LogicLeap Labs

Building apps and AI tools with Claude Code and Codex, every day since launch. The real setups, numbers and mistakes, shared as I go.

[LogicLeap Labs @logicleaplabs]

@trq212 Plan mode was useful to me mainly because Claude is great at generating ideas, so I'd get it to produce a plan, then mark it up and chop and change it.

Writing this out, I realised Artifacts do that just as well. Kill it.

[LogicLeap Labs @logicleaplabs]

@thsottiaux Ultrafast looks....Kindafast

[LogicLeap Labs @logicleaplabs]

202B tokens on Claude Code since May. here's why I stopped trusting Codex https://t.co/5mDmtCaL6a

[LogicLeap Labs @logicleaplabs]

https://t.co/Fpk6EvlHvv

[LogicLeap Labs @logicleaplabs]

@AdamHoltererer @TijnKnp Not when you're out of credits it can't.

[LogicLeap Labs @logicleaplabs]

@AdamHoltererer @TijnKnp ...and swap your api keys every £100 spent?

[LogicLeap Labs @logicleaplabs]

@trq212 I love this!! Thank you

[LogicLeap Labs @logicleaplabs]

@blueemi99 The thing with Opus 5.5 is: it's so good - you don't need to be switching models in the first place.

OpenAI added it as you have to switch pretty much every single message.

[LogicLeap Labs @logicleaplabs]

@_ak_111 @thsottiaux should've read the post that tibo was replying to

[LogicLeap Labs @logicleaplabs]

@thsottiaux Respect for UK/EU users would be a great start

[LogicLeap Labs @logicleaplabs]

@sama If you use @claudeai you will find you can actually trust it much more.

[LogicLeap Labs @logicleaplabs]

@pcshipp I usually wake up and think "what an awful nightmare, thank God I moved to Claude"

[LogicLeap Labs @logicleaplabs]

Codex Cloud isn't free - it comes out of your ChatGPT subscription, same as Claude cloud sessions come out of your Pro/Max plan. Nobody's paying "extra" for either. Difference is Anthropic's just given existing subscribers $100 of cloud credit on top. So if anything, Claude's the free one this month.

[LogicLeap Labs @logicleaplabs]

@X

[LogicLeap Labs @logicleaplabs]

@BenSyne @GeoffreyHuntley Genuinely curious what it was working on for a week? Like what did it ship, and what was checking the output?

[LogicLeap Labs @logicleaplabs]

@GeoffreyHuntley Funny, I get the exact opposite. Opus feels collaborative, like actually thinking a problem through with someone. GPT is throw a prompt and hope it lands.

And every time I leave GPT unattended, it gets stuck in a loop and burns 100% of my usage on a single task.

[LogicLeap Labs @logicleaplabs]

@DavidOndrej1 The Queen wasn't a leader, she was a constitutional monarch. She didn't set a single policy in 70 years, that was Parliament and the PMs. Merkel's fair game, she actually ran Germany. Lumping them together just shows you don't know how the UK works.

[LogicLeap Labs @logicleaplabs]

@donerkebab7 @Priyannkaaaa Not quite. It's one-way. Fable usage counts toward your All models limit, but the Fable meter only tracks Fable. So burning Fable eats into everything else, while using the other models doesn't touch your Fable allowance.

[LogicLeap Labs @logicleaplabs]

Claude has been fantastic for my day-to-day app building. One thing I love: getting it to generate a plan, then marking it up and reshaping it around what I actually want. Claude Code users: what's the workflow you'd hate to lose?

[LogicLeap Labs @logicleaplabs]

@msirajulhaqq I swear. @trq212 can you set up a public profile system so I can credibly share my stats?

[LogicLeap Labs @logicleaplabs]

Is this too much claude usage