Sources
- AI Engineer
- All-In Podcast
- Andrej Karpathy
- Anthropic
- Apple
- Apple Developer
- AWS Events
- ByteByteGo
- Computerphile
- Cursor
- Dwarkesh Patel
- EO
- Fireship
- GitHub
- Google Cloud Tech
- Google DeepMind
- Google for Developers
- Hung-yi Lee
- Lenny's Podcast
- Lex Clips
- Lex Fridman
- Life at Google
- Marques Brownlee
- Microsoft
- No Priors: AI, Machine Learning, Tech, & Startups
- Numberphile
- NVIDIA
- OpenAI
- Perplexity
- Quanta Magazine
- Slack
- The Pragmatic Engineer
- Visual Studio Code
Tech Videos — 2026-05-20#
Watch First#
Why Rust is different, with Alice Ryhl on The Pragmatic Engineer is a genuinely insightful deep dive into Rust’s memory safety, the borrow checker, and its growing integration into the Linux kernel, avoiding the usual marketing fluff.
Highlights by Theme#
Developer Tools & Platforms#
In Skill issue: Lessons from skilling up coding agents to use Langfuse on the AI Engineer channel, Marc Klingen shares a highly practical look at building skills to help coding agents navigate 500-page documentation, actively avoiding issues like hallucinated CLI parameters. Meanwhile, ThePrimeagenHighlights covers Jonathan Blow has an Announcement, showcasing a live demo of a puzzle game targeting the Nintendo Switch 2 built entirely in his custom systems language, Jai. Finally, The Pragmatic Engineer offers a deep technical overview in Why Rust is different, with Alice Ryhl, detailing the language’s RFC process, its unopinionated governance, and its recent graduation from “experimental” status in the Linux kernel.
AI & Machine Learning#
The most technically substantive update is From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents — Cormac Brick, Google on the AI Engineer channel, which details how fine-tuning a 270-million parameter “Function Gemma” model on synthetic data increased function calling success rates from 46% to over 90%. On the research front, Microsoft Research briefly introduces “Intervent” in Test-time verification for AI agents: New from Microsoft Research #ai #agenticai #verification, an LLM-based projection framework that provides real-time formal verification to steer reasoning models away from violations. OpenAI highlights a major milestone in pure math with The Erdős Breakthrough, demonstrating an AI solving a well-known unsolved problem in combinatorial geometry using algebraic number theory. Google also pushed out heavy marketing on Google DeepMind for its new Gemini 3.5 Flash has landed. and multimodal Build your next story with Gemini Omni., touting frontier-level intelligence at four times the output speed of other models.
Hardware & Infrastructure#
For cloud networking, Back to Basics: Understanding Availability Zones in the Context of Networking on the AWS Events channel is a solid whiteboard session breaking down cross-AZ data transfer costs and NLB topology-aware DNS to keep traffic zonal. The AWS Developers channel provides a pragmatic security reminder in Your AWS IAM Starter Checklist, emphasizing that using “AdministratorAccess” to bypass an S3 bug is a dangerous anti-pattern, and outlines how to use IAM Access Analyzer to achieve least privilege. NVIDIA Developer highlights the intersection of HPC and quantum in CUDA Live: CUDA-Q Academic Demo Day, demonstrating how GPUs simulate trillions of shots to synthetically train AI decoders for quantum error correction.
Everything Else#
In a skeptical look at AI-generated UI, the Syntax podcast asks Can AI Make Good Design?, calling out the homogenization of web design where generation tools rely heavily on established patterns and spit out a repetitive, watered-down “warm stew” of existing ideas. Lastly, on Lenny’s Podcast, The impact of war on the hardware industry briefly touches on the urgent push to re-industrialize the Western hardware supply chain in response to rapid, 3D-printed drone prototyping seen in Ukraine.