So Fable 5 has launched with huge hype. They say that its core is way too capable, so it is artificially dumbed down. They said the same thing also about GPT-2… and it is considered now ancient tech…
The price tells you everything you need to know about the direction of closed-source AI.
$10/million input tokens.
$50/million output tokens.
I tested it and it is an impressive model. But do the math for a production agent, consuming tokens all day, every day. A single long-running agent workload can end up costing more than a regular programmer’s salary per month!
The thing is that using it, we are renting intelligence and the landlord sets the rent and each new generation of models consumes more of our funds than the previous one. I’ve seen posts on LinkedIn from people whose API bill jumped from $2,000 to $40,000 in a single quarter, without releasing anything new…
I am sure this models is as good as GPT 5.5 which is kinda the same as GPT 5.2 so at this point it dose not really matter anymore… we are in a place where inteligence starts to matter and we all know this models do not think…
I already said a few weaks back that even GPT-5.3 codex models are too expensive. They’re not really usable for production unless you have a lot of money to spend.
Personally, I’m changing the way I use AI. I design the entire architecture myself like in old days and only use AI when I need help ore stuff like documenation and info. This way, the project remains my own, and I continue learning throughout the process.
I tried vibe coding, and it just doesn’t work. I don’t think it will ever work for anything beyond creating a close clone of an existing app, and even then, its effectiveness is questionable.
This will change drasitcally when we will have gpot 5.5 as a local model boy those will be wired days to have such a tool on your desk and to be able to train it based on your habits, not jsut coding but everything you do, that would b e fantastic but we have to wait for that I thing 1-2 years.
As for AI hype is kinda over, no matter what they do now it will not feel like the initial magic when gpt came out!
I noticed all of these AIs are effectively useless at real-life questions that are not settled. Things like making money, investing, making life or business decisions,
Which kind of tracks to what you said about “beyond creating a clone of an existing app”.
A human is much more opinionated, has their own real life experiences and lessons learned along the way
This is why they want to train models on our day-to-day activities, just like Meta reportedly imposed on its employees. Either way, this approach will not work, not with the current state of these models and we as humbans we will defenitely not agree with this… it goes too far!
These models cannot replace my train of thought when I architect and design my work because they have no real reasoning or imagination. I also need to point out, once again, that this technology is incredibly expensive, and it has become one of the most disliked and hated at this point technologies ever created. Much of that is the fault of the CEOs in charge, with their constant apocalyptic speeches and unrealistic predictions. These are not people acting in humanity’s best interests; they are money-hungry individuals driven by profit above all else.
This situation will not remain unchanged. The companies behind these models essentially built them using the collective knowledge that people freely shared online and who knows how much they stole without approval, yet they take most of the credit and profit for themselves. Mark my words: things will change drastically in favor of creators and ordinary people in the near future. You can already see a shift in the way many CEOs are starting to talk compared to a year or two ago.
As for coding, AI has largely solved many of the implementation challenges, but people often confuse coding with engineering, they are not the same thing. If you’re not a good engineer who also knows how to code, yes knowing how to code reflects in your product quality, your application will only be as good as your own skills and decisions.
For senior developers, AI is something entirely different. It enables us to push projects further and build things faster than we ever could before. The real value isn’t just in writing code, it never was! It’s in knowing what to build, how to architect it, and how to leverage AI effectively combined with you skill to achieve better outcomes.
As I said no matter what they do from now one the hype is over, codex 5.3 is more than enough to help you with any app that you might want to create… and gpt 5 to answer questions until it agress with you no matter what you say
Now things are starting to calibrate, and reality is finally hitting these AI companies. You can see it in their pricing.. it’s becoming ridiculous also al companies that went full AI are waking up to reality as well when they see their bill at the end of the month.
I knew this would happen eventually. These models run on extremely expensive GPUs and consume massive amounts of power. On top of that, the average lifespan of these GPUs is only around 3–5 years at best; after that, they become obsolete and most of them start to malfunction or need to be replaced. From a long-term perspective, it’s not a profitable investment, imagine investing a billion $$ in this s*it and five years later just gone…
As long as these systems rely on transistor-based hardware, the fundamental limitations will remain. No matter what improvements they make to the chips, many of the same problems will still be there. They can optimize performance and efficiency, but they can’t completely escape the physical and economic constraints of the hardware.
