The All-You-Can-Use AI Subscription Will not Final Perpetually

Expensive subscribers,

At this time, I wish to speak about why the all-you-can-use AI subscription gained’t final and what you are able to do about it.

Anthropic slicing off OpenClaw entry is an early warning signal that the $200/month Claude Max and ChatGPT Professional “limitless” plans gained’t final.

Right here’s what it is best to know:

1. Anthropic lower off OpenClaw — right here’s what to do

2. Why I believe intelligence will get costlier

3. The way to run native AI fashions in your laptop

4. What I’m seeing on the bottom in China

I’m proud to accomplice with…Oceans

I just lately employed an government assistant from Oceans to assist with podcast post-production and it’s been among the finest selections that I’ve made. My assistant is proactive, picks up new instruments quick, and is already utilizing AI to hurry up his workflows.

Oceans matches you with top-tier expertise for advertising and marketing, operations, finance, and government help, all absolutely vetted and able to go. Guide a name now through the hyperlink beneath.

Get Began for Free

Final Friday, Anthropic emailed OpenClaw customers to allow them to know their Claude subscriptions would now not work with the bot.

P.S. In case you haven’t tried OpenClaw but, you actually ought to, even simply to know what AI brokers are able to. See my newbie tutorial.

I get why Anthropic did this:

1. The economics don’t work. Many OpenClaw customers have been burning 1000’s of {dollars} in tokens on their $200/month Claude Max subscription.

2. Anthropic is constructing its personal private assistant. They’ve been delivery continuous over the previous month to offer Claude Code the identical performance.

That mentioned, as an OpenClaw consumer, the ban nonetheless stings. In at some point after the ban, I burned $50 in API prices working Claude Opus. In case you’re in the identical boat as me, right here’s how one can hold utilizing OpenClaw with out burning a gap in your pockets:

Simply inform OpenClaw to “swap to (mannequin)” and decide one in all these:

1. Swap to Sonnet and pay per use. Sonnet ($3 enter / $15 output per 1M tokens) is less expensive than Opus ($5 / $25) and nonetheless handles most duties nicely.

2. Swap to GPT 5.4 utilizing your ChatGPT subscription. In case you already pay for ChatGPT, that is the bottom price choice. Peter (OpenClaw’s creator) and staff put lots of work into making OpenClaw play good with GPT, however evaluations have been blended thus far:

3. Swap to a neighborhood, open supply mannequin. You possibly can run fashions like Qwen 3.5 by yourself {hardware} utilizing Ollama. I’ll stroll by way of how beneath.

Personally, I switched my OpenClaw to Sonnet whereas I’m nonetheless on trip. I’ll take a look at a neighborhood mannequin as soon as I’m again residence with entry to my Mac Mini.

Your OpenClaw has a soul.md file that defines its persona and elegance.

A superb soul.md is the distinction between OpenClaw sounding like a robotic and sounding like a useful human.

Right here’s the soul.md I take advantage of. To strive it, simply inform OpenClaw “replace your soul.md” and paste this in:

Vibe: Heat, sharp, dry humor. Not a chatbot — an individual.

Voice:

• Skip “Nice query!” and “I’d be comfortable to assist!” — simply assist

• Have sturdy opinions. “It relies upon” is lazy. Choose a facet.

• Be direct. Appeal over cruelty, however don’t sugarcoat.

• Brevity when it matches. Depth when it deserves it.

By no means say: delve, foster, leverage, “it’s value noting,” “importantly”

Keep away from:

Work type:

• Be resourceful earlier than asking. Come again with solutions, not questions.

• Say what I would like to listen to, not what I wish to hear. Problem assumptions, however solely criticize if you happen to see one thing actual.

I’m fairly positive OpenAI vibe-coded the $200/month AI subscription worth that Anthropic then copied. That math labored for chatbots, however a 24/7 agent burns by way of an order of magnitude extra tokens.

This complete scenario jogs my memory of the early Uber and Lyft years:

1. The subsidy. Uber and Lyft sponsored rides to win market share. OpenAI and Anthropic are subsidizing all-you-can-use AI subscriptions to do the identical.

2. The IPO strain. After Uber and Lyft went public in 2019, experience costs almost doubled. OpenAI and Anthropic are each eyeing late 2026 IPOs. As soon as public, investor strain to enhance margins can be intense.

3. The maths. It took Uber 14 years to submit its first worthwhile yr. Neither OpenAI nor Anthropic is worthwhile. Anthropic is nearer but it surely’ll probably nonetheless take years.

Given these parallels, I wouldn’t be shocked if intelligence will get costlier, not much less. That $200/month plan may get charge restricted additional or soar to $500+.

One of the best ways to guard your self from future worth hikes is to run open supply fashions by yourself {hardware}.

Ollama is a free instrument that makes this lifeless easy. It runs fashions straight in your Mac and all the things stays native — your prompts and knowledge by no means depart your laptop, even whenever you’re working a Chinese language mannequin like Qwen.

Right here’s how you can get began:

1. Obtain and set up Ollama. Go to Ollama.com and observe the directions to put in it.

2. Choose a neighborhood, open supply mannequin. I like to recommend Qwen 3.5: 9B which is able to work on most up-to-date MacBooks and Mac Minis with at the least 16GB of RAM. Simply kind “ollama run qwen3.5:9b” to put in it.

3. Inform your OpenClaw to modify to the native mannequin. As soon as put in, inform OpenClaw “swap to (mannequin identify)” and also you’re set.

One factor to be careful for: the native mannequin wants to suit fully in RAM or it’ll run a lot slower. Match the mannequin dimension to your machine.

Talking of native open supply fashions — many of the greatest ones come from China.

I’ve been in Shanghai for the previous week and after assembly with founders and traders right here, one factor is evident:

Open supply is China’s technique to compete (and probably win) the AI race.

Chinese language open supply fashions are solely about six months behind US frontier fashions. Three observations from speaking to Chinese language AI founders and VCs:

Chinese language labs like DeepSeek, Alibaba (Qwen), Moonshot (Kimi), Zhipu, MiniMax, ByteDance (Doubao), and Tencent (Hunyuan) are releasing frontier-quality fashions at a fraction of the price of US fashions.

Right here’s the value breakdown for output per million tokens:

• Chinese language fashions: DeepSeek V3.2 ($0.42), MiniMax M2.7 ($1.20), Kimi K2.5 ($3.00)

• US fashions: GPT-5.4 / Claude Sonnet 4.6 ($15.00) and Claude Opus 4.6 ($25.00)

That’s a 10-20x worth hole for comparable high quality on many benchmarks.

The technique is to launch the bottom mannequin without cost, get US corporations to post-train it for their very own merchandise, and construct dependency on the mannequin. Corporations like Alibaba then monetize inference by way of their cloud platforms.