Runners

Scale Wall

Next-token scale hit the wall. Labs still selling just-add-GPUs are selling FOMO while world models and sample-efficient agents become the real race.

Next-token scale stopped being the whole race. World models and sample efficiency are eating the just-add-GPUs pitch. Companion fantasy deflates while hybrid attention and JEPA-class bets become the real scoreboard.

world-models, sample-efficiency, hybrid-attention, signal-limits, companion-copium

  1. Midwit trap: fluent autocomplete mistaken for intelligenceThe corporate lie of AI as a harmless co-pilot is collapsing first for the people who actually use it. If your daily output can be replicated by a prompt eng...
  2. Meta CTO deflates the companion fantasyMeta CTO Andrew Bosworth is targeting the venture capital consensus that consumers want to date or befriend highly anthropomorphized AI companions, arguing i...
  3. Brute-force next-token AGI fails the wall testThe AI industry is slamming into a hard wall of its own making, trying to brute-force physical agency by scraping more internet text and paying humans to man...
  4. Monolithic model era loses the architecture argumentLlama 3 success pulled Meta's future bets forward and killed the incremental pipeline. Zuckerberg's founder-mode reset is an admission that one launch can em...
  5. Hybrid attention becomes the new scaling standardThe AI industry keeps pushing for bigger LLM context windows by brute force, but Deepseek V4 just dropped a 1M token bomb proving that's a fool's errand. The...

Markdown source