Claude Fable 5? Simply unbelievable.
I tried Claude Fable 5 over the last few days, taking advantage of the fact that it is included in the subscription until June 22.
My main impressions so far:
- Quality: Unbelievable.
- Speed: Acceptable by today’s AI standards. Compared to a human, of course, unbelievable again.
- Usage / consumption / cost: Also unbelievable, unfortunately. Hopefully not for long. 🙂
I even purchased $50 of extra usage credits just to see how far it goes.
With Fable 5 and Effort=Ultracode, I consumed those $50 in a few minutes, basically on one relatively complex task, which included:
• Analysing a bit of what needs to be done
• Creating a Jira task
• Creating the branch locally
• Making changes across around four files, including tests
• Testing locally
• Committing and pushing
• Creating the PR
- Overall impression: Extremely satisfied.
- Personal feeling: Happy to see the achievements in the field. At the same time, also confused, and honestly a bit concerned about what the future looks like for people whose work happens mostly on a computer.
- Will it affect me personally?
Of course. Affecting me today already. It will affect me much more very soon - probably in ways which I cannot predict today.
Knowing the basic concept of LLMs from Machine Learning days, I am honestly positively surprised by how far the field has come.
Exciting times.