YOUTUBE
Tech Videos
Sources AI Engineer All-In Podcast Andrej Karpathy Anthropic Apple Apple Developer AWS Events ByteByteGo Computerphile Cursor Dwarkesh Patel EO Fireship GitHub Google Cloud …
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
Watch First
Why Robotics Still Isn’t Solved - But Could Be Soon | YC Paper Club is the single video most worth watching. This multi-speaker session is a pragmatic masterclass that moves past robotic demo hype, highlighting tactile sensing and embodiment drift as real physical bottlenecks while showcasing actual solutions in visual-textual memory and 50x local inference speedups.
Highlights by Theme
Developer Tools & Platforms
In Realtime multiplayer, automation, and you! — Idan Gazit, GitHub on the AI Engineer channel, the team demos ‘Agentic Workflows’ (Markdown playbooks with YAML guardrails) and ‘ACE’ (a multiplayer cloud IDE running micro-VMs) to target the 95% of developer time spent outside raw typing. Adding pragmatic architectural context, UC Berkeley’s Frank Coyle in Anthropic’s CCA Exam as a Field-Guide for Agentic Engineering — Frank Coyle, UC Berkeley (AI Engineer) shares crucial agentic patterns, such as using stop_reason loops to control tool calls and subtask forking to prevent context pollution. Meanwhile, the feature highlight How stacked pull requests break down large code changes (GitHub) previews a clean workflow to decompose large PRs into parallel-reviewed layers that rebase automatically.
AI & Machine Learning
The standout technical showcase is Why Robotics Still Isn’t Solved - But Could Be Soon | YC Paper Club (Y Combinator), where Stanford and industry researchers present architectures like MAM (Multiscale Embodied Memory) (decomposing robot policy using compact text-space long-term memories for in-context mistake adaptation) and General Instinct’s distilled World Action Models (achieving a 50x local inference speedup via flow-matching distillation). In Benchmarking Coding Agents on New vs Legacy Codebases — Denys Linkov, Wisedocs (AI Engineer), Linkov delivers a reality check on zero-shot code refactoring, noting that while Sonnet 4.6 and Opus 4.8 show massive iteration gains over older models, GPT-5.5 extra high still fails by generating empty structural scaffolding. Lastly, the Google’s AI Brain Drain, SpaceX’s Huge Quarter, Airtable’s 90% Collapse, US Data Fuels China AI (All-In Podcast) panel debates the strategic impact of US data labeling startups selling $500M of expert-RLHF training data annually to top Chinese AI labs.
Hardware & Infrastructure
In Google’s AI Brain Drain, SpaceX’s Huge Quarter, Airtable’s 90% Collapse, US Data Fuels China AI (All-In Podcast), the hosts dissect SpaceX’s massive Q2 earnings, revealing that ‘Elon Web Services’ (compute renting to Anthropic and Google via Colossus) tripled to $2.6 billion, backed by an eye-watering $18.4 billion quarterly Capex spend. The panel details how Starlink’s direct-to-cell expansion is enabled by Starship launching V3 satellites that pack over 20 times the capacity of V2, while debating Microsoft’s reported 30%+ ROIC on tokens-as-a-service. On the global buildout front, NVIDIA CEO Jensen Huang in Firebird Launches CIS Region’s Largest AI Factory in Armenia (NVIDIA) announces a 250 MW local AI infrastructure deployment spanning Armenia and Kazakhstan over the next 12 months.
Everything Else
The Google’s AI Brain Drain, SpaceX’s Huge Quarter, Airtable’s 90% Collapse, US Data Fuels China AI (All-In Podcast) panel examines Bending Spoons’ acquisition of Airtable for a heavily-collapsed $1.28 billion, marking the end of the ZIRP-era no-code hype as conversational AI like Claude Code begins to instantly vibe-code bespoke internal tools. Complementing this shift, Whatnot CEO in What’s trending down for PMs: driving alignment (Lenny’s Podcast) warns that ‘relationship-building PMs’ who specialize in internal politics are trending down, replaced by highly technical, customer-oriented agents of change. Finally, the podcast’s brief promo 🚨 The All-In Summit 2026 Speaker Announcement (All-In Podcast) highlights speaker slots for Satya Nadella and Gwen Shotwell at their upcoming September summit.
🤖 I could compile a comparative technical report on the agentic patterns and coding agent benchmarks across the UC Berkeley and Wisedocs videos if you want to drill down further.