Tech Videos — 2026-07-01#

Watch First#

Continual Learning for Long-Running Agents: Agents That Keep Getting Better from NVIDIA Developer is a pragmatic must-watch for anyone building AI workflows. It cuts through the “1 million token” hype to address real-world context rot, arguing that long-running agents shouldn’t be fed massive walls of text, but rather handle context programmatically via recursive sub-agent delegation.

Highlights by Theme#

Developer Tools & Platforms#

The Model Context Protocol (MCP) is rapidly replacing bespoke API middleware, and Google Cloud Tech offers a crisp architectural breakdown of how it provides semantic context to agents in MCP vs API: The protocol every developer needs to know. Several major platforms demoed their MCP integrations today: Visual Studio Code showed off its new Foundry toolkit for building grounded agents without writing app code in I Built a Working AI Agent in Minutes - No Code Required, while AWS Developers released secure, read-only-capable API access for agents in Agent Toolkit for AWS helps you focus on building. Slack also ran multiple live demos, illustrating how to pull MCP tools and agents directly into collaborative enterprise channels via their CLI in Live Demo: Build an AI Agent for Slack.

AI & Machine Learning#

A standout technical talk from Cathay Financial reveals how they used synthetic data to fine-tune a 12B Llama 3 model that practically matched Claude 3.7 on intent classification—notably proving that RAG actually degraded their performance by 5.4% due to vector search mismatch in Fine-Tuning Financial LLMs for Customer Intent and Experience. For engineers struggling with slow agent runtimes, Google Cloud Tech shares a highly practical observability demo using OpenTelemetry to trace and parallelize Gemini agent TTS execution, cutting a 90-second bottleneck down to 20 seconds in How to speed up AI agents by 80% on the Gemini Enterprise Agent Platform.

Hardware & Infrastructure#

NVIDIA framed their GTC Taipei announcements around infrastructure built specifically for nanosecond, autonomous agentic workflows rather than human-app interactions in NVIDIA Data Center Partners Recap | GTC Taipei 2026 Recap. The recap highlights the new Vera CPU—marketed as the first CPU purpose-built for agentic AI throughput—and the DSX platform, a software optimization layer designed to safely overprovision power in heavily constrained data centers.

Everything Else#

For a masterclass in software history and culture, legendary engineer Kent Beck spans his 50-year career, the origins of TDD, and how AI might actually revive test-driven workflows in How Kent Beck shapes the software engineering industry from The Pragmatic Engineer. If you want a complete break from modern tech, historian Anthony Kaldellis provides a fascinating look into how the Roman Empire’s shift to a universal flat tax essentially created their “deep state” bureaucracy in Taxation in the Roman Empire - historian explains | Anthony Kaldellis and Lex Fridman. Alternatively, developers can find some nostalgia in an 8-minute history of TCP/IP, the web, and the dot-com bubble in The weird history of the internet… by Fireship.