Tech Videos — 2026-05-15#

Watch First#

The single video most worth watching is Building AlphaGo from scratch – Eric Jang from the Dwarkesh Patel channel. It is a highly technical, rigorous breakdown of Monte Carlo Tree Search and reinforcement learning that clearly connects classical game-solving architecture to the future of large language model reasoning.

Highlights by Theme#

Developer Tools & Platforms#

The AI Engineer channel featured highly pragmatic talks on integrating AI deeply into the SDLC, notably Brian Scanlan explaining how Intercom uses Claude Code and custom MCP skills to drive automated PR approvals while bypassing human bottlenecks (How Building with AI Can Double the Throughput of Your Engineering Team — Brian Scanlan, Intercom). Similarly, Pedro Rodrigues from Supabase demoed a novel approach to resolving LLM context gaps by exposing their documentation directly to agents via SSH, allowing the model to navigate it like a native file system (Combine Skills and MCP to Close the Context Gap — Pedro Rodrigues, Supabase). For front-end engineers, GitHub’s The Download: Microsoft Build, GitHub mobile, open source keyboards & more highlighted Redact, a tiny, AI-generated React-compatible projection runtime built for narrow use cases, demonstrating how LLMs lower the cost of maintaining specialized implementations of large libraries.

AI & Machine Learning#

Eric Jang’s deep dive in Building AlphaGo from scratch – Eric Jang is a masterclass in AI architecture, dissecting how a 10-layer neural network can amortize intractable search problems and analyzing whether we can reliably apply Monte Carlo Tree Search to open-ended LLM reasoning loops. In an enterprise context, Marc Benioff joined the Trump-Xi Summit, Benioff: “Not My First SaaSpocalypse,” OpenAI vs Apple, Multi-Sensory AI, El Niño episode of the All-In Podcast to discuss Salesforce’s $300 million Anthropic spend, arguing that multi-sensory AI models running intelligently at the edge represent the real path forward for agents. Finally, NVIDIA’s brief video Understanding the AI Tokenomics Equation offers a practical reminder that engineers must aggressively optimize inference frameworks like vLLM and SGLang to decouple raw chip performance from the actual cost of running agentic systems.

Hardware & Infrastructure#

On the consumer hardware front, Marques Brownlee reviews a new $99 screenless fitness tracker in Why Fitbit Air Could Be Huge, framing it as a direct, subscription-free competitor to Whoop. For hardware hackers, GitHub’s The Download: Microsoft Build, GitHub mobile, open source keyboards & more notes that Keychron has open-sourced CAD design files for over 100 of its keyboards and mice. Lastly, an interesting infrastructure nugget surfaced during a Jane Street data center tour in the Dwarkesh Patel interview, highlighting the shift to liquid cooling with deionized water and propylene glycol to support dense GB300 AI cabinets that peak at 140kW each.

Everything Else#

If you are skeptical about AI radically changing engineering management, watch Agents Don’t Do Standups: Building the Post-Engineer Engineering Org — Mike Spitz, PFF; PFF completely ripped out scrum and project managers, using Claude to turn a two-engineer “tiger team” into a feature factory deploying 25 times more often than a standard 10-person team. The Fireship channel summarizes the chaotic discovery leaks from the $134 billion Musk vs. Altman lawsuit in I can’t believe this trial is real…, while Syntax highlights a humorous package tracking how often developers curse at their AI coding assistants in F#%K YOU AI.