Good morning from the wAIve team in London.

Today, Thursday 5 November 2026: A day for counting the cost of AI. Apple has quietly cut what its own engineers can spend on AI each day, and MPs say Britain is dangerously dependent on China for the minerals AI is built from. Meanwhile, Meta and OpenAI have both gone shopping for the tools that let whole teams of coding agents get on with the job.

  LATEST  

Apple is significantly reducing the budgets it allocates to AI usage by its own engineering teams, according to reporting published overnight. The cuts affect internal tooling and AI-assisted work across the company. This is not a public-facing change of direction (Apple is still shipping AI features with products), but it does show how even the world's most valuable company is managing the cost of using AI at scale.

THE PLAY: If Apple is rationing internal AI spend, every business deploying AI tools across engineering or operations teams should be asking questions about token use, costs, and value.

  MIDTERMS  

With a new Texas governor-elect, Gina Hinojosa, there are questions around two major data facilities. Texas is the fastest-growing data centre market in the United States and home to some of the largest AI training facilities under development. State permit approvals had stalled before results came in. And of the 41 local ballot measures on data centre development across the country, nearly all voted to restrict construction.

THE PLAY: Any company with planned AI infrastructure spend in Texas now faces uncertainty. Also, expect fresh scrutiny in Congress for AI generally.

  POLICY  

A parliamentary investigation published today found that the United Kingdom is "dangerously" reliant on China for the critical raw materials required for military systems, renewable energy infrastructure, and AI hardware. The Business Select Committee called for the immediate creation of a critical minerals finance facility to fund rare earth mines, deploy grants and equity, and accelerate diversification. The specific minerals at risk include those used in semiconductors, magnets for data centre cooling systems, and advanced chips.

THE PLAY: Ask your tech procurement lead to identify AI hardware that is dependent on Chinese supply chains.

Meta has acquired Conductor, an AI coding startup whose software allows engineering teams to deploy multiple AI agents simultaneously across different aspects of a software project.

Why it matters: The deal signals Meta's intent to accelerate agentic software development across its own engineering organisation, which numbers in the tens of thousands.

OpenAI has acquired Pierre Computer Company, whose libraries already power OpenAI's own Codex coding agent.

Why it matters: Combined with OpenAI's recently launched Git backend, Pando, the acquisition signals a deliberate strategy to build out the entire software development infrastructure stack for AI agents working alongside human developers.

During its busiest week last quarter, Uber reported that only four customers booked a ride using an AI agent. The company made the disclosure to illustrate a wider point: despite enormous industry investment in agentic AI for commerce, real consumer adoption is negligible, for cab bookings at least.

Why it matters: The commercial case for AI agents isn't yet playing out at scale. Agentic commerce is coming, but the Uber numbers tell an interesting story about adoption!

Closing prices, Wednesday 4 November (US markets). Not investment advice.

A privacy-first search engine that lets users search the web without being profiled, with AI features that are optional rather than mandatory.

THE PLAY: Set DuckDuckGo as the default browser search on a device you use for sensitive competitive research this week, and compare the results quality against your current search tool.

ElevenLabs' new voice agent model, built to reason and process thoughts while actively talking.

THE PLAY: Lets you build ultra-low latency, conversational voice agents that dynamically pause, listen, and respond with human-like emotion in real time.

When one of the world's largest ride-sharing platforms, a company that processes tens of millions of bookings each week, reports that only four customers used an AI agent to complete a transaction during its busiest period last quarter, we should probably pay attention.

Businesses across every sector are diving headfirst into agentic AI: software that can browse, decide, and act autonomously on behalf of customers or employees. The marketing pitch is compelling: reduce friction, automate transactions, free up human attention for higher-value work. Billions are flowing into infrastructure, frameworks, and product development.

But the consumer data tells a different story. AI agents struggle with ambiguity, and the kind of messy real-world context that human judgment can deal with. The gap between what an agent can do in a demo and what a customer will trust it to do on their behalf is still wide.

First, any business case for agentic AI without empirical data to back it up is a speculative play. That's a given.

Second, immediate value from AI agents is likely to be internal, for now, not external. Agents automating back-office processes, compliance checks, or document workflows are delivering measurable returns now.

Third, don't read too much into this. Just because the Uber/AI agent proof of concept isn't there yet doesn't mean that it won't be months or weeks from now. If we've learned anything in the last year, it's that the cycle moves very fast.

The question your business should be asking is not "how do we deploy AI agents?" It's how can we develop agents that are so impressive that the customer experience becomes the story. If Uber customers aren't using AI, I'm interested to know why or what Uber's next move might be.

Most executives ask AI to write reports, memos, or briefings without specifying who will read them, so the output is generically correct and specifically useless. The audience anchor technique fixes this by forcing the model to adjust tone, depth, and emphasis for a named decision-maker with known priorities.

It works because specificity collapses the model's range of possible outputs from a broad distribution to a narrow, targeted one: the more precise your constraints, the more useful the result.

Thanks for reading, the wAIve Daily Team. Found this useful? Forward it to someone who needs it. Tell us what you need to know. Reply with your questions.

wAIve Daily is published with Centai Strategy