GPT 5

[sniffer66]

I was impressed to start with, but then ended up in a constant loop.

"Do you want this extra bit?" Yes
"Great, do you also want this extra bit" yes
"Great, before I give you the code, how about this as well?" Yes
Repeat ad infinitum until I demanded it just "give me the f**ing code !!!"

Now it's crashed

[Kai]

I was impressed to start with, but then ended up in a constant loop.

"Do you want this extra bit?" Yes
"Great, do you also want this extra bit" yes
"Great, before I give you the code, how about this as well?" Yes
Repeat ad infinitum until I demanded it just "give me the f**ing code !!!"

Now it's crashed

You can turn off follow up questions in the general settings

But it may not always work, so you may need to give it a custom instruction (trait) to stop doing that unless you explicitly ask for it

I personally see those as optional so don't mind it, the better it knows you the better it can predict your next query or course of action etc

The "agreeable sycophant" trait that it kinda has by default which Shaun mentioned earlier can be a real issue sometimes though, you don't really want to be talking into an echochamber.. something to watch out for IMHO

It can do some serious harm to the wrong person with the wrong mindset, but these are pretty extreme outlier examples : https://www.theguardian.com/australia-n ... ses-ntwnfb

[ShaunWhite]

Liking v5 so far. Coding seems much better with better suggestions and more efficient code. Often a bit long-winded and by the book rather than creative or with maintenance in mind but that's to be expected because it's not no lived experience of it.

[sniffer66]

I was impressed to start with, but then ended up in a constant loop.

"Do you want this extra bit?" Yes
"Great, do you also want this extra bit" yes
"Great, before I give you the code, how about this as well?" Yes
Repeat ad infinitum until I demanded it just "give me the f**ing code !!!"

Now it's crashed

You can turn off follow up questions in the general settings

But it may not always work, so you may need to give it a custom instruction (trait) to stop doing that unless you explicitly ask for it

I personally see those as optional so don't mind it, the better it knows you the better it can predict your next query or course of action etc

The "agreeable sycophant" trait that it kinda has by default which Shaun mentioned earlier can be a real issue sometimes though, you don't really want to be talking into an echochamber.. something to watch out for IMHO

It can do some serious harm to the wrong person with the wrong mindset, but these are pretty extreme outlier examples : https://www.theguardian.com/australia-n ... ses-ntwnfb

It was more that it was stuck in a loop. Until I insisted, it never stopped asking or repeating the questions

[Emtaxx]

Thew new AI 5 is more advance thats for sure. I've paid for it for a day or two just to help me out with building something. Its alot differnt from the previous Gen actually. its has more flow and more explinations, but it defintely throws out a lot more words and would say gets carried away a little. As for building you BAF files. it still can't do that mind you sadly had to stress for a few hours building my own still!

[Kai]

It was more that it was stuck in a loop. Until I insisted, it never stopped asking or repeating the questions

Yeah, it's been known to lose coherence sometimes and become forgetful or delusional

Hence my initial hallucinations remark

[sniffer66]

It was more that it was stuck in a loop. Until I insisted, it never stopped asking or repeating the questions

Yeah, it's been known to lose coherence sometimes and become forgetful or delusional

Hence my initial hallucinations remark

I keep getting lockups and quite a few problems. I might go back to Gemini 2.5 pro for a little longer until they iron out all the issues

Problem is, you get a lockup, and try a Refresh and it loses all the session context so keep having to repeat myself. Or wait for 15m for it to do something. Which could possibly be due to everyone caning it today

[Euler]

After every major release half of silicon valley is in danger of a core meltdown. So this is nomal, but it will settle.

[sniffer66]

After every major release half of silicon valley is in danger of a core meltdown. So this is nomal, but it will settle.

It's not just the speed, I've given it my headers multiple times and its still either asking me what they are or guessing them (incorrectly). Wasting too much of my time repeating myself. Going back to Gemini 2.5 Pro until they sort it out

[weemac]

Sheesh, it makes plenty of errors. It's making up non-existent horses for tomorrow's racing even though I showed it the race card, and it doesn't know well-known footballers because they're relatively recent signings. I also showed it a photo of Frank Lampard who it said it couldn't identify until I typed "Lampard", which prompted it to say "That's Frank Lampard." Great tool which I'll use plenty, but I don't trust its veracity.

[bobalski]

1. Just click here : https://chatgpt.com/
2. And ask it anything, talk like you would with a normal person

It's dead simple, you don't even need an account to start talking to the free ChatGPT version. Conversations etc won't be saved, so you should register if you find it helpful, and you can still use the free version.

Thx Kai, cheers

[wearthefoxhat]

Just spent 15 minutes with GPT 5 knocking together a monte carlo sim for some horse race ratings. (I can also used a variant on football as the season progresses and the stock market at some time)

Q1.png

It outlined and explained some examples. I then asked it for a fully working formulated excel sheet as a starting point. No problem, but there was an issue with the way it ranked a column. "We" got it sorted in double quick time though, and it has a sense of humour too.

MCarlo.png

[weemac]

...... and it has a sense of humour too.

It told me it could refuse to open the pod bay doors.

[goat68]

You still have to be careful, it still makes things up, but that's expected really:

Code: Select all

No, the Betfair Order Stream does not show just your orders.

What the Order Stream actually shows:
The Order Stream in the Betfair Streaming API provides real-time updates for all unmatched orders on the markets you subscribe to — not just your own bets.

It lets you see the full live order book at the individual order level (all users’ orders, with price, size, side, status, etc.).

Code: Select all

Yes! When you subscribe to the Betfair Order Stream for a market, you receive real-time updates on all unmatched orders placed by all users in that market — not just your own orders.