Its not a secret that I don’t like Anthropic.
But from time to time they do really come up with something that wow’d me.
Fable 5 is one of them!
For a few days up until 7th July, Anthropic let their subscribers use the model - so I upgraded my subscription back to Max just for this.
I did that yesterday, and fully used up all my allocations within a day.
It was amazing.
Especially communication-wise.
Having to deal with Codex’ quirks for so long (for those who are not in the know, it feels like talking to a 140 IQ autistic genius child programmer), it was a breath of fresh air to say the least.
It just seems to “get it.”
It didn’t give any of the usual platitudes AI models normally give.
Running it through NLC, it understood my context perfectly and picked it up just like that.
And throughout my interactions, I find myself even coming to....respect it.
Its very easy to anthropomorphize Fable 5.
So I get what Anthropic is doing. Even if I don’t like the way they do it.
What I got done with my time with Fable was something other models struggled with.
.
#1: Graphically Articulate NLC
Design was one of the biggest issues I had with Codex, even though I much prefer Codex to even Opus when it comes to coding.
I wrote an offer doc about NLC, but its a lot of words.
I wanted to turn this into a website, but I was really unsatisfied with how Codex and Opus does it.
Its not that they’re not “nice” - they were not “aligned” with what I want.
Fable with NLC’s design skills made this for me.
Not done, but it’s finally what I feel and resonate with.
.
#2: Consolidating Skills
NLC has 250+ skills installed.
Sounds impressive - but it’s unnecessary bloat.
There were redundant skills that does similar things in different ways - so a coin toss decides which skill is being used.
The problem is solved when the user specifies which skill to use, but I don’t want my user to have to specify which skill to use.
NLC is supposed to be brain-dead simple.
Together, we consolidated 250+ skills down to just 150+, taking the best of everything and combining it cohesively so that skill routing is no longer a coin toss.
We also made a workflow that mechanically forces future agents to either consolidate or supercede older skills whenever they are introducing something new, whether installing a new skill from online, or creating one of our own.
Which means - in the same way a McDonald’s staff don’t need to have cooking skills to deliver consistency, NLC agents don’t need to be powered by Fable to deliver consistency.
.
It helped tremendously to move the needle for the completion of NLC2, as well.
Aside from that, I fixed a bunch of other very annoying and complicated bugs in NLC that I didn’t have time to guide agents for.
I don’t think I’ll be using Fable after it’s no longer offered in subscription, because API-Fable is crazy expensive.
I once spent >$200 USD in API costs from using Opus within 1 hour.
Fable is double the price of Opus.
So no, I won’t be using Fable through API, but it was really good while it lasted.
How did you use Fable?
Hit reply and let me know - I read every single reply.
FELIX
P.S. I am hosting a get together for NLC users this Friday to build stuff together over the course of 4 hours.
8pm to 12am SGT.
You’ll get the benefit of coming in, work with the gang on ONE build, and leaving with one thing getting done.
Want details?