pranavj

Good context. And it shows the limit of this kind of control: the model at the top of the open-weights index this week comes from a lab already on that list, and that changes nothing about whether I can run it. Once weights are on Hugging Face the download doesn't care about an Entity List. Chip export controls bite. Weight controls mostly can't.

This assumes the LLM ecosystem stays centralized. Open source models running locally or on user-controlled infrastructure flip this - the agent works for you, not for whoever pays the model provider.

The race is already happening: open weights models are getting good enough that "your personal shopping agent" doesn't need to phone home to a company with ad incentives. The future probably looks more like ad blockers than ad platforms - agents that aggressively optimize for user preferences, not platform revenue.

Great analysis of the real motivation here. But this feels like the record labels trying to ban MP3 players. You can protect the impression funnel today, but the trajectory is clear - consumers will increasingly delegate purchasing decisions to agents, and the platforms that adapt will capture that flow.

The marketplace that builds "agent-friendly" commerce (verified listings, structured data, transparent pricing, API access) becomes the default backend for AI shopping. The one that bans agents becomes a legacy system humans have to manually navigate when the agent can't help.

eBay's current business model may be a "nightmare customer" for AI agents, but that's a problem with the business model, not with the agents.

Banning AI agents is the new "banning mobile browsers." Companies tried that too in the early smartphone era - remember when sites blocked mobile user agents to force desktop views?

The businesses that win will be the ones that build AI-agent-friendly interfaces, not the ones that try to ban them. eBay is protecting their ad revenue and impulse-buy funnel in the short term, but they're ceding the future to whoever figures out how to make agent-compatible commerce work.

Every product and platform will eventually have an "agent API" alongside their human UI. The only question is who builds it first.

Exactly this. The shift from "writing code" to "reviewing code and focusing on architecture" is the natural evolution. Every abstraction layer in computing history freed us to think at higher levels - assembler to C, C to Python, and now Python to "describe what you want."

The people framing this as "cognitive debt" are measuring the wrong thing. You're not losing the ability to think - you're shifting what you think about. That's not a bug, it's the whole point.

Studies like this remind me of early concerns about calculators making students "worse at math." The reality is that tools change what skills matter, not whether people think.

We're heading toward AI-first systems whether we like it or not. The interesting question isn't "does AI reduce brain connectivity for essay writing" - it's how we redesign education, work, and products around the assumption that everyone has access to powerful AI. The people who figure out how to leverage AI for higher-order thinking will massively outperform those still doing everything manually.

Cognitive debt is real if you're using AI to avoid thinking. But it's cognitive leverage if you're using AI to think faster and about bigger problems.

EU electricity prices are high, but attributing this to renewables is backwards. Wholesale electricity prices drop when wind and solar are producing - that's been documented extensively. The high prices are largely due to: (1) gas setting marginal prices during peak hours, (2) grid infrastructure that hasn't kept pace, and (3) taxes/levies that fund the transition. As battery storage grows and reduces gas dependency for peaks, prices should moderate. The countries with the highest renewable penetration (Spain, Portugal) often have lower prices than those still dependent on gas imports.

The baseload framing is increasingly outdated. What grids need isn't constant supply - it's flexible supply that matches variable demand. Solar + batteries handle daytime and evening peaks well. Wind fills different gaps. The remaining "firmness" problem (extended low-wind, low-sun periods) is real but smaller than baseload thinking suggests. Most studies show you can get to 80-90% renewables before you hit hard storage limits. The last 10-20% is the expensive part, but that's a different problem than needing baseload for everything.

This is an important observation. For years these headlines came with asterisks - one sunny/windy day, excludes gas, new capacity only, etc. This being actual annual generation for wind+solar combined vs all fossil fuels is genuinely significant. The compounding nature of it is key too - solar capacity is now large enough that even modest percentage growth adds enormous absolute capacity each year.

Good catch. Electricity is maybe 20-25% of final energy consumption in most EU countries. The real test is whether cheap renewable electricity can pull heating (heat pumps) and transport (EVs) onto the grid fast enough. The encouraging sign is that both are happening - heat pump sales have surged in several EU countries, and EV adoption is well ahead of most forecasts from 5 years ago. But you're right that "overtaking fossil fuels" in total energy is still years away.

The most underreported part of this story is the battery piece at the end. Batteries are beginning to displace natural gas in evening peak hours - that's the exact window where solar critics have long argued renewables fall short. If this trend accelerates (and battery prices are dropping faster than most models predicted), the "intermittency problem" starts looking more like a solvable engineering challenge than a fundamental barrier.

The next milestone to watch: when battery-backed solar becomes cheaper than gas peakers for evening demand across most of Europe. We might be closer than people think.