The infrastructure wars are shifting to a new front: who can help humans and AI agents communicate better. While everyone else optimizes for speed and cost, the real value is in bridging the understanding gap.
01
Vercel CEO lays out the blueprint for "agentic infrastructure"
Guillermo Rauch argues that cloud infrastructure built for human developers won't work for AI agents. His vision: infrastructure that configures itself for agents, deploys long-running agent processes instead of static pages, and eventually becomes an agent itself that holds the pager and self-heals when things break.
Why it matters: If you're building agents that need to deploy code, manage databases, or scale compute, the current DevOps playbook assumes a human is running the show. Rauch is betting that assumption breaks down fast.
Anthropic's Thariq Shihipar calls prompting "the new public speaking"
The Anthropic engineer says prompting will remain a high-leverage skill like writing or public speaking, describing it as "the skill of talking to agents, mediated by the harness." His goal is growing the bandwidth between humans and agents to help them understand each other better.
Why it matters: While everyone debates whether prompt engineering will disappear, someone actually building the agent infrastructure thinks it's becoming more important. The skill isn't going away, it's evolving into something more fundamental.
Andrej Karpathy warns of a growing "AI capability gap"
The former OpenAI researcher says too many people tried ChatGPT's free tier months ago and let that experience define their view of AI capabilities. He points to viral videos mocking AI mistakes while noting that people aren't seeing the latest models in action.
Why it matters: The person who helped build GPT is saying public perception of AI is dangerously outdated. If decision-makers at your company are still thinking in terms of 2023 ChatGPT, they're missing what's actually possible today.
Google product manager Josh Woodward hinted that Gemini will soon deliver "even more creative model replies," sharing an example of more engaging AI output.
Product leader Peter Yang preps his kids for an "AGI proof career"
Yang shared his approach to training his children for careers that will survive artificial general intelligence, though he didn't detail the specific strategies.