I think very shortly the conversation will switch to running inference and training for US-based workloads outside of the US. The same way the mining industry was knee-capped.
Training is a very easy thing to move out, since latency doesn't matter, but increasingly for inference (heavy reasoning-based workflows) where the applications are not very latency sensitive, we'll see workloads move internationally.
Countries that'll do really well: abundant energy, easy permitting, proximity to the US.
Chris and I left Stripe in the last few months to work together on a new company - CoPlane.
We’re building a small, world class founding engineering team and seeking agentic and curious problem solvers, excited about working at the nexus of data and finance.
If that sounds like you, please dm!
Chris and I left Stripe in the last few months to work together on a new company - CoPlane.
We’re building a small, world class founding engineering team and seeking agentic and curious problem solvers, excited about working at the nexus of data and finance.
If that sounds like you, please dm!
Advice I gave a founder earlier this week which generalizes, I think:
If you’re narrating a go-to-market approach for an AI startup where the business gets interesting in N years, you’re implicitly short AI capabilities progress over N years. Do you really want to make that bet.
Speech input interfaces are going to become much more prevalent moving forward for two main reasons: transformers have drastically improved robustness for speech recognition (across many languages), and LLMs are enabling natural language interfaces across many product use cases.
Too many consumer facing tech-enabled startups built platforms, but underestimated the engineering talent to support them through phases of growth and maturity. They initially shipped fast but really poor quality product, leading many core flows to be completely broken.
In other words, it is possible to keep maintenance cost low for a complex system — but after a certain point, there is a need to be thorough as you’re building out.