news.volyx.in

Darkbloom – Private inference on idle Macs (darkbloom.dev)

501 points by twapi · 90 days ago · 252 comments on HN

Article summary

Darkbloom is a platform that enables private AI inference on idle Macs, allowing users to earn money by providing compute resources. The platform uses Apple's secure hardware to ensure that inference data remains private and secure. Darkbloom claims to offer comparable model performance at 50% lower cost than typical API providers. The platform is currently in public alpha, with operators keeping 100% of inference revenue.

Main themes

  • Private AI Inference
  • Secure Hardware
  • Cost-Effective Computing
  • Mac-Based Earnings
  • Attestation and Verification
  • Scalability and Business Models

What commenters say

  • The platform's reliance on Apple's secure hardware is a significant advantage, but it may not be foolproof against determined attackers.
  • The business model seems too good to be true, with some commentators questioning the feasibility of paying off a Mac mini in 2-4 months and making a significant profit thereafter.
  • The lack of a true hardware enclave for third-party code is a fundamental flaw in the platform's security strategy, making it vulnerable to exploits.
  • The use of Apple's secure hardware and attestation mechanisms can provide strong evidence of the integrity of the inference process, but it is not a guarantee of security.
  • The platform's scalability and potential for growth are significant advantages, but they also raise questions about the feasibility of operating at large scales and the potential for competition.
  • Some commentators are skeptical about the platform's claims of privacy and security, arguing that it is impossible to guarantee the confidentiality of data sent to external servers.
  • The comparison to other cloud-based inference services is unfair, as Darkbloom's use of idle Macs and secure hardware provides a unique value proposition.
  • The potential for abuse and exploitation of the platform's payment model is a concern, with some commentators arguing that it could be used to generate fake or malicious traffic.