Every published a hands-on review of GPT-6 Astra, OpenAI's new model. The reviewers found it strong at writing, operating software and visual design. They rated Anthropic's Fable better at the instincts needed to build a product. Astra wrote the first draft of the review from a single prompt. Every's cofounder Dan Shipper did not realise a person had not written it. The full set of tests and the final verdict sit behind a paid membership. Every also scheduled a live session on Fable 5.1 and GPT-6 Astra for 4 September, open to paid members. The assessment is Every's own and is not a controlled benchmark.