Harvard–Perplexity paper on agent usage

The DAS crew focused on Nvidia’s decision to open source its Nemotron model family, what that signals in the hardware and software arms race, and new research from Perplexity and Harvard analyzing how people actually use AI agents in the wild. The second half shifted into Google’s new Disco experiment, tab overload, agent driven interfaces, and a long discussion on the newly announced US Tech Force, including historical parallels, talent incentives, and skepticism about whether large government programs can truly attract top AI builders.

Key Points Discussed

Nvidia open sources the Nematron model family, spanning 30B to 500B parameters

Nematron Nano outperforms similar sized open models with much faster inference

Nvidia positions software plus hardware co design as its long term moat

Chinese open models continue to dominate open source benchmarks

Perplexity confirms use of Nematron models alongside proprietary systems

New Harvard and Perplexity paper analyzes over 100,000 agentic browser sessions

Productivity, learning, and research account for 57 percent of agent usage

Shopping and course discovery make up a large share of remaining queries

Users shift toward more cognitively complex tasks over time

Google launches Disco, turning related browser tabs into interactive agent driven apps

Disco aims to reduce tab overload and create task specific interfaces on the fly

Debate over whether apps are built for humans or agents going forward

Cursor moves parts of its CMS toward code first, agent friendly design

US Tech Force announced as a two year federal AI talent recruitment program

Program emphasizes portfolios over degrees and offers 150K to 200K compensation

Historical programs often struggled due to bureaucracy and cultural resistance

Panel debates whether elite AI talent will choose government over private sector roles

Concerns raised about branding, inclusion, and long term effectiveness of Tech Force

Timestamps and Topics

00:00:00 👋 Opening, host lineup, StreamYard layout issues

00:04:10 🧠 Nvidia Nematron open source announcement

00:09:30 ⚙️ Hardware software co design and TPU competition

00:15:40 📊 Perplexity and Harvard agent usage research

00:22:10 🛒 Shopping, productivity, and learning as top AI use cases

00:27:30 🌐 Open source model dominance from China

00:31:10 🧩 Google Disco overview and live walkthrough

00:37:20 📑 Tab overload, dynamic interfaces, and agent UX

00:43:50 🤖 Designing sites for agents instead of people

00:49:30 🏛️ US Tech Force program overview

00:56:10 📜 Degree free hiring, portfolios, and compensation

01:03:40 ⚠️ Historical failures of similar government tech programs

01:09:20 🧠 Inclusion, branding, and talent attraction concerns

01:16:30 🏁 Closing, community thanks, and newsletter reminders

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Anne Townsend, and Karl Yeh