Real AI LLM capabilities?

 Andrej Karpathy says only 5,000 people have seen what AI can really do - The New Stack

The AI Usage Hierarchy: According to Andrej Karpathy's "napkin math," the global population is split into distinct tiers regarding AI exposure:

• ~6 billion people have barely touched an LLM.
• 1 to 2 billion use chatbots casually.
• ~20 million use frontier models professionally for math and code.
• Only about 5,000 people have access to the most powerful frontier systems inside elite labs.

The "Vibe Coding" & Agentic Shift: Karpathy notes a massive evolution in software development, transitioning from casual coding autocomplete to "agentic engineering," where developers act as directors reviewing complex changes made by multi-hour autonomous agents rather than writing code manually.

Where Capability Peaks: Karpathy argues that AI capability peaks in domains that are digital, verifiable, and economically valuable (such as coding and math), because results can be tested and measured via reinforcement learning.

Enterprise Bottlenecks: Box CEO Aaron Levie notes that while agent swarms (parallel agent workflows) represent the future, most enterprises are a long way from utilizing them. Obstacles include high costs at scale, varying model capabilities, security/compliance concerns, and the need for significant change management.

Async/Await languages design

A Design Space Exploration of Async/Await @brown.edu

Many programming languages now provide the async/await keywords for expressing concurrency. The design rationale is pretty consistent: to make concurrent programs look more like straight-line code (see: Python, Rust, or Swift). We therefore describe the paradigm which encompasses async/await as straight-line asynchrony, as opposed to using event loops or callbacks.

...how similar or different is async/await between languages?

The short answer is a lot more different than expected.