@omarsar0 What's the logic behind using a model with inferior context performance for decisions that rely on ability to reason over large amount of information accurately?
Fable's ability to reason over long context is not much better than Opus 4.8 and worse than GPT 5.5.
I haven't used GPT 5.6 yet, but Fable for me has been more like an exceptionally good yes man, but what I really want is yes-and man.
When I use Fable to help me think about conceptual systems where thinking needs to happen few levels deep for any meaningful progress, it has exceptionally hard tendency to revert back to surface-level "midwit insights". If you ask it to analyze its mistakes, it does perfectly articulate that it went back to surface. But even with this analysis, it just can't help itself.
Once you get a turn where it's at the right level, it does offer genuinely good and helpful insights about your own thinking to spark further ideas, but I honestly can't remember Fable coming up with a single novel insight.
Soon, those 5,154 names will be a single model name.
The moment a single AI is capable of discoveries that took 5,000+ top human minds decades, the level of progress will be so fast that only other AIs will be able to work fast enough to productize the incoming tsunami of discoveries.
New discoveries are great but they need to be productized for humans to benefit.
14 years ago today, CERN announced the discovery of the Higgs boson.
Here are all 5,154 authors of the Higgs boson discovery paper - the scientific article with the largest number of contributors.
The very base layer of the internet is controlled by 12 companies, and 9 of them are under US law.
The internet was designed to withstand a nuclear attack, but it wasn't designed to withstand a political one.
The internet is not a long-term sustainable *global* network.