switched to a cheaper model and felt the intelligence was too low to accept (for example, Claude 3.5).

You get what you pay for; if you pay ten cents, you get three cents' worth, making it hard to choose.

2. Subscription mentality vs Pay-per-use mentality

Being able to manage life is certainly a good quality for older generations. However, the downside is that when I want to discuss certain issues, I always think of first looking for the subscription-based Gemini and GPT, rather than OpenClaw, since the latter is pay-per-use.

However, this constant switching really disrupts the experience, U1S1. I feel a bit like the older generation in the past, going out for dinner and having to turn off all the lights and air conditioning at home.

3. Human-like vs AI-like

I feel that one reason people like OpenClaw is that its way of speaking is more human-like. On the other hand, if you haven't trained GPT Gemini, it deliberately behaves like an AI. I guess OpenClaw incorporated a human persona during its setup, which has won over many who don't like to train AI.

4. Domestic products always provide ample supply

One major way to alleviate Token anxiety (people are really doomed now; smartphones cause charging anxiety, electric vehicles cause range anxiety, and playing with AI even brings Token anxiety) is to switch to DeepSeek. Use it freely without worrying about costs.

Domestic products are always about ample supply, but in terms of intelligence, they definitely lag behind Claude Opus 4.5 and 4.6. Moreover, there are some limitations in terms of values.

The price difference is 20 times… can you believe it?

5. Urgently need hybrid models

I urgently need a modified version of OpenClaw that is dual-model, similar to new energy and hybrid vehicles.

For casual chatting, literature searches, and completing daily tasks, use the cheaper model that provides ample supply;
But when I need to write code, switch to the most expensive Claude Opus 4.5/4.6 automatically, giving me truly usable code, not just toys.

Use a chicken knife for chicken, and a beef knife for beef.

6. Contextual cleanliness should be used with caution

I have a contextual cleanliness obsession. Ever since I learned that every additional line of context for AI exponentially consumes computing power, I particularly dislike completing all tasks in one dialogue box. If the tasks are unrelated, I will not hesitate to open a new dialogue box.

However, OpenClaw seems to not give me the chance to reopen the dialogue box. Watching my dear assistant think with 100K/200K of context every time makes me feel really uncomfortable.

Because this context includes all my casual chats, failed tasks, and inquiries about OpenClaw operation commands, all this messy context gets mixed with my Vibe Coding requirements… especially when using the expensive Opus 4.5 API, my heart bleeds.

7. Everything requires an API

Everything requires an API; even searches need an API. The brave search AI I bought is not enough for free, and I have to pay. For Bitcoin prices, I also need the Coingecko API. In short, various large purchases and expenses.

8. Security issues

Security issues are not as serious as I imagined when I was a cloud player, especially regarding injection attacks, which I was a bit overly concerned about. It has its own security protocols. However, I still wouldn't store important assets and passwords in it.

Finally

When I said I was on the verge of bankruptcy at the beginning, it was a joke; it's not that expensive. But to be honest, every time I generate a dialogue, it's actually a "micro-investment" of your money and time.

AI has a saying that goes well—every time you burn Tokens and time doing Vibe Coding, you are playing a slot machine, betting on it to spit out a perfect piece of code, hoping it’s not a pile of garbage that can never be fixed.

Wishing everyone good luck in their "gambling."

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。