Anthropic Finds a Conscious-Like Layer in Claude
New Anthropic research identifies a "global workspace" inside Claude — a small slice of active representations that the model can reason with, mirroring how only a fraction of human brain activity reaches conscious awareness. The team calls it "J-space" and says it can be read, audited, and shaped to keep models trustworthy as they grow more capable. They've also partnered with Neuronpedia on an interactive demo of the methods on open-weights models.
SpaceX Shares for 2 Million Kids
SpaceX is donating 2 million shares to Trump Accounts for 2 million children, to be held until they turn 18 — framed as making every child a shareholder in the American Dream. Brad Gerstner will appear on CNBC Monday from the White House ahead of a joint NYSE/Nasdaq bell ringing to celebrate the launch.
AI Code Review Consensus Hardens
Developers are converging on a blunt take: less manual review, more automated validation. Theo argues you should be reviewing a much smaller share of your code than five years ago, and that critical code deserves more automated tests, not more hand-checking. ThePrimeagen agrees — if code is important, the answer is more automation around it, not less.
Fable Buzz + Apple's AI Hardware Bet
Fable is drawing rare genuine enthusiasm from developers and researchers, with multiple builders calling it "insanely good" and worth paying for when it leaves their current subscription tier. Separately, Apple and Broadcom have extended their chip partnership through 2031, with all signs pointing to the deal covering Apple's first chips built specifically for AI servers.