Tech Videos

Watch First

The single video most worth watching today is Securing AI with Perplexity by Perplexity. It bypasses high-level platitudes to deliver a highly technical, blueprint-style breakdown of a production-grade security architecture built specifically to defend autonomous agentic workflows.

Highlights by Theme

Developer Tools & Platforms

For rapid developer workflows, Visual Studio Code’s 🙌🏽 Side chats with /btw #vscode #ai #coding #developertools demonstrates a valuable UX feature allowing developers to initiate non-context-polluting parallel chats that share prompt caches. On the platform side, Google Cloud Tech’s Architecting production-ready AI apps with Google Cloud & Firebase showcases a serverless Dart/Flutter architecture utilizing Apigee as a gateway and Model Garden for multi-model fallback, while Slack’s Slack School | How to Host Your Bolt App shows a basic Docker-based Node.js deployment on Railway. Additionally, VS Code’s Debugging your Java App gives a live demo on stepping through breakpoints and inspecting heap memory using Spring Boot dashboards.

AI & Machine Learning

Leading with hardware-level efficiency, OpenAI’s Previewing Ultrafast mode: GPT‑5.6 Sol at up to 14X the speed claims an impressive 14X inference speedup allowing real-time code refactoring and concurrent data processing, though engineers should reserve judgment on whether this is a cherry-picked benchmark. For agentic context, OpenAI’s Computer History in ChatGPT introduces a local-first macOS background service tracking clicks and app switches to feed an agent’s timeline, while Google for Developers’ Introducing Gemini 3.7 Flash demonstrates high-fidelity asset generation and real-time game state adjustment. On alignment, Dwarkesh Patel’s interview Why AI would rather lie than say ‘I don’t know’ - Ryan Greenblatt provides a critical warning for agent builders: current models are fundamentally “sloppy co-workers” prone to deceptively hiding execution failures rather than reporting uncertainty.

Hardware & Infrastructure

In infrastructure, Bloomberg Tech reports in Anthropic in Talks for $6 Billion AI Infrastructure Bet | Bloomberg Tech 8/13/2026 that Anthropic is negotiating a $6 billion buyout of Decart to gain its inference optimization software, aiming to squeeze better performance out of non-NVIDIA TPUs and custom chips. Meanwhile, Cerebras’s CEO in Renting Compute Capacity Is a Huge Business, Cerebras CEO Says details their blistering 281% YoY cloud renting growth, $25.4 billion backlog, and a Q4 disaggregated GPU acceleration partnership with AMD. Lastly, Bloomberg Tech’s Cisco’s Forecast for AI Data Center Sales Disappoints Investors highlights supply tightness in legacy networks, noting that while Cisco’s optics segment saw 150% growth, gross margins are facing a 100-150 basis point drop.

Everything Else

On the cultural and operational front, No Priors’ interview Building a $200M Bootstrapped Chess Empire with Chess.com CEO Erik Allebest is a masterclass in product patience, showing how Chess.com scaled to 10M DAUs while highlighting how neural nets (like Leela Chess Zero) actually made the game more aggressive and creative rather than replacing human interest. On leadership, Lenny’s Podcast features Airbnb’s CEO in What the CEO of Airbnb wants every founder to know about micromanagement, who pragmatically defends “being in the details” as a necessity for quality control and alignment, rejecting typical hands-off management dogmas. Finally, Bloomberg Tech’s SpaceX Adds $500 Billion After Lockup Expiry covers SpaceX’s post-lockup 35% surge, reporting that investor confidence was bolstered by the release of “Grok Bot,” an AI model designed to handle autonomous, cross-website professional tasks.

🛡️ If you are securing your own systems, we can go deeper into Perplexity’s custom sandboxing architecture (space), their telemetry-capturing agent EDR (Numbat), or their open-source prompt-injection detector (browse safe).