Nvidia and Palantir demonstrate smaller model advantage

Nvidia has shown that a more compact language model, when trained on domain-specific operational decisions, can substantially outperform a much larger counterpart at supply-chain allocation tasks. According to reporting by TNS contributor Paul Sawers, Nvidia's 30-billion-parameter Nemotron 3.5 Lightning achieved 86.7% accuracy in determining optimal placement for constrained components, whereas the company's 550-billion-parameter Nemotron 3 Ultra managed only 55.5% accuracy on the same problem.

The Lightning model was refined using actual decisions from Nvidia's operations team, with Palantir's platform serving as the bridge between those operational choices and the underlying supply-chain information. The two organizations initially disclosed this initiative in June and formally announced the results on Thursday. Both companies are encouraging other enterprises to adopt this methodology, leveraging proprietary business data while maintaining ownership of their custom models.

The findings suggest that organizations selecting a language model for a particular use case may benefit from choosing a smaller architecture and tailoring it through fine-tuning to their specific business context. However, the performance improvements did not extend uniformly across all tasks. The question remains: which supply-chain challenges saw the most improvement through fine-tuning, and which problems proved resistant to this approach?

The AI-speed SOC: A peer workshop for security leaders

Alert volumes are climbing, and AI is helping attackers move faster than most SOCs can respond. On September 15, a small group of CISOs and SOC leaders will meet behind closed doors to define what security operations will look like when AI becomes core to its architecture.

  • A closed-door working session under Chatham House Rules, not a webinar you watch
  • Small-group roundtables with peers who run your same reality, capped at 25 spots
  • Open only to CISOs, SOC Directors, and IR leaders with teams of 10+

Top of the stack

Shopify rebuilt its mobile app in 12 weeks after abandoning React Native

Shopify is moving away from React Native toward native mobile development, leveraging AI agents to reconstruct its Shop app in just 12 weeks and reimagine how code repositories function with agent-driven development. In 2020, the company made a strategic commitment to React Native that gained traction across the developer community by enabling teams to write mobile functionality once rather than maintaining separate implementations in Swift and Kotlin. On Thursday, Shopify announced it is reversing that decision.

What else is new?

TDS launches ShipAI: functional implementations, not marketing presentations

Towards Data Science, a sister publication, has introduced ShipAI, a curated collection of genuine AI implementations: screen-recorded demonstrations, editorially reviewed, and preserved as substantive reference material rather than ephemeral social feed content.

  • Access genuine, operational AI implementations rather than theoretical case studies
  • All submissions undergo editorial review before inclusion in the archive
  • Top implementations are highlighted monthly as essential viewing

Developers working on AI projects can submit their work for consideration and visibility.

Source: The New Stack