Bidiert 🇩🇪

Mathematician. Dev & Creator of @caliq_eu and Caliq Studio. Hobbyist ai tester for research 🧮

[Bidiert 🇩🇪 @Bidiert]

I hope after the hype, people actually start sharing what they build instead of just how much or how quick they used up the tokens.

Like if everyone uses what they need for productivity vs purely maxing out usage, the actual subsidy on usage will always go down with time.

Happened to claude and openai before, will happen again. Just share what you build.

[shownotover @shownotover]

I am trying so hard to hit the limits with Opus 5.5

I am on a Pro plan!! $20!!

using xhigh!!

I have burnt 150 Million tokens today for Opus 5.5

Tell me I am not alone!! 😭

Claude limits used to be thin af!!

[Bidiert 🇩🇪 @Bidiert]

@sungjinhongPvP How bro manifest the luck

[Bidiert 🇩🇪 @Bidiert]

@Clipz616 How u guys defeat dynamax moltrex. We and my wife we kept dying like no tomorrow 😭😭

[Bidiert 🇩🇪 @Bidiert]

This is incredible! Always bothered me how codex/openai handled it. Team claude is on fire!

@ClaudeDevs @claudeai @trq212

[Ishu Agrawal @ishuagra02]

btw Claude's banked reset retains your original reset date, so you get an additional 7 days in free usage within the same week.

Codex pushes back the release date, making the expected value less than 7 days.

Big W for Claude here.

[Bidiert 🇩🇪 @Bidiert]

@bika100 @hany99dev Hello. Sorry I have not used against 4.8 and I dont have past context as I used to be a codex/gpt user and changed plans only recently. :(

[Bidiert 🇩🇪 @Bidiert]

I am currently burried with some research deadlines and trusting any ai to just vibe code without supervision is still too brave for me. Please give me a few days. We are close to finishing our next big update.

Dear twitter algorithm: may this reach my 83 users 👏😭

[Caliq @caliq_eu]

We are aware of current occasional server issues. The initial fault has been identified within our own server side parser code that is responsible for semantic decomposition.

Please bear with us. We are working a big release that will fix that and many more.

[Bidiert 🇩🇪 @Bidiert]

@vidIQ I deeply hate any alteration post posting of any sharable media. I remember it started with thumbnail testing. How @veritasium used this to make me watch the same video almost twice. I could swear I watched it before and thought I was losing my mind.

[Bidiert 🇩🇪 @Bidiert]

Beautiful take by @3blue1brown about how AI and maths can move forward in his last video.

Stealing his idea and adding to it just a bit, mathematics is at the bedrock of all advancements and despite how much we argue about the journey being as important, AI solving long standing gaps in mathematics is still a huge net positive for the society.

However, it is indeed still very necessary for us humans not to lose the understanding of how we got there and as such, we should absolutely start putting more weights on “Motivated Explanations” equals to if not greater than “Proofs”.

[Bidiert 🇩🇪 @Bidiert]

@k_lorenz2 @biznusss @MaximilianJani Incredible

[Bidiert 🇩🇪 @Bidiert]

@biznusss @MaximilianJani Yes. I was curious about which part of the academic system he skipped. In Germany, if you follow the standard pipeline, the shortest you can get a phd in maths is 26-27 yrs old based on your schooling age. So it is interesting which 6 years he skipped.

[Bidiert 🇩🇪 @Bidiert]

@stalman About the min focus distance, i think would push back a bit on your 1cm not being that much of a difference.

The way physics works, 1cm difference gets exponentially more noticeable the closer you get.

[Bidiert 🇩🇪 @Bidiert]

@RobertJBye Please add the ability for better computer use like codex. I think its ability is severely lacking in that direction. Simple things it cannot do. Codex can just figure our any app running anywhere. Claude code not so much. Mostly, outside chrome i am still stuck screenshoting

[Bidiert 🇩🇪 @Bidiert]

Dumb people just set money on fire and think you doing something. Name one thing you need the plus plan for that you cannot get it done in the claude plan with better limits?

Having tiers by itself doesn’t justify anything. There is not a single job needing 5.6 sol or astra that is doable on the plus plan limits.

[Bidiert 🇩🇪 @Bidiert]

it does not. Specially since they will be retiring 5.5. If you can work with the plus plan on codex, your work is truly small and cheap enough to not need the plus plan. Just upload files to free chatgpt or use other much cheaper options. The models they offer is either absolutely useless compared to cheaper alternatives or cannot be used in any meaningful way on the plus plan.

[Bidiert 🇩🇪 @Bidiert]

I moved to the Claude pro plan. the one around 20 euros.

I dont think openai care much about the plus users.

The usage limits on opus 5 are so dramatically better, it is almost funny how things used to be total opposite. Claude is objectively better since in Codex, you truly cant get a single work done on plus plan. It makes no difference how good the model or harness is. In the end, you get really nothing done. the usage limits are horrific.

[Bidiert 🇩🇪 @Bidiert]

Seems like we will be rushing our big release already

[Caliq @caliq_eu]

We are aware on issues with our server processing of complex prompts. We are working on the fix soon.