I like Claude. It is one of the more fun AI Chatbots I have used. It can write your code, edit your novel, summarise a 200 page document and even explain advanced quantum physics in basic english. However, when you spend as much time with it as I have, you will start bumping into edges. They aren’t bugs or failures of any kind, just walls and here are five of them.

It can’t paint you a picture or a movie

Open ChatGPT or Gemini or Grok and type “Generate a sketch of a backflipping dinosaur.” Seconds later, you will have something. You won’t always like it but it is still something. Do the same with Claude and you will get words about what the dinosaur would look like. The words may be helpful, but they still aren’t the dinosaur you wanted to see backflipping.

Also read: OpenAI defended Anthropic in its feud with US govt: Here’s how

Claude has zero image or video generation capability. No Imagen or DALL-E equivalent that they have tucked away. If you need visuals, it will hand you SVG code or a detailed prompt to paste into Midjourney and send you on your way. It’s like if you asked an art critic to paint your portrait, you will get the insight but not the execution.

Personalisation is still a dream

Imagine feeding Claude every document you have ever written, every process and quirk of your style and having it just know and recognise you. That dream exists but it’s not available to regular or even Pro users for that matter.

Fine-tuning Claude on your own data is strictly an enterprise and API affair. What you get instead is context. You can paste things in, build a project and repeat yourself across sessions. It works but it is a workaround, not a solution. The model remains fundamentally the same for everyone.

The usage window runs out

Claude Pro is not unlimited and it will remind you of that fact exactly when you need it to fix a code right before your submission window. There is a rolling usage window of about 5 hours and if you’ve been deep in a mammoth coding session or grinding through a stack of documents, you will eventually hit a wall and get a polite message telling you to come back later.

Also read: Iran-US-Israel war: Guy used AI to build 24-hour replay of Operation Epic Fury

Anthropic knew this was a problem which is why they launched a Max tier at $100 a month. So the limitation is real enough that they built a whole new product category around solving it. That tells you everything you need to know. 

It speaks in turns

Claude has a voice. What it lacks is conversation. With ChatGPT’s Advanced Voice Mode or Gemini live, you can interrupt it mid-sentence and it will stop, recalibrate and roll with you. It feels more alive. Claude on the other hand has a voice mode that is turn based which can be a real nuisance at times. 

Audio and video processing

Slide Claude a pdf or a spreadsheet or a photo and it will dig right in. Hand it a podcast episode, meeting transcript or a video clip and it will stare at you blindly like a deer in headlight. Claude can not process any audio or video. It just simply can’t. On top of that, you have the file uploads. They are limited to 30mb per file and 20 files per conversation in a standard chat. For people who work with recordings, calls and videos, this means a detour to a transcription tool before Claude even enters the picture.

Also read: Grok didn’t really predict US strike on Iran: Here’s why