We Tested Fable 5.2 Early. It’s a BEAST.
Categories: Startup, Product, AI
Summary
Claude Fable 5.2 is a "spiky" paradigm shift—S-tier for coding and knowledge work but B-tier for writing—that enables solo builders to create professional software (like a full video editor) in weeks for thousands of dollars instead of millions, fundamentally reshaping how creative tools get built.
Key Takeaways
- Fable 5.2 enables end-to-end professional software creation through prompting alone. One engineer built EveryCut (full video editor with Premiere + Descript features) in weeks, replacing what would normally cost millions in development investment.
- The model uses an 'overnight model' workflow—set complex tasks to run asynchronously in the cloud while you work elsewhere, with estimated completion times provided. This is fundamentally different from iterative laptop-based development.
- Fable 5.2 shows extreme task-specific variance (S-tier coding, A-tier design, B-tier writing). The real value comes from identifying which specific workflows match the model's strengths rather than treating it as a general replacement.
- Professional software creation democratization threatens existing vendors. If serious creatives can now build custom tools matching their exact workflows, companies like Adobe must reconsider the value proposition of traditional software platforms.
- Real-world testing methodology matters more than benchmark scores. Every's frontier team tests on actual work (coding, writing, design, knowledge work) rather than public benchmarks, providing day-zero actionable intelligence on model capabilities.
Related topics
Transcript Excerpt
best measure of a new model is whether it causes you to try things that you never would have tried before. And by that measure, Claude Fable 5.2 is a true paradigm shift of a model. kind of nuts, but it's a spiky kind of nuts. Like there's a certain kind of task that it's going to absolutely blow your mind and be unlike anything else you've ever tried before. And for everything else, it's just sort of normal. Like you probably actually just should use Opus 5.5 So the model just dropped. You're probably wondering how is this review already out? And that's because at Every we have a frontier team whose job it is to get all the models before they come out and test them rigorously on all of our work. not just testing them on big public benchmarks or flashy X demos. We're actually testing them …