PulseAugur
EN
LIVE 01:28:26

Anthropic AI models show signs of global workspace; Gemini 3.5 Pro leak surfaces

Research from Anthropic suggests that AI models may have spontaneously developed a "global workspace" that supports functions associated with conscious access. Separately, a leak claims Google's Gemini 3.5 Pro will rival Fable5 and GPT-5.6, though the source expresses skepticism about the claim's magnitude. AI

IMPACT AI models may be developing complex internal architectures, while competition intensifies with new model releases and leaks.

RANK_REASON The cluster contains commentary on research and a speculative leak about a model release, rather than a primary announcement.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Anthropic AI models show signs of global workspace; Gemini 3.5 Pro leak surfaces

COVERAGE [2]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    ‪Ethan Mollick‬ ‪@emollick.bsky.social‬ · Some really interesting research from # Anthropic that # AI models have spontaneously developed a workspace that "appe

    ‪Ethan Mollick‬ ‪@emollick.bsky.social‬ · Some really interesting research from # Anthropic that # AI models have spontaneously developed a workspace that "appears to support the functions associated with conscious access" Demo of how this works: https://www. neuronpedia.org/qwen…

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    @intheworldofai NEW Gemini 3.5 Pro LEAKS! # Google Is Back and Will Rival # Fable5 & GPT-5.6? Nah not so much ... https:// youtu.be/-z2QiGjtOzk?si=g8674c _m2bkg

    @intheworldofai NEW Gemini 3.5 Pro LEAKS! # Google Is Back and Will Rival # Fable5 & GPT-5.6? Nah not so much ... https:// youtu.be/-z2QiGjtOzk?si=g8674c _m2bkg6c6A # AI # LLM 🗞️ 7/5/26