Train-Time Scaling: How AI Learns from Its Own Reasoning — Stanford CS329A Part 6
Stanford CS329A Part 6 connects STaR, DeepSeekMath/GRPO, and DAPO to show how verified reasoning can be turned into persistent train-time improvement.
AI & Technology
Capability Is Not Reliability — Stanford CS329A Part 5 Agent Evaluation
Stanford CS329A Part 5 compares METR, GDPval, and DeepScholarBench to show why agent capability, reliability, context, and real-world deliverable quality must be evaluated separately.
Where Should an AI Agent Get Its Feedback? — Stanford CS329A Part 4
Stanford CS329A Part 4 connects ReAct, execution feedback, and Constitutional AI to one question: where should a self-improving agent get corrective signals it can actually trust?
When the Verifier Can Be Wrong — Stanford CS329A Part 3
Stanford CS329A Part 3 traces robust verification from outcome verifiers and process reward models to Math-Shepherd, weak-verifier ensembles, and verifier distillation.
Test-Time Compute Is a Resource Allocation Problem — Stanford CS329A Part 2
Stanford CS329A Part 2 shows why inference scaling is not just about generating more samples, but about allocating compute across search, verification, revision, fusion, and architecture design.
Why AI Progress Is Moving Beyond Bigger Models — Stanford CS329A Part 1
Stanford CS329A Part 1 traces the shift from pretraining scale to chain-of-thought, post-training, inference-time compute, verifiers, and agentic feedback loops.
Self-Improving AI Agents Aren’t One Model. They’re a Feedback Loop.
Stanford’s CS329A shows why self-improving AI is less about one smarter model than a closed loop of search, verification, tools, learning, and long-horizon evaluation.
Elon Musk’s 2022 Future Vision, Revisited in 2026
In 2022, Elon Musk linked autonomy, humanoid robots, brain-computer interfaces and Starship into one optimistic future. Four years later, their uneven progress shows why vision, technical direction and timelines should be judged separately.
What Elon Musk’s 2017 TED Talk Reveals About a Future Worth Building
A 2017 TED conversation connects tunnels, electric cars, solar energy, reusable rockets and Mars through one larger idea: progress should solve real problems and make the future feel worth anticipating.
AI Data Centers Need More Power. The Harder Question Is Who Pays for It
AI data centers are forcing utilities and regulators to solve two problems at once: how to add power quickly, and how to keep the cost and risk of new infrastructure from landing on ordinary ratepayers.