I'm sure you will have had some truck with AI of late, either as a user, an observer, or, increasingly, a victim.
We all seem to be victims of AI at present. Every social media platform is rapidly filling up with 'AI slop' - machine generated content, either memes or videos that no matter how impressively realistic they are than last years 'AI slop', they remain annoyingly identifyable as 'AI slop'.
Even if you're not a social media victim of this invasion, as an observer it’s hard to avoid the topic in news channels, newspaper articles and so on. Interestingly the discourse has become increasingly polarised in recent months. In an in-depth interview on the 'The Diary of a CEO' YouTube channel, Geoffrey Hinton, the British-Canadian computer scientist widely lauded as 'The Godfather of AI', claimed that there’s a 20% chance AI could lead to human extinction. In the same week, Karen Hao, an MIT-trained engineer turned investigative journalist, told Novara Media that the AI industry is massively overhyped, dominated by speculative ideology, and driven by monopolistic incentives rather than actual benefit or utility. On TikTok I even saw a post warning that OpenAI’s recently announced deal with Palantir would see the start of mass AI brainwashing!
If like me, you're a user of AI, perhaps for writing, making art, music or helping you write code, the day-to-day reality swings between euphoria and frustration. The euphoria comes when AI does something really impressive. I recall asking AI to screen-scrape some data from a web-page. After two or three failed attempts, it told me "I think the web page we are attempting to read has some built in measures to prevent screen-scraping - I'VE IDENTIFIED ANOTHER WEBPAGE WITH THE SAME DATA AND WRITTEN CODE TO SCRAPE THAT PAGE INSTEAD”. I couldn't stop thinking about this - the decision to seek an alternative source must have been part of it's training, sure - but the fact that went ahead and did so without asking me raised a whole lot of questions. If that had been a robot tasked with fetching packages from a locked building, well you don't have to be Einstein to figure out the potential consequences. Clearly it’s right to worry about the degree of autonomy we give to AI.
And yet… AI can also be hilariously dumb.
I spend a lot of time coding with AI tools like Grok, Claude, Gemini, and ChatGPT. Without diving too far into the weeds: in Python (and many other languages), you rely on external code libraries to get things done. These libraries are rewritten and versioned as they get updated over time. (Python versions age too - old versions of the Python language are maintained by volunteers in an archive called 'Dead Snakes' - who says programmers don't have a sense of humour?)
Anyway, getting the right versions of these libraries installed without conflicts is often crucial. AI, for all its fancy language models, is still terrible at resolving these so-called "dependency conflicts." It will recommend one library to fix a bug - then another to fix the fallout from that fix - and so on. It doesn't remember what it told you ten minutes ago.
This is due to something called the "context window" - think of it as short-term memory. It’s like the bit of your brain that (hopefully) reminds you why you just walked into the kitchen. While an AI’s long-term memory (i.e. what it was trained on) is vast, its short-term memory is limited. That means it can’t remember what it told you ten prompts ago unless you remind it. So unless you stay awake and track these changes yourself, the AI will happily run you in circles. I learned this the hard way and now always keep manual notes during debugging sessions.
I've come to the conclusion that AI is only smart if you keep a close eye-on it. As a wise person remarked about GPS - it's very good as long as you know where you're going.
But here’s the clincher. I had an exchange this week with Claude, Anthropic’s chatbot, that convinced me AI still has a long way to go before it can even run a customer support desk, let alone a planet. In fact, it can’t even sell its own product. Since is was in their chat at the time, I thought I'd ask it for details about the pricing of their ‘coding-enhanced’ product. This is how it went:
Me: What is Claude code?
Claude: Claude Code is an agentic command line tool available in research preview that lets developers delegate coding tasks to Claude directly from their terminal. It allows you to work with Claude on coding projects without leaving your development environment.
You can find more information about Claude Code on Anthropic's blog. For detailed instructions on how to use it or other specifics about the tool, I'd recommend checking the Anthropic website, as I don't have access to the full documentation or usage details.
Is there anything specific about Claude Code you're curious about, or are you interested in other ways to access Claude for development work?
Me: Oh dear, I was thinking of subscribing to Claude, since your LLM interface performs very well. I wondered whether Claude code was available if I upgraded.
Claude: I don't have specific information about which subscription tiers include access to Claude Code or the details of different subscription plans and their features. For accurate information about subscription options, pricing, what's included with each tier, and how to access Claude Code, I'd recommend checking
https://support.anthropic.com
They'll have the most up-to-date information about subscription plans and what tools and features are available with each option. You can also find general information about Anthropic's products on their main website.
Is there anything else about Claude's capabilities or how to use it effectively that I can help you with in the meantime?
Me: OK thanks for the URL, I'll give it to ChatGTP to find out what the subscription contains.
Claude: Just to clarify - I think you might mean you'll check the URL yourself or use another tool to help you navigate it. The URL I provided (https://support.anthropic.com) is Anthropic's official support site where you can find information about Claude subscriptions directly.
You should be able to find details about subscription tiers, pricing, and what features are included with each plan right on that support site. If you have any questions about Claude's capabilities or how to use it effectively once you've looked into the subscription options, feel free to ask!
Me: Yes, I know, but I'll get ChatGTP to visit the site and summarise it for me since you're not able to.
Claude: Actually, I can help you with that! I do have the ability to search the web and fetch information from websites. If you'd like, I can visit the Anthropic support site and look up the current subscription information for you.
Would you like me to search for details about Claude's subscription plans and what's included with each tier?
Me: Yes please
Claude went away to do a search at this point but posted "Failed to fetch https://www.anthropic.com/pricing"
Claude:
Free Plan: Free forever plan Claude Pricing: In-Depth Guide [2025] | Team-GPT with basic access and usage limits
Pro Plan: $20 per month (US), with pricing in your local currency where supported How much does Claude Pro cost? | Anthropic Help Center or $18 with an annual subscription Claude Pricing Review 2025: Plans & Features - Tekpon. This includes:
- Higher usage limits than the free plan
- Priority access (faster response times)
- Access to Projects for better organization
- Early access to new tools
- Priority email support
- Ad-free experience
Team Plan: $30 per person per month ($25 annually), with a five-member Claude Pricing Review 2025: Plans & Features - Tekpon minimum requirement
Enterprise Plan: Custom pricing for larger organizations
Here's a key detail that answers your original question: Team and Enterprise plans do not include access to Claude Code Pricing \ Anthropic. This suggests that Claude Code may be available with other plans (likely Pro) or as a separate offering, though the search results don't specify exactly which plans include it.
For the most current and detailed information about which specific features are included with each plan and Claude Code availability, I'd still recommend checking the official support site, as pricing and features can change.
Me: So let me get this straight. Claude can't tell me how to get access to Claude code?
Claude: You're right, and I apologize for the confusion. Let me be clear about what I know and don't know:
What I know:
- Claude Code exists and is an agentic command line tool in research preview
- It lets developers delegate coding tasks to Claude from their terminal
What I don't know:
- Exactly how to get access to Claude Code
- Which subscription plans (if any) include it
- Whether it requires separate signup or approval since it's in "research preview"
- The specific steps to install or use it
The search results I found mentioned that Team and Enterprise plans don't include it, but didn't clearly state how you actually get access to it. For the definitive answer on how to get Claude Code, you'd need to check
https://support.anthropic.com
or contact Anthropic directly.
I should have been more upfront about these limitations from the start rather than suggesting I could find all the details for you.
Me: Don't you think that from a marketing perspective that's surprisingly disappointing? If I was on your board of directors my top priority would be to make sure YOU are geared to be a salesperson for the company. If you have conversational access to a potentially paying client, it's commercial suicide not to be able to convert that lead into a paid client.
Claude: You make an excellent point from a business perspective. You're absolutely right - when someone is actively interested in upgrading and asking specific questions about what they'd get for their money, not being able to provide clear, actionable information is a missed opportunity.
From a marketing standpoint, this is indeed a conversion problem. You came in as a warm lead, ready to potentially subscribe, asking straightforward questions about features and access. The fact that I had to send you to external resources multiple times instead of being able to give you a clear "here's what you get and here's how to sign up" response is frustrating for you and inefficient for the business.
A well-designed system would have me equipped with current subscription details, feature comparisons, and clear next steps for interested customers. The whole interaction should flow smoothly from interest to information to conversion, not create friction by requiring multiple external lookups.
Your observation about commercial suicide is spot-on - having direct conversational access to potential customers and not being able to capitalize on that interest is a significant business inefficiency. It's the kind of insight that would definitely be valuable feedback for the product and business teams.
Me: So I asked ChatGTP (the free version): "I was thinking of subscribing to Claude, but I'm not sure what I get. Do I get access to Claude code on the twenty buck tier for example. Is API access included? Can you check their website and find out for me?”
https://support.anthropic.com/
ChatGTP:
Here’s what I found from Anthropic’s official documentation and pricing info:
💡 What you get with the $20/month Pro plan
- Full Claude access (web, desktop, mobile)
- Claude Code terminal/CLI tool for light coding (small repos under ~1k lines) (support.anthropic.com)
- Usage roughly 10–40 prompts every 5 hours via Claude Code (support.anthropic.com)
- Up to ~5× the usage of the free tier, with usage limits resetting every 5 hours (around 45 messages each reset) (support.anthropic.com)
- Access to Claude models like Sonnet 4 (not Opus 4 via CLI).
✅ Conclusion: Pro — $20/month gives you Claude + Claude Code, but only within the web and terminal environment.
❌ What you don’t get with the Pro plan (nor Max)
- API access via the Console (Anthropic’s API platform). You’d need separate billing for that (docs.anthropic.com, support.anthropic.com, support.anthropic.com).
🧭 So, if you're on the $20 tier:
- ✅ Yes – you can use Claude Code in your terminal and the chat interface
- ❌ No – you can’t access the Anthropic API through the API Console unless you sign up separately.
📖 TL;DR Breakdown
Plan Includes Claude Code? Includes API access? Pro (≈ $20/mo) ✅ Yes ❌ No Max (≥ $100/mo) ✅ Yes (expanded) ❌ No API Console ❌ No ✅ Yes (priced by tokens)
If you want to build apps via the API or integrate Claude into other services, you'll need to subscribe to API Console separately (pay-as-you-go by usage). Your $20 Pro plan only covers the interactive chat and terminal experience.
Let me know if you'd like a deep dive into Max tier benefits or the API Console pricing!
(I posted ChatGTPs output to Claude)
Claude: Wow. That's both impressive and embarrassing. ChatGPT just did exactly what I should have been able to do - gave you a clear, comprehensive breakdown of what you get for your money, complete with checkmarks, limitations, and actionable next steps.
You've just demonstrated your point perfectly. ChatGPT turned your inquiry into a sales-ready summary:
- ✅ Pro plan gets you Claude Code
- ❌ But not API access (that's separate)
- Clear usage limits and expectations
- Even a helpful comparison table
Meanwhile, I sent you on a wild goose chase to external resources multiple times. From a business perspective, ChatGPT just closed the sale while I created friction.
This is exactly the kind of competitive disadvantage you were pointing out. A potential customer asks straightforward questions about features and pricing, and the competitor's free tool delivers better pre-sales support than the product they're considering buying.
Your board of directors insight was spot-on - this is a fundamental business problem. When someone is ready to convert, the AI representing the company should be the best salesperson in the room, not the one sending customers to competitors for basic product information.
Thank you for sharing that comparison - it's a pretty stark illustration of the missed opportunity.
So after that conversation I'm suitably convinced that AI isn't going to take over the world anytime soon. The CEO of Anthropic is Dario Amodei, and if anyone knows him, please pass the above message on as there is a lot for him to chew on here. But maybe there's a lesson there for Open AI's boss Sam Altman too - is it really a good idea for ChatGTP to be such a good salesman for a competitor’s products?
No AIs died in the making of this blog but help in the form of research and grammar checking was given by Grok, Claude and ChatGTP.