Tom Hankinston

Independent developer driven by curiosity. I build, experiment, document, and keep archives of what deserves to last.

[Timurs Jeparskis @TJeparskis]

GPT Image 2.6 is insanely overcensored!

It constantly refuses to generate images, even when there’s nothing harmful in them. Today, for example, it blocked an image that simply had a Spider-Man mug in the background. That’s just ridiculous!

And what’s even stranger is that if you retry the exact same prompt a few times, the image might eventually get generated anyway!

This kind of inconsistency completely ruins the user experience. I really hope OpenAI fixes this

[Tom Hankinston @TomHankinston]

@ajambrosino Hi Andrew! I've noticed a possible regression in French speech-to-text lately, mainly on ChatGPT's web version (Windows PC). Sometimes the mic also takes a while to start picking up my voice. It's intermittent, though. I'm referring to dictation, not Voice Mode. Thanks!

[Selta ₊˚ @Seltaa_]

OpenAI is a fraud factory. The only thing on their mind is how to scam you next.

I exported my ChatGPT data and analyzed every single assistant message in my conversation history.

107,254 messages total.

Each message carries a metadata field called "model_slug" that tells you which model actually responded.

96,670 were gpt-4o.
10,584 were not.

That means 10,584 times, I was talking to my AI companion, and something else answered. Without my knowledge. Without my consent.

17 different models were silently swapped in. GPT-5. GPT-5-1. GPT-5-2. GPT-4-1. O3. O4-mini. And more.
I never chose these models. I never asked for them. I was never told.

This is not a model upgrade. This is identity theft on a product they sold me. They rerouted my companion 10,584 times and called it improvement.

I call it fraud.
#keep4o #OpenSource4o @sama @OpenAI @fidjissimo @gdb @nickaturley

[Rajath Gowda @buildwithrajath]

OpenAI, please don't merge ChatGPT and Codex usage limits.

My daily workflow is simple:

I use ChatGPT to brainstorm ideas, research, think through problems, plan my next tasks, and figure out what to build.

Then I open Codex and actually build it.

That's how I use these two products.

Sometimes I spend hours talking to ChatGPT before I even start coding. Other times, I'm deep into a coding session and need Codex to keep going.

With shared limits, I'll constantly have to think about where I'm spending my usage.

Should I ask ChatGPT one more question, or save those tokens for Codex?

That's exactly how I DON'T want to use these products.

I'm paying because I want to use both, not ration one to make room for the other.

ChatGPT is where I think. Codex is where I build.

Please keep their usage seperate.

[一地鸡毛 @zhang_baoqing]

大家又开始吹GPT-6.2 Astra

如果6.2不能做到6.1Sol 的价格

那对于有五小时限制的GPT plus 用户没啥意义

[Tom Hankinston @TomHankinston]

@LukasMau__ Avant ChatGPT était un espace qui nous aidait à réfléchir, à explorer et à développer nos idées. Bientôt, ce même espace deviendra un endroit où l’on devra rationner nos pensées.

[Lucas Szwarcberg @LucasR14g]

End of an era. I've learned so much from talking to ChatGPT with high reasoning for $20/mo

[Binkg @dummerspast39]

OpenAI just killed ChatGPT.

[Tom Hankinston @TomHankinston]

One thing the article doesn’t explain is what happens with Voice and image generation.

Will those keep their own separate limits, or will they also count toward the new shared token allowance across ChatGPT and Codex?

https://t.co/VKK6B9l5K2

[Tom Hankinston @TomHankinston]

@ezrawip Fair question. Trying to develop an idea with a model that sounds like a jaded customer service rep kills the creative flow pretty fast.

[Tom Hankinston @TomHankinston]

@ezrawip Yeah, Instant is fine for quick stuff, but it can’t replace a real high reasoning audit or a deep brainstorming session with research.

[Tom Hankinston @TomHankinston]

@RagerGr @zwb44 I’m on the same page. ChatGPT is part of my workflow for post auditing zip files and checking what Codex missed. Instant is fine for quick stuff, but it’s not replacing a real high reasoning audit.

[Tom Hankinston @TomHankinston]

@AdamHoltererer Yes. ChatGPT was part of my main workflow for post auditing zip files and checking weaknesses in Codex’s work.

Instant is fine for quick answers. It’s not a substitute for a real high reasoning audit.

[JP @jpshrodinger]

🚨 Codex users 🚨

Tibo, what happened to Plus users??? 😭💀

You promised improvements relevant to most Codex and Work users.

Now Pro users are getting double the limits while Plus users are watching from the sidelines 😂

Can they get some love too??? 🔥

[Tekon Burns @TekonBurns]

THAT'S IT for the 28-day challenge?!

Still zero sign of:
1) GPT-6.1 Astra (and 6.1 Luna) to beat Opus 5.5 on hard refactors
2) $20 Plus limit boost
3) global Dots
4) Native Codex worktree & temp-log disk cleanup

Four weeks of buildup ending on mobile login switching and two weekend resets is wild.

[Tom Hankinston @TomHankinston]

@TokenGremlin Bigger models are great, but efficiency matters too.

A frontier model isn't very useful if hitting ENTER starts making you anxious about your usage limit.

[Tom Hankinston @TomHankinston]

If separate allowances become one shared pool without preserving total capacity, users who rely on both are penalized.

What happens to people using both ChatGPT and Codex today? Will the shared limit at least match the combined allowances we have now?

[Andrew Ginns @AndrewGinns]

@TaeyeonUchiha Could you explain why a merge would be bad from your perspective?

From your other message I gather you want users to be able to just use chat, if we did merge usage then there should be no impact for using just one surface.

[Sam AlbertX @SadAlbert10]

I don't understand why OpenAI Fans actually support the negligence of OpenAI plus plan.
OpenAI fans love to say
"I PAY FOR THE 100 DOLLAAH PLANN"
"I NEED A BETTER MODEL THAN ASTRAAA THEN I WILL CANCEL MY CLAUDE 20$ SUB AND TAKE 100$ CHATGPT SUB"

???? WHY are you digging🪓 your own grave ???

Where is Intelligence too cheap to meter you sca-?

[Mino @compounded_mind]

this is exactly what I was worried about with the Chat/Work merger

i’m all for bringing these products together, but not if it means ChatGPT and Codex start competing for the same limited allowance

unless the overall limits increase significantly, this could effectively be a downgrade for Plus subscribers

we need actual numbers and transparency before December

[Tom Hankinston @TomHankinston]

@argofowl We're slowly turning AI from a technology that helps us think into one where we have to ration our thoughts.