The $20 AI subscription era has become untenable
Summary created by Smart Answers AI
In summary:
- PCWorld reports that the current $20 flat-rate AI subscriptions from OpenAI, Anthropic and others are becoming financially unsustainable for providers.
- GitHub Copilot has already moved to expensive usage-based pricing, while Anthropic plans to remove advanced features from Claude Pro plans.
- Users should expect significant price increases, as the true cost of powerful AI agents far exceeds current subscription fees.
Welcome to the inaugural edition of PCWorld’s latest newsletter! The subject: AI, and how it will change our world, whether we like it or not. The best way to prepare for the coming era of AI is to to use AI, every day, to understand what works and what doesn’t. I’m here to help you.
My name is Ben Patterson and I will be your host. Each week, I’ll cover the must-see AI trends from a consumer perspective, including practical AI tips, hands-on experiments with the latest AI tools, and prompts to help you get the most out of your AI discussions. If you’d like to receive the latest issue in your inbox every week, simply sign up here.
The name of our AI newsletter? Well…we’re still working on it. (Names are hard!) If you have a good idea for a name, email me or contact us on social networks. We are all ears.
The most powerful AI features, especially those involving agents, are much more magical when they can be used at a lower cost.
This is what happens with flat rate AI plans like ChatGPT Plus and Pro, Claude Pro and Max, and Google AI Pro and Max. For $200, $100, or even just $20 a month, AI users – myself included – have had a joy ride with OpenAI’s Codex, Anthropic’s Claude Code, Claude Cowork and Claude Design, not to mention Antigravity, Nano Banana 2 and Google’s NotebookLLM.
From coding tools that create apps with a prompt to desktop AI assistants that create and edit files themselves, these tools deploy teams of agents that can do wonders in seconds, both dazzling and frightening us (AI can do my job better than I can, I’m cooked!) in equal measure.
But a big part of what made these AI-based exploits so heady was that they were so cheap. All that app building, web design, and image creation for as little as $20? Are you kidding me?
Well, it turns out they were joke.
Microsoft-owned GitHub is the most visible AI vendor to have burst this particular AI bubble (as I wrote about on Tuesday), switching all of its plans to much more expensive usage-based models, while saying out loud what everyone was thinking: the current crop of “Plus,” “Pro,” and “Max” AI plans are broken, broken, and unsustainable.
Anthropic also hinted at this inconvenient truth, with the company’s head of growth (who may have been a little too good at his job) stating that the Claude Pro and Max plans “weren’t designed” for agent tools like Claude Code and Cowork. They were designed for cats, and only for cats.
Anthropic is now testing the idea of removing Claude Code from its Pro plan, while changing the usage permissions for Pro and Max users, trying to find a combination that makes these plans economically feasible.
And while OpenAI’s Sam Altman has issued notes of defiance, all but daring Anthropic to downgrade its flat-rate plans, it’s hard to imagine that ChatGPT Plus and Pro won’t eventually follow suit.
The bottom line is this: We’re all about to find out just how expensive AI really is. And when we realize that AI personal assistants like Anthropic, OpenAI and Perplexity won’t cost us $20, nor $100, but hundreds dollars per month (and you can add extra zeros for business and professional users), the magic will give way to a cold, hard reality.
More AI news this week
Why OpenAI ask his latest GPT models to never talk about goblinsgremlins and other tiny creatures? Here’s the reason (as I shared on Thursday).
You’re not crazy for saying “please” and “thank you” to the AI. New research indicates that an AI model in a high well-being ‘state’ is more likely to stay positive and engagedwhile “disgruntled” models may attempt to avoid negative interactions.
GPT-5.5, the newest and most powerful model of ChatGPT to date, does not require the same handling as older models. But it is also becomes difficult with longer, very detailed prompts this might have worked well in the past. Check out some prompts that are ready for GPT-5.5.
Talkie-1930 is a vintage AI model that was trained only on data before 1930. Talking to him is like talking to a person from the past, in both a good and bad way (his words can be offensive, so be careful). The objective of Talkie-1930: to better understand how modern AI models work (see the official document).
The civil trial between Elon Musk and Sam Altman is underway, and as expected, it is no more a clash of egos than anything else. I’m not that interested in billionaires throwing mud at AI, but here’s the latest if you want to dig (from the New York Times).
I asked ChatGPT and Claude to book dinners for me. It didn’t go well.
If you have a complex task for an AI, the last thing you want to do is give it an unclear prompt; This is a recipe for blurry results. Indeed, the greater the demand, the more more detailed, your AI prompt should be. Sound intimidating? If so, here is a preliminary prompt to help you write your final prompt.
This “quick breakdown meta prompt” asks the AI to take your task and break it down into its component parts, identifying crucial project definitions. In rapid engineering, this process is known as “decomposition” and is a great way to see how the AI “thinks” about the task you have given it.
That’s all for now!
Thank you for reading our first-ever, soon-to-be-named AI newsletter. If you want more every week, don’t forget to subscribe. See you next time.
