A pitch built around distrust of the frontier labs
Niteshift, from two engineers who helped scale Datadog, is entering the crowded AI-coding market with a pointed thesis: why would a company hand the code that runs its products directly to model makers like OpenAI and Anthropic, when those same labs keep launching apps that compete with their own customers? The startup launched with a $7 million seed led by Greylock's Jerry Chen, plus angels including Reid Hoffman, Datadog's founders, Braintrust's Ankur Goyal, and Reflection AI's Misha Laskin.
The Datadog analogy
CEO Sajid Mehmood reaches for his own history to make the case. In Datadog's early days, plenty of e-commerce companies refused to build on AWS while Amazon was busy putting retailers out of business - and that wariness fed Datadog's multicloud business. He sees the same dynamic now as AI labs push into verticals like legal, healthcare, and finance, the trend some call the "SaaSpocalypse."
What Niteshift actually sells
- It isn't trying to replace Claude Code or Codex; it argues it reduces dependence on them.
- Its AI coding cloud routes between models - Claude, GPT, open source, and others - based on each project's needs, so teams can switch rather than commit to one vendor.
- It sells infrastructure, not intelligence-as-labor: it charges like a cloud provider on per-minute usage, not by tokens.
The catch
Model independence isn't a new idea, and the field is brutal: Cursor (reportedly a SpaceX acquisition target), Cognition (which just raised $1 billion at a $26 billion valuation), Amazon Bedrock, and gateway player OpenRouter (which raised $113 million at a $1.3 billion valuation) all have a head start. Niteshift's answer is the founders' scar tissue - they lived the exact growing pains big engineering orgs now face with AI-generated code, and argue the infrastructure to run, test, and verify that code autonomously in production should be built by people who've done it at scale.