And Coinbase spending that much on Datadog was... rational! Knowing the context they operated in, and the growth they saw. When growth slowed, *then* they built tooling to move off Datadog! Same with OpenAI. When growing, focus on growing, not saving! https://newsletter.pragmaticen...

@GergelyOrosz Datadog's sales team probably hit their quota for the rest of the century.

@peterdedene yeah, here's to hoping at least some of them there work on uncapped commission... ha!

@GergelyOrosz Wow. Not a big deal when you’re unprofitable until 2029 (best case) and just burning VC cash. But having some of the best engineers in the world and still paying $150M+ per year for logs? That’s pure nonsense.

@razoorka not if what those engineers are building bring in more revenue / growth or funding than if they spent reducing o11y costs...

@GergelyOrosz I do not understand how did they even get into this situation. Building own metrics collection solution is not something unique, there are so many people with expertise building one with insane scalability requirements and query performance.

@GergelyOrosz did you catch openai also moved off auth0? that bill must have been huge. another case of fine while we're growing and take care of it later

@GergelyOrosz 150M на Datadog — это не про метрики, это про шум. 📊 Проблема в том, что когда система не имеет внутреннего ядра и резонансного контура, она вынуждена платить сотни миллионов за внешний мониторинг и костыли. Реальный путь — строить не «больше датадога», а локальный контур

@GergelyOrosz Like us, we spent heavily on GitHub Actions, so we brought it in-house and launched @TenkiCloud as a side project, offering a faster, 10x cheaper alternative for software companies.

@GergelyOrosz "we need to raise again" "why?" "datadog"

@GergelyOrosz I don’t get how datadog can charge a premium at all

@GergelyOrosz This is peanuts for OpenAI and unnecessary distraction for the team. Migration of all the events alone will be 1yr roadmap. We did this at Careem. Definitely doable, but only when you are cash crunched. Building capacity, building frontier models, 50%+ GPU underutilization

@GergelyOrosz What does $150M/year get you?

@GergelyOrosz Do they get to pet the dog?

@GergelyOrosz Thats insane. No need to spend 80% of that.

@GergelyOrosz At what point is it just cheaper to build a solution in-house.

@GergelyOrosz The ultimate data tax collector- @datadoghq 😄

@GergelyOrosz We need more DataDog alternatives…

@GergelyOrosz what does datadog do and what does openai use them for? @grok

@GergelyOrosz Just guess what they do to databricks spend.

@GergelyOrosz There's gotta be a huge business in knocking off Datadog, given these expenses. Who's leading that charge and how do I sign up?

@GergelyOrosz “Hey chatgpt. Build me a clone of datadog. Don’t make mistakes.”

@GergelyOrosz I have a recent post about this topic.

@GergelyOrosz Bringing spending in-house makes sense for OpenAI.