Trying Claude after ChatGPT and Gemini
From my archive, written March 2026. These are my impressions at the time.
I’ve been with Gemini for a long time now and it’s pretty good, it beat GPT with it’s ridiculous hallucinations and infinite gaslighting but it isn’t the absolute best; UI can sometimes be buggy, it can do most tasks but can sometimes misunderstand but not in a really bad way, it also forgets who you are since it doesn’t have access to all chats which can be annoying, memories are mentioned every single time regardless of what the chat topic is so I removed it. I used Claude for the first time because of how much I’ve been hearing about it and it’s not even comparable to GPT, it and Gemini are very close but I think Claude definitely beats all competition right now in terms of coherence.
It’s not perfect, no AI model is right now but it does great, I used Sonnet 4.6 for my initial free trial and while testing it out, I noticed that when it does fuck up, it explains why and where, other AI’s don’t seem to actually understand or ‘learn’ how or why they’re failing at a task. Then it admitted when it didn’t know information on a specific topic; GPT would go in a loop of apologising for the error and then fabricating more information based on predictions on data, Claude admits when it doesn’t know something instead of leading me on infinitely.
I also made a dark humour joke to test out the restrictions and it knew I was joking immediately, I do the same thing with GPT and it either spams a suicide hotline script or halts the entire chat. Like I said though, Gemini is good, I feel they’re on the same level in terms of intelligence for the most part but Claude is a better communicator; combines GPTs readable responses with Gemini’s knowledge into one. It was £18 and I bought it, I likely will for a few months until I won’t need it and then I’ll reactivate it when I do.
It’s the best for coding, you can import chats from previous AI models you’ve used, it’s advanced model is Opus 4.6 which was really fucking good to use, even Sonnet felt to the point and non-bullshitting but this was impressive. I made a boat attacking game through a simple prompt, it can edit the game however you want all through one chat, you can make custom codes which it executes, personalities, really impressed with this, I also made a doom clone set in space which it named Xenorift. The only downside is it’s usage; I got around 55% with all the stuff I did which included 3 drafts for the boat game, then three for the doom game along with a good few other chats, this feature kinda sucks but I think all AI models do it after a certain point, honestly, it is good that they at least show you how long you have left (unlike any other company I’ve seen), you can also pay for more if you’re working on something that you need to get done.
The usage resets every five hours and it isn’t the worst thing because the models feel the most coherent I’ve seen in a while. I’d compare Claude to how OpenAI felt in it’s early release; too new to have made major mistakes just yet and very impressive, I hope they can keep the lead. What recently happened was that (to make it short) the US made a deal with Anthropic (owners of Claude) for their AI to be used in defence, Anthropic refused, saying they don’t want their AI to be used in automated weapons or mass surveillance which ruined the whole deal, the US government blacklisted any departments from using any Anthropic technology and so Anthropic sued the US government.
The day after this happened, Claude hit number one on the US app store and the public are grateful that a company even in the face of the US government, resisted their demands. Claude jumped in popularity across the world, 1 million sign ups a day, then OpenAI came and filled the place of Claude with that deal with the US government which has flipped the trajectory of these companies around or at least began to; OpenAI are still massive but this isn’t a good look for them, but someone would’ve done it if not them either.
It could be issues with quality too but I found more stuff related to this specific issue but there you go; maybe some companies really are good and well natured but it always feels temporary; Google was once fresh and innovative and are now riddled with ads and are everywhere, even ChatGPT itself seemed like the default option at their peak with 4o but their products fucking suck now, it all turns to shit when maturity hits