Models, harnesses and the next year¶
Separate observed performance from product choices and forecasts. This draft does not establish a universal model ranking.
Working notes¶
- The sessions show complementary Claude and Codex roles, with each catching mistakes the other missed.
- Matched briefs and sandbox access changed apparent comparison outcomes.
- The corpus does not directly evaluate open models or open-source harnesses.
- Provider-hosted and self-managed cloud workflows also lack direct evidence here.
Source: course evidence synthesis, §7–8. These are selected observations, not general performance guarantees.
Subscriptions and 2027¶
- [TODO: verify subscriptions, employer support and six-hour usage limits before the course.]
- [TODO: substantiate the opinion that competition is improving open models; “capitalism is doing its job” is a framing, not a benchmark.]
- [TODO: distinguish demonstrated 2026 capabilities, 2027 forecasts and work that still needs domain expertise.]
Before the course¶
- [TODO: select a compact worked example and its verification artifact.]
- [TODO: rehearse the live walkthrough and prepare a fallback.]