Thought #25 - Claude Gets Smarter, GPT-5 Gets Colder

Memory, context, and why not everyone is in love with the latest AI releases.

Author

Laura Gemmell
August 13, 2025

Hi lovely humans,

Phew - yet another big week. Not so much news this week - but two really big releases. GPT-5 is here. And we’re whelmed. Also, Claude finally gets memory!

I also wrote this whole thing from scratch this week - rather than AI assisted as I had too many opinions.

Our Week in AI

Has mostly been taken up testing our new releases. More of our random day-to-day next week.

AI New Releases

GPT-5

The long awaited release of GPT-5 has been … interesting.

It’s a bit of a long story, so I’ve written a blog:

Why Have So Many People Taken the OpenAI Release Personally?

If you have been on the internet in the past week (particularly on Reddit), it has been really hard to avoid the overwhelmingly negative reaction to the new OpenAI (specifically ChatGPT) release. Her...

www.taughtbyhumans.com/blog/gpt5-release-community-response

tldr; people got upset, they have returned GPT-4o.

Back to the release - it’s fine. Works well for some things, not so good for explaining tech or long form writing. Still getting tripped up by maths and counting letters. Seems to be hallucinating less. More detailed thoughts in our blog:

Can You Ever Just Be Whelmed? Well, that’s how GPT-5 makes me feel.

“I know you can be overwhelmed. And you can be underwhelmed, but can you ever just be whelmed?”“I think you can in Europe.” Clueless I love any excuse to quote a 90s chick flick, but this quote did f...

www.taughtbyhumans.com/blog/trying-gpt-5

On the API side - GPT-5 is now cheaper than all the GPT4.1 models. We haven’t yet tested it out on our platform, but will do so soon.

Memory has long been the saving graces of ChatGPT (this is where the system remembers things you’ve said in other chats - weirdly one of the most useful and most annoying thing at the same time). So it’s a BIG deal for Anthropic to announce this.

Currently this is only for users who are paying for max or above ($100 a month plus), so we (as a cheap little startup) don’t have access.

But one of our advisors - a hardcore vibe-coder and Claude fan - has tested it out for us. In his words - Omg so “Claude --resume” which lets you pick past chats (if you’re a max subscriber using Claude Code - check it out). He has also mentioned using Claude to write its own claude.md file - he’s vibe-coding on a whole other level.

Claude Context Window Increased

Anthropic have also increased the context window (basically how much Claude remembers in one scenario) on Claude Sonnet 4 to 2 million tokens (for simplicity these can be thought of as words). This is 5 times more than before, so quite a big change.

Interestingly for those using the APIs, they have introduced tiered pricing:

Pricing is nearly double for prompts above 200k tokens

AI News

Not Quite News, But Worth a Read (or Listen or Watch)

  • Should IT and HR merge in the AI workplace? - study

LinkedIn AI Poll

How are data privacy concerns changing how people use AI tools?

The data privacy nerd in me is thrilled to see the majority of people are not letting it stop them use AI, but are thinking about the data they share!

Vote in this week’s poll - please!

This week we want to know what you think of GPT-5.

Final Thoughts

As always we hope this was helpful!

Feel free to share this with anyone who might find it useful.

Next week, more testing of the new releases and some thoughts on using AI.

Laura
Always learning