From vibe coding to vibe capitalism

Bolt has been a pleasure this week. I wrote and implemented an internal tool in hours — our CTO dropped it on a sub-domain and voilà: a Saturday-morning, post-breakfast noodle deployed to the team by Monday. This thing had been on my backlog for a year, and I built it in four hours with no coding knowledge, using the front-end of our main site so it could be adapted, absorbed or iterated later.

This was one of those better-to-test-in-production ideas — like so much in scaling-up, where it's hard to justify the calories when the RoI is contentious. The ability to go from idea to production at lightning speed without dev resource is the thing underpinning the momentum in valuations and growth in this market. It's a combination of scary, exhilarating, dumbfounding and disorienting.

From vibe coding to vibe marketing

The fastest-scaling companies are yet to come — the ones who'll embrace tooling and risk, who reach good-enough/excellent and are ready to point and shoot. Unlike Replit and Bolt leading the 'anyone-can-code' space, there are no clear marketing winners yet. Adobe introduced its Experience Platform Agent Orchestrator with ten built-in AI agents, tackling audience targeting, content creation and site optimisation — customising content whether a visitor arrives via a TikTok ad or a search query.

HubSpot built Breeze on the data-enrichment side; it needs a beefy enterprise upgrade, but next to Salesforce, HubSpot is the marketing platform — if you integrate martech, you integrate into HubSpot. Its VP of Ecosystems Scott Brinker notes the industry now relies on over 14,000 tools, a 27% global growth driven by AI martech. I've also spent time with Clay and the Clay community in the UK; it raised $40M at $1.25B in January and just announced a community follow-on round. The B2B/GTM services market is so enabled by this type of tool — what it can do is somewhat magical.

GTC 2025

Last week was Nvidia's GTC 2025, a big deal in the AI event space (shout-out to the video delivery by RainFocus and CaptionHub partner Kaltura). It showcased record attendance and a slew of announcements, from new chips to personal AI 'supercomputers'. Jensen Huang pushed back on competition and tariff concerns, insisting demand for high-performance hardware remains robust — but with major players exploring in-house chips and competitors like DeepSeek offering efficient alternatives, Nvidia faces mounting challenges. Its moves into quantum computing and diversified edge AI signal a strategy to secure future growth.

Two talks stood out for the autonomous-driving geeks. Wayve's CEO Alex Kendall outlined a low-cost, hardware-agnostic approach that integrates with existing sensors and doesn't rely on HD maps — end-to-end data-driven learning using real-world and synthetic data, aiming at a scalable ADAS path to Level 4 autonomy. And Rivian founder RJ Scaringe spoke on cars, AI, scale and supply chain. These are the modern-day future builders.

Market moves

• Perplexity is in talks to raise $1 billion at an $18 billion valuation, with ARR reaching $100 million and an agentic browser, Comet, on the way.
• Disney, with Nvidia and Google DeepMind, unveiled a physics engine named Newton for lifelike robotic characters bound for its parks.
• Nvidia announced the Blackwell Ultra processor, DGX Spark and DGX Station desktop AI supercomputers, and the Dynamo software platform.
• The UK's Alan Turing Institute is democratising weather forecasting with AI models called Aardvark, running on desktop computers rather than supercomputers.

A few more from the macro file: Meta AI finally launched in the EU, albeit with a restricted feature set under GDPR; Apple replaced John Giannandrea with Mike Rockwell to lead Siri's AI; SoftBank agreed to acquire chip designer Ampere Computing for $6.5 billion, deepening its AI-infrastructure focus; the Bank of England held rates at 4.5%; and the OECD's interim outlook saw global growth stabilising amid geopolitical and supply-chain strains.

Land Rovers and public ownership

On public ownership — something I've been writing about lately. This weekend I learned that Land Rover, during the late-'60s turmoil that hit most UK automotive, was brought under public ownership. One British Leyland rule was that Land Rover deliver two new models a year; in its first year of operation, 1970, one of those was the brand-new Range Rover, which did surprisingly well. So it's not all binary and bad when it comes to public ownership. Till next week.