Content at the speed of culture.
Chinese AI startup DeepSeek's R1 model offers ChatGPT-like capabilities, despite claiming to have spent only $5.6 million on computing (compared to the billions spent by OpenAI, Google, Meta etc.)
This alleged breakthrough has triggered a sharp selloff in U.S. tech stocks, with NVIDIA losing nearly USD$600 billion in market value—the largest single-day loss in history. I can't see how or why we're seeing such a market-moving effect without any real transparency around the true cost of R1's development and maintenance, given at least 15% of NVIDIA'S revenue comes from Singapore (driving distance to China,) the clear security risk to enterprise and the global pressure Trump's announcement of Stargate has injected into the AI race.
However, it's undoubtedly a 'Sputnik moment' (great call by Marc Andreesen) that has been making everyone in AI bash their head (including my gigantic one) against the idea of ‘defensibility’ over the weekend.
This is likely the beginning of foundational model commodification: they are becoming cheaper, more open sourced and infinitely more expansive in choice. When everyone has the same powerful models in their pocket, the competitive advantage is domain expertise and experience. How can you understand user problems deeply enough that you can build THE interface users rely on to solve these problems with AI.
Technology advantage is temporary, with foundational LLMs narrowly competing on benchmarks each day. Interface loyalty is forever. Look at Google - they have maintained the exact same UI for 28 years despite insane product development for a reason. Look at Chat.GPT - its chat-based UX and ecosystem of plugins have made it the public interface to generative AI?
We’ve seen this before. In 2006, AWS made cloud storage accessible on a pay-per-usage basis at a fraction of the cost, disrupting an industry convinced that the real picks & shovels $$$ was in selling servers. As infrastructure became commodified and consolidated into a handful of incumbents, trillions of dollars have been and continue to be made at the application layer. No one calls Spotify an ‘AWS wrapper’; its value to users is too great. And this value isn’t because its cloud computing service of choice is superior - it’s because it has become our interface to daily listening, playlists and annual Spotify Wrapped.
In short, GPT ‘wrappers’ such as Perplexity and Brand Ninja on the application layer are looking highly competitive in a more evened playing field. The questions are: who will garner the UX, the feedback loop, the network effect, and the distribution to build the Spotify for law, the Slack for content coordination, or the Netflix for accounting? These companies will be exceptionally capitalized, ruthlessly focused on their specific customers, and led by a team of both exceptional engineers and domain experts (or at least highly capable first-principle thinkers with excellent listening skills).
They will design the UX specifically for their niche, integrating all necessary plugins, tools, and skills. They will build a marketplace where domain experts can share their successes on the platform, making it increasingly useful for others. They will create an information moat: the more time users spend on the platform, the better results it will drive for them. All this will be gamified and delightful.
Additionally, they will have a strong enterprise sales department, recognising that no matter how the wild west frontier evolves, medium-to-large businesses will always value people who can distill the chaos into hand-held implementation with clear outcome, security, SLAs, and relationship-led account management. They will rightly invest greatly into community and brand. Most importantly, they will remain completely LLM-agnostic, ensuring their users consistently benefit from the latest advancements in technology.
I couldn’t be more excited about where Brand Ninja is today, and it feels as though DeepSeek has only sharpened our competitive edge as a truly industry specific but LLM agnostic platform. Let’s build!
