AIAcademy realm
Frontier Change
Current and near-future model capability, labs, releases, and fast-moving AI shifts.
106 chapters · 385 lessons · 372 structured sources. Source types include research paper, government research program, economics paper, government export-control document.
3 direct Practice Lab projects use Policy note, Evaluation sheet, Paper club brief.
Learner pathways
- I am new to AI
- I build with AI
- I need to understand risks
- I want frontier updates
Practice
- Write a frontier update brief · Policy note · 25 minutes
- Maintain a model capability watchlist · Evaluation sheet · 25 minutes
- Lead a scaffolded frontier paper club brief · Paper club brief · 30 minutes
Chapters
- Before transformers
AI Foundations · Where did AI come from? · 4 lessons - The transformer (2017)
AI Foundations · Where did AI come from? · 4 lessons - From GPT-3 to ChatGPT
AI Foundations · Where did AI come from? · 4 lessons - Five years in: 2026
AI Foundations · Where did AI come from? · 3 lessons - Tokens, context, prediction
AI Foundations · How does AI think? · 4 lessons - Training: the three steps
AI Foundations · How does AI think? · 4 lessons - Hallucination, sycophancy, confidence
AI Foundations · How does AI think? · 4 lessons - Reasoning models (the 2024-2026 leap)
AI Foundations · How does AI think? · 4 lessons - What models do well
AI Foundations · What can AI do, what can't it · 4 lessons - Where models struggle
AI Foundations · What can AI do, what can't it · 4 lessons - How we measure capability
AI Foundations · What can AI do, what can't it · 4 lessons - Reading model announcements skeptically
AI Foundations · What can AI do, what can't it · 3 lessons - Tools and function calls
AI Foundations · Reasoning, agents, and tools (2024-2026 leap) · 4 lessons - Agents
AI Foundations · Reasoning, agents, and tools (2024-2026 leap) · 5 lessons - Multimodality
AI Foundations · Reasoning, agents, and tools (2024-2026 leap) · 4 lessons - The current model lineup (2026)
AI Foundations · Where the field is right now · 5 lessons - What's hot in research right now (2026)
AI Foundations · Where the field is right now · 4 lessons - Open questions
AI Foundations · Where the field is right now · 3 lessons - The staged-release pattern
AI Foundations · The 2027 model lineup · 3 lessons - GPT-6 and continuous memory
AI Foundations · The 2027 model lineup · 4 lessons - Gemini 4 and the world-model bet
AI Foundations · The 2027 model lineup · 3 lessons - The open-weights split
AI Foundations · The 2027 model lineup · 4 lessons - The China frontier in ten labs
AI Foundations · The 2027 model lineup · 4 lessons - Why the old benchmarks died
AI Foundations · Benchmark literacy after MMLU · 3 lessons - The new spine
AI Foundations · Benchmark literacy after MMLU · 4 lessons - Reward hacking benchmarks
AI Foundations · Benchmark literacy after MMLU · 3 lessons - How to read a model card honestly
AI Foundations · Benchmark literacy after MMLU · 3 lessons - Reasoning models in production
Building with AI · Prompting & context engineering · 4 lessons - Open-weights training infrastructure
Building with AI · Fine-tuning & post-training · 3 lessons - The basic problem
AI Safety & Society · What 'alignment' means · 4 lessons - How alignment is currently done
AI Safety & Society · What 'alignment' means · 5 lessons - The Mind of the Model
AI Safety & Society · What 'alignment' means · 5 lessons - Honest mistakes
AI Safety & Society · Hallucination, deception, sycophancy, scheming · 3 lessons - When models tell us what we want to hear
AI Safety & Society · Hallucination, deception, sycophancy, scheming · 3 lessons - When models seem to deceive
AI Safety & Society · Hallucination, deception, sycophancy, scheming · 5 lessons - Why these matter for everyday use
AI Safety & Society · Hallucination, deception, sycophancy, scheming · 3 lessons - Persuasion as a capability
AI Safety & Society · AI manipulation & persuasion literacy · 3 lessons - Spotting manipulation
AI Safety & Society · AI manipulation & persuasion literacy · 4 lessons - Defense
AI Safety & Society · AI manipulation & persuasion literacy · 4 lessons - Responsible Scaling Policies
AI Safety & Society · Frontier-lab safety · 4 lessons - How labs evaluate themselves
AI Safety & Society · Frontier-lab safety · 5 lessons - Reading lab claims skeptically
AI Safety & Society · Frontier-lab safety · 4 lessons - What we know empirically
AI Safety & Society · Jobs, economy, society · 4 lessons - Tasks vs jobs
AI Safety & Society · Jobs, economy, society · 3 lessons - Education systems
AI Safety & Society · Jobs, economy, society · 4 lessons - Inequality and access
AI Safety & Society · Jobs, economy, society · 3 lessons - How AI gets governed
AI Safety & Society · Governance, openness, the open vs closed debate · 4 lessons - Open vs closed weights
AI Safety & Society · Governance, openness, the open vs closed debate · 5 lessons - Civil society and AI literacy
AI Safety & Society · Governance, openness, the open vs closed debate · 3 lessons - Forming your own view
AI Safety & Society · Governance, openness, the open vs closed debate · 3 lessons - Reward hacking in production
AI Safety & Society · When agents fail · 4 lessons - Memory poisoning and indirect prompt injection
AI Safety & Society · When agents fail · 3 lessons - Evaluation awareness
AI Safety & Society · When agents fail · 4 lessons - Multi-agent drift
AI Safety & Society · When agents fail · 3 lessons - Hallucinated case law
AI Safety & Society · The 2026 incident files · 3 lessons - The copyright trifecta
AI Safety & Society · The 2026 incident files · 4 lessons - The AISI rebrand cycle
AI Safety & Society · The 2026 incident files · 3 lessons - The preemption battle
AI Safety & Society · The 2026 incident files · 4 lessons - How reasoning models work
Frontiers · Reasoning models, deeply · 5 lessons - Math, science, and proof
Frontiers · Reasoning models, deeply · 5 lessons - Open questions in reasoning
Frontiers · Reasoning models, deeply · 3 lessons - Where interpretability is now
Frontiers · Mechanistic interpretability frontier · 3 lessons - Probing affect and persona
Frontiers · Mechanistic interpretability frontier · 4 lessons - Open questions in interpretability
Frontiers · Mechanistic interpretability frontier · 3 lessons - Where agents are now
Frontiers · Computer-using agents and long-horizon work · 4 lessons - Where they fail
Frontiers · Computer-using agents and long-horizon work · 3 lessons - Open questions in agents
Frontiers · Computer-using agents and long-horizon work · 3 lessons - Generation
Frontiers · Multimodal and world-model frontier · 3 lessons - World models and embodiment
Frontiers · Multimodal and world-model frontier · 5 lessons - Open questions in multimodal
Frontiers · Multimodal and world-model frontier · 3 lessons - Capability questions
Frontiers · Open questions, mid-2026 · 4 lessons - Safety questions
Frontiers · Open questions, mid-2026 · 4 lessons - Geopolitics and access
Frontiers · Open questions, mid-2026 · 4 lessons - Society and you
Frontiers · Open questions, mid-2026 · 3 lessons - METR Time Horizon 1.1
Frontiers · Long-horizon autonomy · 4 lessons - Orchestration patterns
Frontiers · Long-horizon autonomy · 4 lessons - Multi-agent systems
Frontiers · Long-horizon autonomy · 3 lessons - Humans in the loop
Frontiers · Long-horizon autonomy · 3 lessons - Genie 3 to Genie 4
Frontiers · World models and embodied AI · 4 lessons - Wayve, Tesla FSD, robotaxi commercialization
Frontiers · World models and embodied AI · 3 lessons - Humanoid foundation models
Frontiers · World models and embodied AI · 3 lessons - What physical AI means in 2027
Frontiers · World models and embodied AI · 3 lessons - GPUs as the dominant accelerator
The Physical Stack of AI · Chips and accelerators · 4 lessons - NVIDIA roadmap — Blackwell, Rubin, what next
The Physical Stack of AI · Chips and accelerators · 4 lessons - The challengers — AMD, TPU, Trainium
The Physical Stack of AI · Chips and accelerators · 4 lessons - Specialist inference chips
The Physical Stack of AI · Chips and accelerators · 3 lessons - How a chip gets made
The Physical Stack of AI · Fab and packaging supply chain · 3 lessons - The CoWoS bottleneck
The Physical Stack of AI · Fab and packaging supply chain · 3 lessons - HBM — the binding memory constraint
The Physical Stack of AI · Fab and packaging supply chain · 3 lessons - Chip geopolitics — export controls and diffusion
The Physical Stack of AI · Fab and packaging supply chain · 4 lessons - What a hyperscale AI datacenter looks like
The Physical Stack of AI · Datacenters and gigawatt campuses · 3 lessons - Stargate — the $500B project
The Physical Stack of AI · Datacenters and gigawatt campuses · 4 lessons - Hyperion, Colossus, Project Rainier, Stargate UAE
The Physical Stack of AI · Datacenters and gigawatt campuses · 4 lessons - Cooling, water, and local pushback
The Physical Stack of AI · Datacenters and gigawatt campuses · 3 lessons - AI energy demand projections
The Physical Stack of AI · Energy and grid · 3 lessons - The nuclear-AI nexus
The Physical Stack of AI · Energy and grid · 4 lessons - Grid limits and siting politics
The Physical Stack of AI · Energy and grid · 3 lessons - Power-purchase agreements explained
The Physical Stack of AI · Energy and grid · 3 lessons - Why every country wants its own compute
The Physical Stack of AI · Sovereign AI compute · 3 lessons - Europe — UK Isambard, France Jules Verne
The Physical Stack of AI · Sovereign AI compute · 3 lessons - Stargate UAE, Saudi PIF, IndiaAI Mission
The Physical Stack of AI · Sovereign AI compute · 4 lessons - China and chip sovereignty
The Physical Stack of AI · Sovereign AI compute · 3 lessons - Cost per million tokens — the unit economics
The Physical Stack of AI · The inference economy · 4 lessons - Why Opus dropped 67% in February 2026
The Physical Stack of AI · The inference economy · 3 lessons - Neoclouds vs hyperscalers
The Physical Stack of AI · The inference economy · 3 lessons - Intelligence too cheap to meter — fact vs reality
The Physical Stack of AI · The inference economy · 3 lessons