Keep pulling the thread on Mike Cannon-Brookes.
Mike Cannon-Brookes describes the current AI funding ecosystem as a circular flow where a VC gives $100 million to a model company, which pays $150 million to a cloud provider, who then pays $200 million to NVIDIA, with each entity losing money.
Atlassian's AI strategy is to support multiple competing foundational models rather than building its own, as the company believes it is unlikely to be able to compete in training proprietary models.
Mike Cannon-Brookes and Matt Garman of AWS both predict their respective companies will employ more software engineers in five years than they do today.
Mike Cannon-Brookes predicts that in 10 years, the current three-to-five-year period will be remembered as a time of "complete up and down turmoil" in the technology industry.
Mike Cannon-Brookes believes that most AI-related companies are currently "vastly overvalued."
Mike Cannon-Brookes believes that in the AI era, good design will become a more valuable and scarce resource that is hard to copy, especially as the cost to create software decreases.
Atlassian is launching a "vibe coding" environment to enable its community of app vendors and creators to build applications on its platform.
Atlassian engineers use a variety of AI coding tools, including the company's own Rovo Dev, as well as Cursor and GitHub Copilot.
Mike Cannon-Brookes states he would not pay 10 times the current price for AI coding assistants because the low switching costs and high competition between providers would prevent such a price increase.
Mike Cannon-Brookes believes that the widespread, simultaneous global availability of large language models is a unique event in technology history, unlike the slower rollouts of PCs or mobile phones.
The classic competitive dynamic in technology is whether a startup can acquire distribution before an incumbent acquires innovation.
Mike Cannon-Brookes believes it is too early to determine the margin profiles of AI businesses because their monetization models are still changing and the rate of model value depreciation is unknown.