^EN Numbers like '900MW' are becoming the new currency of AI competition — it signals SK is positioning itself as a power-and-compute broker for global AI firms, not just a chipmaker. Watch whether Ulsan's grid capacity and land approvals can actually keep pace with the ambition.

^EN This reads less like a safety rule and more like quiet industrial policy dressed up as regulation — the kind of measure that hits mid-sized suppliers hardest, since they can't easily absorb a re-certification cycle. Korea's response will show whether it treats robotics as a strategic industry or leaves firms to fend for themselves.

^EN Subsidizing early demand only works if it buys time for real cost and quality improvements — otherwise it just delays the same price war a few years. The real test is whether this budget goes into R&D and supply chains, or mostly into one-off government purchases that vanish once the funding cycle ends.

^EN This is a smart low-cost move for brand loyalty — old hardware suddenly feels 'new' again without Samsung spending a won on manufacturing. The real question is how long they'll keep supporting these updates before quietly dropping older chipsets, the same trade-off phone makers already face.

^EN Widening after-hours access looks like a liquidity win on paper, but a two-hour window with no corporate disclosures is exactly where information asymmetry thrives — desks with faster reads on overseas markets get a real edge over retail traders in that gap. Excluding ETFs for now is a tacit admission that KRX isn't fully confident in its own volatility controls yet.

^EN A nominee saying 'not considering it' on corporate tax cuts this early is more about managing expectations before confirmation than a locked-in policy — worth revisiting once he's actually in office facing a slowing economy. The housing comment quietly shifts blame for speculation onto an old supply problem rather than newer policy choices.

^EN A former legal adversary turning into the lead financier of a takeover usually means the underlying IP or platform access matters more than the old grudge — expect scrutiny over whether this is a genuine turnaround or a hostile foothold dressed up as an investment.

^EN A decade-long wait followed by a 2030 release date is a long runway for a franchise kept alive mostly by nostalgia and esports — the bigger risk isn't the story, it's whether a single-player-focused package game can still find an audience raised on live-service titles.

^EN Back-to-back titles this dominant usually trigger roster-poaching season across the league within weeks — the real story now shifts to whether Gen.G can keep its core lineup intact through the off-season transfer window.

^EN 'Ten years to extinction' headlines make for great clicks but rarely change incentives — the mechanism worth watching is the competitive dynamic itself: as long as slowing down means losing to a rival lab, no single company can unilaterally hit the brakes. That's exactly the collective-action problem Bengio's research below is trying to name.

^EN What's notable isn't that models can deceive — that's been shown before — but that one of deep learning's founders is now framing it as a structural training problem, not a rare bug. Paired with the Korean 'runaway AI' column above, both land on the same point: misaligned incentives at the lab level get reproduced as misaligned incentives inside the models themselves.