Capital is chasing agents. October still has three clocks.

Capital is chasing agents this week, and the useful brief for agency principals is not the headline price — it is which three clocks you are measuring against. Suppose you walk into a client meeting with one chart: Polymarket's board for which company has the best AI model end of October, about $1.4 million in volume, resolving 31 October. Traders currently price Google near 64%, Anthropic near 37%, and OpenAI under 1%. A parallel board for the best Chinese AI company prices Xiaomi first, then Alibaba, with Moonshot and others further back. That is a crowd forecast, not a lab ranking — and the prices move. It is still the kind of number a client will ask you to explain before they ask you to ship.

What did the labs claim this week?

The week's lab catalyst is Google's own announcement. On 30 September, Koray Kavukcuoglu at Google DeepMind introduced Gemini 4 Argon as built for deep reasoning across long-horizon workflows — software engineering, legal and finance knowledge work, cybersecurity defence — with a phased Fairwind rollout before broader access. The same week, OpenAI said it disrupted a coordinated campaign designed to extract protected reasoning from its models, called the pattern adversarial distillation, and attributed a core cluster of activity to individuals associated with Moonshot AI. Xiaomi, first on the Chinese board, said MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index and claimed the strongest open-source slot available while still trailing the strongest closed models. Primary statements, different interests: read them as attributable claims, not as a settled hierarchy.

Where is the capital going while October settles?

Capital is not waiting for October's crown to settle. SoftBank Group said on 1 October it executed the third and final $10 billion tranche of its OpenAI follow-on via Vision Fund 2, completing $30 billion of follow-ons and taking cumulative investment to $64.6 billion with roughly 13% ownership. Further down the stack — and closer to brand work — OuterSignal raised a $22 million Series A for agentic one-to-one personalisation across consumer brands. Photon, building agents over iMessage and WhatsApp, closed a $4.5 million seed; its CEO told TechCrunch the shift from apps to agents is directionally inevitable. The money is already arguing that the channel you sell into will be negotiated by agents as much as by apps.

Which three clocks should agencies brief against?

For agency and brand principals, three clocks are running at once. The first is which frontier model clients will ask for by month-end. The second is which Chinese open-weight stack shows up in cost-sensitive or Asia-facing briefs. The third is which agentic vendors — personalisation, messaging, search visibility — VCs from the US, China and Europe are already funding into the channel. Fluency is already on the shelf. What is not on the shelf is a governed answer to who may mutate spend, copy or creative without a named human, who may refuse it, and whether "so what" still has an owner when an agent posts at seven in the morning.

What are the limits of this cut?

Limitation: Polymarket is a crowd forecast, not an evaluation. Argon is not yet broadly available. Moonshot had not immediately commented in reporting checked for this draft. SoftBank's OpenAI stake is ownership of a lab, not a product recommendation. Odds above were rechecked live as of 2 October 2026 — Google near 64%, Anthropic near 37%, OpenAI under 1%, about $1.4 million volume on the model crown board; Xiaomi still leads the Chinese board. Recheck both Polymarket links before you brief a client.