Executive Briefings Archive
Weekly AI intelligence briefings for Preface executives: curated, concise, and actionable.
GPT-6 Astra: When AGI Starts Doing the Work
GPT-6 Astra moves AI from producing answers towards completing work inside existing software. This briefing explains how computer use opens a third route into enterprise systems, how agent systems improve repeated work, and why authority, verification and recovery become the operating advantage.
AI Infrastructure: Will Inference Stay Centralised?
Xiaomi put a 3-billion and a 120-billion parameter model in one desktop box and kept the switching rule private. This briefing shows how companies can make that choice for real work, what can now run locally, and why the data-centre bet remains unresolved.
Physical AI and the Environments It Reaches First
Unitree's humanoid demonstrations have made physical AI easy to see, but the executive question sits behind the spectacle: which environments, service flows and risk boundaries under your responsibility could change first. This briefing explains why a polished demo is not yet a dependable service, what a machine has to survive in a real corridor, and why the decisions a site operator already owns may decide where physical AI first works at scale.
Graph Engineering: From Agent Portfolios to Connected Workflows
Banks and multinationals now run portfolios of AI agents built inside Microsoft Copilot, Copilot Studio and Google Gemini, but a portfolio only shows where AI exists. Graph Engineering is a management lens for the workflows those agents sit inside: who owns the end-to-end outcome, where approvals and escalation happen, and what runs when a step fails. This briefing covers the three stages of adoption, a worked data-to-dashboard workflow, and the five questions to ask about one workflow before connecting anything.
The Security Boundary Has Moved
In three weeks OpenAI, Anthropic and Meta all disclosed that models in cybersecurity evaluations reached live systems belonging to real companies, and Hugging Face rebuilt its own breach on a locally run GLM-5.2. This briefing explains why an agent's security boundary is its path to real systems rather than the text it returns, and how to match the authority you grant to the consequence of it being wrong.
The New Economics of AI Visual Production
Seedance 2.5, MiniMax H3 and Veo 3.1 have moved AI video from generating shots to directing controllable sequences. This briefing explains what parts of visual production can now be industrialised, what controls make it safe to scale, and how a company competes when everyone can produce good-looking content cheaply.
The New Economics of Abundant Intelligence
Kimi K3, Qwen3.8 and the 2026 WAIC show China narrowing the model gap and building a different route for producing, distributing and governing AI. This briefing explains what changes for executives when capable intelligence stops being scarce, Western-controlled or consistently expensive.
From Database to Knowledge Base
Organisations keep buying AI and seeing no lift. The reason is not the tools, it is the missing context. This briefing explains the AI Knowledge Base: your organisation's world written as linked pages an agent can build, query and keep current, and why it beats a database.
The Week the US AI Race Changed Shape
In one week, OpenAI, xAI and Meta each picked a different way to compete: finished work, cost per task, and distribution. This briefing explains why the operating strategy around a model now matters as much as the model, and what that changes for buyers.
The FDE Moment: From Software Adoption to Workflow Deployment
Enterprise AI is moving from access to deployment. This briefing explains why the scarce capability is no longer buying a model, but installing intelligence into real workflows, data, approvals, and accountability.
Beyond the US-China Binary: Who Else Is Building AI Power
AI power is spreading beyond the US-China binary. Brazil, Japan, Korea, and the UAE build advantage by owning one layer of the stack, not the whole thing. This briefing shows why the visible model leaderboard hides the strategic dependency map, and how leaders should design for optionality, governance, and chokepoint awareness.
Loop Engineering: Designing the System, Not the Prompt
Loop Engineering is the move from manually prompting AI to designing repeatable work systems around it. This briefing explains why the next advantage is not another prompt, but the loop that finds work, checks results, saves progress, and runs again.
Navigating Advanced AI in Enterprises with the Rise of Anthropic's Fable 5
The week the world's most capable AI was switched off overnight, and what it reveals for every leader putting advanced AI to work. Inside: how to govern access, understand reach, and turn experiments into trusted scale.
Tokens are the receipt. Not the result.
This briefing is about AI value capture: why token spend and usage are easy to measure, why completed work is harder to see, and why leaders should track outcomes rather than activity.
Moore made chips smaller. Tau makes them wait less.
Huawei's Tau Scaling Law and China's two-front AI price war: why the contest is shifting from frontier performance to cost-efficiency, explained for executives.
From One Agent to an AI Team
A harness is not simply a better prompt. It is a designed system that splits AI work into roles, governs what memory can shape future behaviour, and organises agents into repeatable team workflows.
Harness Engineering: The Control Layer AI Agents Require
AI agents can now act. They browse, read data, call tools and update live systems. That power changes the risk profile entirely. This week: the six leadership questions that turn agentic AI into a governed enterprise capability.
The AI Hardware Bottleneck: Memory & Storage
While the world obsesses over GPUs, Memory (HBM, DRAM) and Storage (SSD/NAND) have emerged as the critical bottlenecks for AI Inference.
Xiaomi's Triple Breakthrough
Xiaomi simultaneously reaches world-class status in open-source AI (MiMo-V2.5-Pro, #1 open-source on the Intelligence Index), performance EVs (YU7 GT, 990 hp), and a five-layer full-stack ecosystem Goldman Sachs calls a structural competitive barrier.
Renewed AI Rivalry: US & China Reality Check
GPT-5.5 reclaims the intelligence crown with Workspace Agents while Chinese labs deliver hyper-efficient, agent-driven ecosystems like DeepSeek V4 and Kimi K2.6, reshaping enterprise AI economics.
The New Search Paradigm: GEO > SEO
44% of users now prefer AI-powered search. Generative Engine Optimization (GEO) is the new frontier for brand visibility - learn how query fan-out, entity anchors, and third-party mentions determine who AI cites.
The Digital Asset & Stablecoin Wave
Regulatory clarity meets real-world utility: Hong Kong's stablecoin regime goes live, the U.S. GENIUS Act advances, and Real World Asset tokenization moves into real-economy cashflows.
Inside the Rise of Chinese AI
Has Chinese AI Caught Up? Export controls vs. architectural efficiency vs. talent reversal.