‘There’s no magic money tree’: Alex Depledge on the Budget, startup lobbying and why she left the UK government
THE SO WHAT
A former UK startup adviser walking away while stressing "no magic money tree" is a reminder that public support for tech is cyclical and politically constrained. Founders in the UK should plan as if grants and tax breaks will be slower and more conditional, and build funding plans that don’t assume a friendly Treasury.
READ THE SOURCE
MORE FROM THE WIRE
Anthropic mints record gains in record time
Anthropic’s IPO-era returns compress the venture timeline for frontier labs—capital is learning that model-scale bets can reprice in just a few years. For operators, this means faster iteration in the vendor landscape and more board pressure to show a credible AI posture tied to one of these newly liquid ecosystems.
Startups & VentureSuperMe, which generates AI-powered chats with simulated profiles of prominent industry experts to let users ask questions, raised $20M
SuperMe’s $20M for AI-simulated expert chats is another step toward ‘synthetic mentorship’—career advice decoupled from real networks. HR and L&D teams should assume employees will seek guidance outside company channels and decide whether to endorse, integrate, or explicitly guardrail these tools.
Startups & VentureSource: Dulo, a startup co-founded by Google veteran Sebastian Thrun to build "foundation models for hardware design", raised a $130M seed
A $130M seed for Dulo to build foundation models for hardware design is a clear bet that chip and board layout will follow software—iterated by AI, reviewed by humans. Hardware teams should start isolating design steps that could be handed to models and where verification and tooling need to harden first.
Startups & VentureCanada, Nvidia Back Cohere in Round at Over $20 Billion Value
A $20.5B valuation for Cohere, backed by both Canada and Nvidia, cements the model-maker as a strategic asset in the non–Big Tech LLM camp. Enterprise buyers now have a state- and chip-aligned alternative to hyperscaler-native models—leverage that in negotiations on data residency, pricing, and GPU access.