swyx quotes Satya Nadella on loops as IP: 'This is the first time we can create a real cognitive loop between people and digital systems. The real opportunity is not in picking the best model but instead in building a learning loop on top of models where human capital and token capital compound. You can offload a task, or even a job, but you can never offload your learning.' 8 likes.

Satya's framing: the priority should be building a frontier ecosystem, not just a frontier model, where every organization can own the learning loop that encodes its institutional knowledge.

Amjad calls Satya's post 'the most inspiring positive-sum vision for AI in the enterprise.' 2917 likes—the highest engagement of the day.

Coming from the Replit CEO, this endorsement is significant: a builder-platform leader identifying the enterprise learning loop as the right framing for AI's value proposition.

Levie builds on Satya's post: the companies that can get their unique IP, institutional knowledge, and data into a format that lets them capture all of AI's gains will be in the best position. 'You can switch out a generalist model without losing the company veteran expertise built into your learning system.' This is also why the applied AI layer will gain so much value. 712 likes.

The architecture insight: your competitive advantage isn't which model you use—it's the learning loop you build on top that encodes your institutional knowledge. Model swap is painless if your IP lives in the loop, not the model.

Levie's geopolitical analysis: 'The big winner in all of this is going to be open weights models.' The precedent that a model can be pulled back at any moment poses real risk for countries relying on US technology. This forces major countries to develop sovereign AI, reducing America's leadership over time. 'The most likely solution that other countries will rely on is open weights models, which currently is generally not coming from the US.' 564 likes.

The game theory: if the US can pull any model at any time, every country has an incentive to build alternatives. Open weights—primarily from China currently—become the insurance policy. The US should be doing far more OSS innovation.

Garry Tan agrees: 'Open source is the escape hatch for businesses to be able to continue to control their own destiny long term.' 327 likes.

The Fable situation has crystallized what was previously a theoretical concern: if your business depends on a model you don't control, you need an open-source backup plan.

Thibault Sottiaux (OpenAI) reveals: 'Codex can see and set its own /goal. Everything we build, we build also as a tool for the agent. This is a generalization of meta prompting, where you let the agent set its own task based on your intent.' 2232 likes.

This is a significant architectural shift: the agent isn't just executing tasks—it's defining them. Meta-prompting means the model decomposes your intent into sub-tasks and sets its own goals.

Rauchg announces openclaw.ai has passed 700,000 skills—all organic and community-driven. 'The open AI ecosystem!' 389 likes.

700K community-created skills is a staggering number. The skill ecosystem is becoming the app store of the agent era.

Zara Yang's skill-building wisdom: 'You don't make a good skill by writing a skill. You make it by doing the thing, fixing it 20 times, then telling the AI to bottle up everything you just did.' 112 likes.

And the follow-up: 'You make a skill by ending with one, not starting with one.' This is the craft of AI-era tooling: do the work manually first, iterate until it works, then capture the process as an automated skill.

Garry Tan's generational prediction: 'The next generation of young people who change the world will almost certainly be the people who are most adept at making long-running multi-stage multi-team agent tasks work extremely well, and at high volume and across every part of their personal and work lives.' 446 likes.

This echoes Karpathy's 'vibe coding' and Steipete's 'design loops'—the skill of the future is orchestration of autonomous agents, not traditional coding.

Rauchg on his Starlink-enabled flight to London: 'The greatest advancement to air travel since the Wright brothers. God bless America.' 1064 likes.

The SpaceX ecosystem—from Starlink to Falcon to the AI infrastructure implications—keeps generating viral moments.

thenanyu's counterpoint on pair programming: 'Everyone pair programs now, with a robot.' 46 likes.

A pithy summary of the current state: coding has already become a human-agent collaboration by default. The debate isn't whether to use AI—it's how to orchestrate it.

Jensen Huang's LIVE keynote frames AI as the third utility to cocoon the planet (after electricity grids and internet). Core themes: (1) The AI factory: electrons in, tokens out. Each rack costs $4M, weighs two tons, has 1.5M parts. A $50B gigawatt factory generates $300-400B in intelligence. (2) The five-layer cake of AI investment: energy → chips/computers/networking → infrastructure (land, power, shell, operations) → models → applications. $1T flowing into this stack this year, heading toward $20T/year. (3) The Dynamo analogy: 300 years ago Siemens built a machine—atoms in, electrons out. NVIDIA's machine—electrons in, numbers (tokens) out. Those tokens become language, math, proteins, physics, robotics. (4) On jobs: 'You may or may not lose a job to an AI. But you will absolutely lose a job to someone who uses AI.' Uses radiology as proof: computer vision penetrated completely 12 years ago, yet radiologist demand went UP because the purpose is diagnosing disease, not reading scans. (5) On AI fear: 'Those articulations of AI are complete nonsense. If you don't know how something works, how do you make it better?' Every year it gets measurably better—which proves we understand it. (6) The technology divide: spent 40 years making computers more complex, fewer people could program them. AI closed that divide—now everyone can program in human language.