AIAcademy update cohort · current

2026 Q2 frontier refresh

The first public AIAcademy update cohort tracks frontier-lab source refreshes, product-created learning surfaces, and the lessons that need review before the next quarterly refresh.

8 updates · 5 source-linked · 3 product-created

Key questions

• Which frontier-lab claims moved into the monthly review lane?
• Which learner-facing surfaces were created during this cohort?
• Which lessons should authors revisit before the next cohort ships?

Updates

• GPT-5.2 system-card update enters the monthly review lane
  Track 6 frontier lessons should treat GPT-5.2 claims as current-refresh material and separate measured system-card behavior from launch commentary.
• Clinical AI guidance gets a sharper professional-boundary source
  Health-adjacent examples should route learners toward clinician review and avoid presenting model answers as diagnosis or treatment decisions.
• Knowledge-work adoption sources move work lessons from novelty to workflow design
  Work lessons should emphasize review ownership, delegation boundaries, and measurable workflow fit rather than generic productivity claims.
• Enterprise model-card reading becomes a builder skill
  Builder lessons can use Command A as a current example for reading context, deployment, retrieval, and evaluation claims before adopting a model.
• Social-platform AI usage needs separate treatment from classroom or workplace usage
  Safety and family lessons should distinguish public social AI behavior from private assistant use, especially around persuasion, identity, and crowd dynamics.
• Practice Lab first pass is live
  Each track now has one real-world project and each learner pathway has a capstone artifact, with templates saved into existing lesson artifacts.
• Chip now exposes local grounding references
  Tutor answers can carry source chips from the local curriculum, library, resources, and source graph; quiz and high-stakes guardrail modes are deterministic.
• Auth, tutor, and league privacy hardening shipped locally
  Auth and tutor now use D1-backed rate-limit buckets, and weekly league visibility requires synced learner opt-in.