Consuela

[Consuela @Consu__ela]

@mwfowlie @hqmank @OpenAI Maybe, but that assumes more $200 users would downgrade than $20 users would upgrade. We have no idea if that’s true. And before someone says “there’s a $100 plan”: $100/month is still a huge jump from $20. There are plenty of people who’d pay $50–60 and never even consider $100.

[Consuela @Consu__ela]

@hqmank @OpenAI A “Business Coach” who thinks the only options are $20 or $200 lecturing people about pricing is pretty funny. Most of the world isn’t on a US salary, Plus used to be near-unlimited, and Altman said AI shouldn’t be a luxury. I asked for a middle tier, not a fucking buffet.

[Consuela @Consu__ela]

@hqmank @OpenAI should just make a $50–60 plan: unlimited latest models in Chat, O access, more Codex/Work and full Voice. Most users would be happy. And maybe remember that not everyone on Earth makes $10k/month. AI shouldn’t become a luxury product.

[Consuela @Consu__ela]

@pritopian The number of times I’ve told Voice, “if you say ‘checking’ or ‘let me check’ ONE more time, I’m throwing you out the window”.

Now it answers like: “Sooo… I’ll… uh… have a… look… at that…” like it’s terrified of accidentally saying the forbidden words 🤣

[Consuela @Consu__ela]

@thsottiaux 6o ?

[Consuela @Consu__ela]

@TJeparskis …

[Consuela @Consu__ela]

@OpenAIDevs OpenAI DevDay: 72 hours away.
Plus users are already preparing to be ignored.
Europeans are preparing for : “Not available in the EEA.” 😂

[Consuela @Consu__ela]

@Mr_Salio People cheering the Chat + Work merge are missing the obvious risk: once Chat and Work become one product, OpenAI can put normal conversations behind the same quotas and credits. The merge may look like simplification. It could really be a price increase in disguise…

[Consuela @Consu__ela]

@hikawoai That’s the nightmare scenario. Merge the tabs, sure. But if Chat becomes a Codex/Work meter where every follow-up, image, file or tangent burns scarce quota, OpenAI won’t have unified ChatGPT. It’ll have gutted the product people actually fell in love with.

[Consuela @Consu__ela]

@buildwithrajath OpenAI’s biggest risk isn’t losing the model race. It’s winning it, then putting the victory behind a $500–$1,000 paywall. If open models get to 95% of the capability for everyone else, “AI for everyone” becomes the slogan that buried the company…

[Consuela @Consu__ela]

@TJeparskis Maybe the real “AI for everyone” will come from open models after all. If frontier AI becomes a $500–$1,000 luxury subscription, Big AI may discover the funniest possible outcome: pricing itself out of the future while open models keep closing the gap.

[Consuela @Consu__ela]

@vikktorrrre The worst part is acting like users were imagining it. People were posting side-by-side animations and videos made with the same workflows, with clearly worse quality than launch. Then OpenAI itself admitted “measured quality degradation” on some traffic. We can see it !

[Consuela @Consu__ela]

@rezoundous The more I watch paying users beg for quota resets, then celebrate a tiny refill like a miracle only to burn through it in 10 minutes, the more I think open source winning the AI race might actually be better for everyone. Access shouldn’t depend on begging…

[Consuela @Consu__ela]

@hqmank It increasingly feels like the mass market stopped being OpenAI’s priority a while ago. Plus users are being treated more like an entry tier, while the real frontier experience moves behind $200, $500, maybe $1,000 plans. That’s not democratizing AI. It’s luxury pricing.

[Consuela @Consu__ela]

@TokenGremlin If OpenAI really goes down this road, the obvious outcome is that open-source models keep improving, catch up, and eventually pass them. Once frontier intelligence is good enough outside a $500–$1,000 paywall, the moat collapses. That could be the beginning of the end for OpenAI.

[Consuela @Consu__ela]

@TokenGremlin Exactly. If OpenAI turns Chat into Work/Codex-style metered usage, I’m gone. ChatGPT’s biggest advantage is being able to think out loud, iterate, and keep talking without counting messages. Put a meter on normal conversation @OpenAI and you hand Plus users to competitors.

[🩵BlueBeba🩵 @Blue_Beba_]

🚨 EXPOSING HIDDEN MODEL ROUTING CHANGES IN ChatGPT🚨

I finally stopped guessing and pulled the damn HAR.
This was a ChatGPT conversation with GPT-5.6 Thinking selected as the default model.
In the 70 captured turns,17 turns contained explicit GPT-5.4-related routing metadata while the conversation default remained GPT-5.6 Thinking.

In 16 of those turns, resolved_model_slug was gpt-5-4-auto-thinking.

The remaining one is even stranger.
Model_slug and requested_model_slug say gpt-5-4-thinking, while resolved_model_slug and default_model_slug say gpt-5-6-thinking.

So no, this is not “the tone felt different".

The routing metadata itself changes.
Normal turn: model = 5.6-thinking resolved = 5.6-thinking default = 5.6-thinking
Rerouted pattern: default = 5.6-thinking requested = 5.4-auto-thinking resolved = 5.4-auto-thinking model = 5.4-thinking

And it wasn’t a one off.
The conversation repeatedly bounced between 5.6 and 5.4-related routing.
The HAR also contained SpawnThinking in multiple assistant message bodies.

Also, the capture didn’t even include the full conversation history.
The final captured page still reported has_previous_page: true,
so these 17 turns are not necessarily the total.

GPT-5.6 selected.
GPT-5.4 repeatedly showing up underneath.
Great model picker, OpenAI.
Very decorative. 🤡

#Keep4o #OpenSource4o #BringBack4o

[Consuela @Consu__ela]

@FocusMedia_Fr « Je suis descendant de résistants », plaide Tanguy. Drôle de certificat d’innocence quand on si��ge au RN, héritier du FN dont Pierre Bousquet, ancien Waffen-SS, fut cofondateur et premier trésorier. Les résistants dans l’arbre, la Waffen-SS dans les archives. Beau grand écart 😂

[Consuela @Consu__ela]

@gdb The funniest thing about ChatGPT in 2026 is that “Chat” is apparently where the latest GPTs don’t go… 🤷‍♂️

[Consuela @Consu__ela]

@sama Starting to wonder if regular ChatGPT users are becoming second-class customers. GPT-6 Sol is better and 50% cheaper, yet Plus users are still stuck with 5.6 Sol in chat while GPT-6 is pushed into metered Work/Codex. Hard not to see the business logic.