THE SWELL #28
SAIL Weekly Digest August 1–7, 2026 | Issue #28
The week opened with a shift in tone for mathematicians: AI is doing original work, and people are starting to notice. Kai Williams sat down with twenty mathematicians to talk about what it feels like to watch their field get eclipsed in real time.
Our other SAIL authors turned towards the practical side of that acceleration, looking at how to actually run agents day-to-day, keep serving costs manageable, and track the open-source models taking on a bigger share of the workload.
The Week in AI
Google centralized its AI leadership in Mountain View under Koray Kavukcuoglu to run research and operations, while DeepMind co-founder Demis Hassabis steps back from day-to-day management to serve as Google DeepMind Chairman and Alphabet’s Chief Scientist.
DeepMind’s AlphaProof and AlphaGeometry 2 hit silver-medal performance at the International Mathematical Olympiad, marking a landmark achievement in formal reasoning and Lean-checkable mathematical logic.
The frontier price war deepened after OpenAI slashed GPT-5.6 Luna input costs by 80% to $0.20 per million tokens, as CFO Sarah Friar confirmed OpenAI models now reach over one billion active users.
This Week from SAIL Authors
Managing the Machines
🔮 Seven lessons for managing AI agents — Azeem Azhar distills what running agents day-to-day actually requires, plus the 50-tool stack Exponential View uses. — Exponential View
The Inference Engineering Masterclass — Baseten’s Philip Kiely and Ali Taha on the engineering behind serving models fast and cheap, fresh off a $13B Series F. — Latent Space
FCC’s Adam Chan on the New Robot Rule — Jordan Schneider on import substitution as industrial policy for the robot supply chain. — ChinaTalk
Keeping Score on Open Models
Introducing our Artifacts Hub and Adoption Dashboard — Nathan Lambert on scaling Interconnects’ curation and measurement of the open ecosystem. — Interconnects
Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontier — Florian Brand and Nathan Lambert on the proliferating capacity to train strong models. — Interconnects
The Human Ledger
Mathematicians are grappling with the possibility that AI might eclipse them — Kai Williams interviews twenty mathematicians about rapid AI progress in their field. — Understanding AI
🔮 Leopold & exponential markets; transformative GLP-1s; runaway AI & the future of safety — Azeem Azhar on markets pricing the exponential, the runaway-AI safety debate, and AI agents as citizen rights. — Exponential View
📈 Data to start your week — Hannah Petrovic charts AI blurring job boundaries, GenAI entertainment rising, and El Niño shaking stability. — Exponential View
Full Library
FCC’s Adam Chan on the New Robot Rule — Jordan Schneider | ChinaTalk
🔮 Seven lessons for managing AI agents — Azeem Azhar | Exponential View
Mathematicians are grappling with the possibility that AI might eclipse them — Kai Williams | Understanding AI
The Inference Engineering Masterclass — Philip Kiely & Ali Taha, Baseten — Swyx | Latent Space
Introducing our Artifacts Hub and Adoption Dashboard — Nathan Lambert | Interconnects
📈 Data to start your week — Hannah Petrovic | Exponential View
Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontier — Florian Brand & Nathan Lambert | Interconnects
🔮 Leopold & exponential markets; transformative GLP-1s; runaway AI & the future of safety — Azeem Azhar | Exponential View
