Tuesday, June 2, 2026
The infrastructure of the agent era is here.
June 2 · 8 videos
GitHub projects 14 billion commits by 2025.
Agents are driving the volume.
Dell AI server revenue jumped 800 percent.
Andrew Ng says anyone can build apps in 30 minutes.
The Pope issued a 235 page warning on human dignity.
Fine-tuning is now 10 times cheaper than frontier APIs.
“The code was always just an annoying technical layer in between to create what we wanted.”
AI Engineer Melbourne 2026 Keynote Livestream | Day 1
Unknown · AI Engineer · 6 min
Watch on YouTube →The AI Engineer role matures into a foundational pillar of software development. This keynote marks the transition from experimental wrappers to industrial grade AI systems in the APAC region.
- Regional tech hubs like Melbourne are becoming critical centers for AI implementation.
- The AI Engineer role has diverged significantly from traditional Data Science.
- Conferences serve as primary drivers for industry standardization in emerging markets.
- Focus is shifting toward systems integration and deployment rather than just model experimentation.
- Networking in regional hubs is now as critical as global knowledge for career advancement.
Alastair Reacts to Tony Blair's Attack on Labour
Alastair Campbell · The Rest Is Politics · 61 min
Watch on YouTube →Rory Stewart and Alastair Campbell contrast the Pope's moral framework for AI with Tony Blair's tech-optimist political strategy. They examine the existential risks of AI to labor and democratic truth.
- Pope Leo XVI issued a 235 page encyclical titled Magnifica Humanitas regarding AI ethics.
- Tony Blair published a 5,600 word essay urging the Labour Party to embrace competitive AI.
- Indifference to facts is identified as a precursor to totalitarianism in the age of populism.
- AI risks mass unemployment with one CEO citing 67,000 potential job losses in call centers.
- The hosts propose a four part framework for AI: fatalism, tech optimism, regulation, or stoppage.
- Work is viewed as a vocation that universal basic income cannot easily replace.
Build Your Own App In Just 30 Minutes! Full Course with Andrew Ng
Andrew Ng · DeepLearningAI · 25 min
Watch on YouTube →Andrew Ng demonstrates how AI has democratized software creation for non-coders. The focus is on using structured prompting to build functional web applications rapidly.
- The 5 Building Blocks of Prompting are Goal, Input, Output, Layout, and Special Features.
- Software creation has shifted from manual syntax writing to high level instruction giving.
- Minimum Viable Product time to market has shrunk from weeks to minutes using AI.
- Prompting skills are platform agnostic and apply to ChatGPT, Gemini, and Claude.
- Users are encouraged to troubleshoot bugs by describing symptoms to the AI rather than reading code.
What Lies Beneath the API — Benjamin Cowen, Modal
Benjamin Cowen · AI Engineer · 12 min
Watch on YouTube →Benjamin Cowen from Modal argues that successful AI products eventually move from general APIs to custom fine-tuned models. This shift is driven by cost, latency, and the need for specific business logic.
- Custom models can cost as little as one tenth of frontier API expenses.
- Supervised fine-tuning and reinforcement learning can be implemented in 300 lines of Python.
- Three signals for fine-tuning: high API costs, performance plateaus, or strict latency needs.
- Intercom and Pinterest are already beating frontier APIs using specialized models.
- Data collection and evaluation development should begin long before training starts.
Task Fidelity Scaling Laws — Kobie Crawdord, Snorkel
Kobie Crawford · AI Engineer · 20 min
Watch on YouTube →Kobie Crawford presents research showing that task quality is the primary driver of performance in model fine-tuning. High quality data yields a 5x improvement over low quality data for the same compute budget.
- Fine-tuning on high quality tasks improved the base model by 6 percent compared to 1 percent for low quality.
- High quality tasks average twice as many tool calls and more reasoning tokens.
- Ambiguity in tasks prevents models from learning actionable patterns and should be avoided.
- Model improvement can be masked by the quality of the evaluation tasks themselves.
- Expert in the loop curation is essential to avoid saturation on public benchmarks.
GitHub’s Agent Era: 14x Commits, 200M Developers, Copilot’s Next Act — Kyle Daigle
Kyle Daigle · Latent Space · 84 min
Watch on YouTube →GitHub COO Kyle Daigle discusses the platform's transition to an AI operating system. The company is preparing for a massive surge in agent driven code production.
- GitHub projects 14 billion annual commits by 2025, a 14x explosion driven by agents.
- The platform has surpassed 200 million developers as AI lowers the barrier to entry.
- Micro-skills are replacing brittle Mega-skills as the building blocks of AI functions.
- The Model Context Protocol (MCP) is being used to bridge silos like Slack and GitHub issues.
- GitHub is rewriting its legacy infrastructure to handle the scale of agentic activity.
- Ambient AI will allow agents to act with the judgment of human leaders by using full context.
How Lovable self-improves every hour — Benjamin Verbeek, Lovable
Benjamin Verbeek · AI Engineer · 19 min
Watch on YouTube →Benjamin Verbeek explains how Lovable uses automated feedback loops to improve its AI platform hourly. The system is designed for non-technical users who need zero friction.
- Lovable generates over 200,000 projects daily for mostly non-technical users.
- The Vent Loop allows agents to report platform bugs directly to developer Slack channels.
- The Lovable Stack Overflow clusters successful unblocking strategies to inject as context.
- Holdout groups are used to prune context rot and maintain system efficiency.
- The goal is to ensure a technical mistake happens exactly once and is then automated away.
Dell's Comeback Marks a Turning Point in AI
Josh · Limitless Podcast · 28 min
Watch on YouTube →Dell has transformed into a primary infrastructure provider for the AI era. The company is solving power and cooling challenges for massive GPU clusters.
- Dell's AI server revenue surged 800 percent year over year.
- The stock price increased 240 percent in 2024 as Dell became an AI arms dealer.
- NVIDIA's Blackwell based DGX Station and RTX Spark chips enable local frontier model usage.
- On-prem AI is becoming a priority for sensitive industries like banking and healthcare.
- Michael Dell's leadership is credited with navigating the transition from legacy servers to AI factories.
References
PeoplePope Leo XVI · Tony Blair · Andrew Ng · Benjamin Cowen · Kobie Crawford · Kyle Daigle · Benjamin Verbeek · Michael Dell · Jensen Huang · Satya Nadella · Rory Stewart · Alastair Campbell · Mitchell Hashimoto · Thomas Dohmke · Nat Friedman
ToolsGitHub Spark · Model Context Protocol · Modal · Snorkel AI · Lovable · NVIDIA Blackwell · RTX Spark · ChatGPT · Gemini · Claude · WorkIQ · FoundryIQ
PapersMagnifica Humanitas