learn a core skill - writing, design, editing create pow - reach out to people with that pow - get active on twitter and linkedin - get an intern - do good work there i.e create more pow - switch org.

pls actively go and seek out your firsts in your 20s, your first $1k, first heartbreak, first flight abroad, first LinkedIn post, first offer letter, first everything that makes your heart go brrrrr

A year ago, we barely ranked on Google and this year, 44.3% of Flexprice’s inbound leads came from AI search. I’ve put together everything we tried, what worked, and a few additional resources at the end of the article for anyone looking to do the same.

We do billing for @SimplismartHQ, one of the best teams building inference infrastructure and working with them has made us realise how early an inference company needs AWS-level billing infrastructure. A customer might switch GPU types, scale from two replicas to six during a traffic spike, or shut a deployment down overnight. Billing needs to track how long each replica ran, which hardware it used, and whether model loading and idle time count as billable usage. It also needs to apply the right rates, commitments, and prepaid credits, even when a workload moves to another GPU mid-month and these are everyday customer requirements. When billing cant handle them, engineers end up reconciling usage logs with commercial agreements every month instead of building the product. We automated this for Simplismart with @tryflexprice, giving their team back 30% of its daily engineering bandwidth. We've been closely observing how inference platforms approach billing, from GPU metering and pricing models to commitments and prepaid credits. The public pricing pages and billing docs of 30 platforms, including Baseten, Together AI, Fireworks AI, CoreWeave, Modal, and Groq, offer an interesting look at how these systems work. A report covering the findings and what they mean for billing infrastructure is in the works. Well share the document with you all soon.