kplcode

Web development & Consultancy. latest iOS App PunchScore

[kplcode @Kplcode]

Adding #AppleWatch support for PunchScore iOS app. Test the power of your punch from your Apple Watch & get latest boxing news at a glance

[kplcode @Kplcode]

@ClaudeDevs Opus 5 was also completely nerfed and consumed millions of tokens that generated code which had to be redone.

[kplcode @Kplcode]

@theo I had the exact same experience. I used Fable and opus until I had consumed 100%. The went into gpt-5.6 sol ultra and realised that ape some tasks, when steered, is actually better than Fable

[kplcode @Kplcode]

Can I suggest something that would be genuinely useful.

- resets are banked

- you can set the date of your subscription schedule. This would allow me to overlap my Claude subscription with GPT. So when one gets maxed, the other is only 2.5 days from being reset

Professional development using these tools is becoming unpredictable. Fine for vibe coding users, not good for professionals with deadlines and targets.

The alternative is having only one frontier model (Claude or gpt) and the second subscription being a Chinese offering (practically unlimited)

Currently running Claude and gpt doesn’t work and leaves gaps

[kplcode @Kplcode]

@OpenAI The only models I use are

Fable 5 ultracode
Opus 5 ultracode
GPT-5.6 sol ultra

[kplcode @Kplcode]

But the issue is you have to chat using Work in order to get the model selector to display. The model selector isn’t present in normal chat mode. You may find less people use chat for chat and the app using Work for chat and an iOS device to chat via the model picker. You have to use Work or browser not the native app. I think this one result in more people adopting Work.

[kplcode @Kplcode]

@Tesco your new “show pin “option on the mobile app only shows the pin that was originally sent with the card initially. Meaning it’s showing the wrong pin if the pin has been changed by an ATM machine since it was initially issued.

[kplcode @Kplcode]

@just_sahitya_ Opus 5 is better at coding and engineering. I’m not sure about creative writing etc.

[kplcode @Kplcode]

@sama This is exactly what it’s best at. This is the best, most useful example of that feature.

[kplcode @Kplcode]

So people have stated my negative reaction to resets being forced and not banked. I was on 80% remaining and 1 day until my natural reset.

Imagine you have an £80 gift card, and tomorrow you’re due another £100 gift card.

Instead, someone swaps your £80 card for a £100 card today and cancels tomorrow’s £100 card.

You’ve gained £20, but lost £100.

Overall you’re £80 worse off.

[kplcode @Kplcode]

I think @OpenAI are trying to move to a single app that does everything and no real control or view of what’s going on under the hood. The app is called ChatGPT, then when you open it you choose option ChatGPT or Work. Isn’t work ChatGPT then? It say options “Chat only”, “Work”, “Coding”

[kplcode @Kplcode]

@thsottiaux Is that a good thing?? Serious question - could that indicate that it’s being used less for engineering tasks? Engineers using alternative options?

[kplcode @Kplcode]

Personally, I use cursor with OpenAI vs extension, with Claude VS extension, custom CLI for Jira automation and slack integration. I don’t use cursor’s own agent for building, I only use the cursor agents for live monitoring on all processes queued, pending and progressing. It also makes team collaboration much simpler. Everyone has their own setup preferences, that’s what works for me.

[kplcode @Kplcode]

If you pace and schedule your usage then the resets work against you. If you burn through it all then they work for you. For people working in the professional field, managing their tokens and workload, it works against us. I’ve raised this a few times but isn’t taken on board.
When @AnthropicAI do a reset they keep your dates unchanged. So it can only benefit you.

[kplcode @Kplcode]

@thsottiaux Please stop doing this, this actually works against people as you change the date too. If a person is more than half through their week but has more than half remaining, your reset is worse off. Please do banking or make it optional.

[kplcode @Kplcode]

@melvynx So far, yes. Though, when a new model is released it’s great, then gets nerfed after the first week. It feel like GPT-5.6 Sol is worse than it was. That said, Fable is amazing still

[kplcode @Kplcode]

My views on the current AI landscape from an engineering view, and the models I personally think are the best, in order:

Fable 5
Opus 5
GPT-Sol 5.6
Opus 4.8

These are the only models I use. If I want to use lesser/cheaper models, I’d still have one of the above models orchestrate things, but I’d state to use lesser models for the grunt work.

[kplcode @Kplcode]

@mois_u Yes this is an issue actually. Something @AnthropicAI seem to have overlooked. If they are unwilling to provide banked resets then they should give you the option to opt out of automated resets

[kplcode @Kplcode]

@JustinBleuel @ChatGPT An engineering version, just a single engineering powerhouse. Not the vibe version, but one that is geared for engineers. I’d like to move from CLI eventually…..

[Claude @claudeai]

Introducing Claude Opus 5.

It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price.