The context · August 2025
OpenAI introduced GPT-5 for developers in August 2025, presenting a new option for coding, reasoning and tool-using applications. [1]
A proposed project brief inspired by the period’s developments. This is an idea for exploration, not an announced Sansa product, an approved roadmap or a claim of completed client work.
Capture the feature’s actual job
Choose one established AI feature and collect approved examples representing its normal use. Add examples of ambiguity, missing evidence and unsupported requests. Define the answer format and the behaviours that would make an output unacceptable.
Run a paired evaluation
Compare the current and candidate configurations using the same inputs and reference material. Separate differences caused by the model from differences caused by prompts or retrieval. Record response time, total request cost and the amount of human correction required.
Deliver a release decision
This project concept would produce an evaluation report and a documented rollout gate. It should allow three outcomes: adopt, continue testing or stay with the current model. Keep the old configuration deployable until the new one has demonstrated acceptable behaviour in the application’s real operating conditions.
Source & context
OpenAI · GPT-5 for developers, 7 August 2025This retrospective was written for the archive in September 2026. The linked primary source documents the announcement or event; the practical interpretation and proposed approach are Sansa’s editorial perspective. Public examples do not imply a client relationship. Product capabilities and guidance may have changed since the period discussed.
Another perspective · August 2025
A new model release is a reason to test, not a migration plan
Continue readingWorking through a similar question?
Talk it through with Sansa