Welcome back to LLM Decode 👋
AI is moving beyond generating answers and into the systems that run businesses. NVIDIA is using AI to accelerate how future chips are designed, while startups are putting AI directly inside live sales conversations to help people make better decisions in the moment.
Here’s what matters today.
AI Is Starting to Design the Chips Behind AI
NVIDIA has introduced cuPotential, a new CUDA-X library designed to accelerate atomic-scale simulations used in semiconductor research and materials science. The goal is to help researchers and AI agents model how atoms and molecules interact during the development of advanced chip materials. NVIDIA says early testing showed 2–3× additional acceleration on workflows that were already GPU-optimized.
NVIDIA is also working with chip-design startup ChipAgents on Build-A-Chip-Bench, a benchmark designed to test whether AI agents can handle an end-to-end chip-design workflow. Instead of evaluating AI on isolated engineering tasks, the benchmark looks at stages including specifications, implementation and verification.
The bigger idea is a feedback loop: AI is increasingly being used to improve the design and manufacturing of the hardware needed to run AI itself. As chips become more complex, automating parts of engineering could help reduce development time and improve how efficiently future processors are built.
Why it matters
AI is entering hardware engineering: The technology isn't stopping at software and content.
Chip design could become more agentic: AI systems may eventually handle multiple stages of engineering workflows.
Faster simulations can shorten development cycles: More experimentation becomes possible without proportionally increasing compute time.
AI could help improve AI infrastructure: Better chip designs can ultimately translate into more capable and efficient models.
Your Next Sales Coach Could Be AI - In Real Time
Sales AI is moving beyond analyzing calls after they happen. Startup Othello has raised $8 million to build an AI system that listens to live sales conversations and provides representatives with recommendations while the conversation is still happening.
The system is designed to understand the context of a conversation and surface useful information or suggestions in real time. Instead of waiting for a salesperson to review a call transcript later, AI can act as an in-the-moment decision-support layer.
Othello says its technology has been trained on 20 million hours of sales conversations and can integrate with tools such as Salesforce, Google Workspace and Slack. The company says customers including Microsoft and Capital One are using the technology, though those performance claims are company-reported.
Why it matters
AI coaching is becoming real-time: The value shifts from analyzing what happened to influencing what happens next.
Expertise can become more accessible: Less-experienced salespeople could get contextual support during difficult conversations.
AI becomes part of the workflow: Instead of opening a chatbot, employees can receive assistance where work is already happening.
Human judgment remains central: AI can suggest the next move, but the salesperson still owns the relationship and decision.
4 AI Workflows to Try
Add AI Before the Decision, Not After: Instead of using AI only to summarize meetings, experiment with having it surface key risks, missing information or next steps before a decision is made.
Build Context Into Your AI Tools: Connect AI to the information employees already use CRM data, documents, project history or previous conversations to make recommendations more relevant.
Use AI as a Second Pair of Eyes: For important emails, sales proposals or presentations, ask AI to identify objections, unclear claims or missing information before you send them.
Look for “In-the-Moment” AI Opportunities: Ask: Where does someone repeatedly stop working to search, check information or make a routine decision? That's often a strong candidate for AI assistance.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.