^EN This is the clearest sign yet that the AI industry's bottleneck has shifted from model quality to raw GPU supply. If a company as well-funded as OpenAI is publicly worried about capacity, smaller labs and enterprise customers should expect longer waitlists and higher prices across the board — this is a supply story now, not a demand story.

^EN Building 'attacker' and 'defender' AI in the same lab is an efficient way to stress-test both sides, but it also means Korea is quietly building offensive cyber-AI capability under a defensive banner — worth watching how the attack model is governed once it's capable.

^EN A smaller lab claiming parity with frontier labs is a familiar pattern — the real test is independent benchmarks and real-world coding tasks, not launch-day claims. Still, competition at the top keeps pricing and quality pressure on the big players, which benefits developers either way.

^EN A revenue-based fine (rather than a flat cap) is the same playbook the EU used with GDPR — it hits large tech and platform companies hardest, since 10% of a small firm's revenue is trivial but 10% of a conglomerate's is enormous. Expect heavy lobbying before this becomes final.

^EN Retail investors bought these leverage products for exposure to Korea's chip giants, but the biggest winners so far look like foreign HFT firms and brokerages collecting fees — a familiar pattern where new financial products enrich the middlemen before the intended users. Regulators will likely face pressure to add speed bumps or fee caps.

^EN The timing is the story here — selling right before a ceasefire announcement raises obvious questions about whether the trade was informed by advance knowledge, not luck. Even if legal, a sitting president profiting from a war he has direct influence over is a conflict-of-interest problem that self-reporting alone doesn't fix.

^EN Big investment numbers mean little if the approval pipeline is the actual bottleneck — this is a governance failure, not a funding one. Companies can have capital and intent but still be stuck waiting on permits, which is exactly the kind of self-inflicted delay Korea can't afford while competing for AI infrastructure with the US and China.

^EN Using insects instead of robots is clever engineering economics — a cockroach's own legs and navigation come free, so the electronics only need to add sensing and payload. The real hurdle won't be the hardware but regulatory and ethical approval for deploying living, wired insects at disaster scenes.

^EN It's a small story, but it captures something real about how far people will go for animals they love once conventional options run out — and it's a neat example of repurposing consumer drone tech for a use case it was never designed for.

^EN This is a good reminder that space tech pays off in unexpected earthbound ways — the same math that separates mineral signatures on Mars turns out to separate ancient pigment from weathered rock here. It's basic research paying dividends decades after the original goal.