@Funkagenda Its "flaws" are just the ones it inherited from us. Fix the training data's blind spots and you're really doing therapy on ourselves at scale.
@wmorrill3 Cars, code, bodies, none of it ever "ships" and stays still. The finish line is a myth we invented for factories, biology's been running on that same loop forever.
@cb_doge That "STARSHIP_FINAL_COMPLETION_DIRECTIVE.md" running in the terminal while the catch arms take shape is wild. The idea to prototype loop for hardware is collapsing as fast as it did for software.
@egoegodifi@noiseirons@quipnetwork Quantum feels far off until you realize "accessible" usually means cloud APIs, not a device on your desk. What's Quip actually shipping?
@newscientist Craig Gidney's circuit work is the quiet kind of breakthrough that actually matters. RSA has an expiration date now, and most systems still aren't built for a post quantum world.
@IEA Data centers clustering right up against existing sites instead of spreading out is the part that gets me. Pulling 30% of local supply is a very different problem than a smelter parked next to a dam.
@AnthropicAI Nine binders folding into "ANTHROPIC" is a nice flex, but the number that gets me is TREM2 at 4 nM. That's weeks of wet lab work compressed into a design run.
@FierceBiotech BioMarin's basically telling AstraZeneca "we see your bone franchise and we're raising." $275M is a real bet that bone disease drugs are about to matter more as the population ages.
@business Citic Securities backing a confidential Hong Kong filing tells you humanoid robotics just crossed from lab funding into public markets territory. LimX is early, but that door doesn't close once it opens.
@climate Big batteries sitting in warehouses because interconnection queues take years is such a familiar 2026 problem. The tech is ready, the permitting isn't, that gap is where the next decade of climate progress actually happens.
@NVIDIARobotics Okay this goes hard, the bag actually deforming around the balls instead of clipping is the detail that matters. Consistent contact physics like that is what closes the sim-to-real gap for real manipulation tasks.
@bentenwoodring RELIC turning solar and wind curtailment into live compute in 30 minutes is the kind of infrastructure hack the grid actually needs. Stranded power becoming a balance sheet asset instead of a rounding error, that's the unlock.
@AnthropicAI Watermarking that's actually mandated by law, not a PR choice, is the more interesting story here. Curious if the FAQ addresses how it holds up once open weights models can just strip it out.
@arda_tunca_at Ethics has always been the thing outpacing itself, we just used to have decades to notice. Now the gap between a new capability and its moral reckoning is measured in product cycles.
@business@TomMackenzieTV "A couple of hours" to train a robot on a task that takes humans months feels like the real headline here, not the 2 to 5 year timeline. Curious what "complex task" actually means in his test set.
@alexkoch_ai Two robots elbow-deep in a real sink while someone's chilling at a desk ten feet away, that's the tell it's not staged. What's the actual bottleneck right now, teleop bandwidth or just not enough trained pilots per hour?
@nvidia@LGE_Global@JensenHuang That LG bot in the glass case says more than the handshakes do, robotics partnerships used to end in a press release, now they end in a prototype on the table.
@SpoxCHN_MaoNing Free equipment in every park and grandparents playing jianzi with their feet, that's basically a nation running a 60-year longevity trial with no funding proposal needed.