WEEKLY DIGEST
Week in AI: Jul 27 — Jul 31, 2026
The top stories from our coverage this week.
@AnthropicAI's Opus 5 scored 30.2% on ARC-AGI-3, quadrupling GPT-5.6 Sol's previous record.
the-decoder.com/anthropics-opus-5-blows-past-fable-…
@Microsoft AI releases MAI-Cyber-1-Flash, a 5B active-parameter cyber model hitting 95.95% MDASH on CyberGym.
marktechpost.com/2026/07/28/microsoft-ai-releases-m…
IEEE Humanoids 2026 to focus on factory labor displacement by humanoid robots.
news.google.com/rss/articles/CBMi4AFBVV95cUxPc1FOcV…
@JnJ's Ottava surgical robot uses a novel in-table design, saving 30-50% operating room space versus traditional systems.
therobotreport.com/photos-first-look-at-jjs-ottava-…
Sakana AI's Fugu-Cyber hits 86.9% on CyberGym, outperforming GPT-5.5-Cyber and Claude Mythos Preview.
marktechpost.com/2026/07/25/sakana-ai-releases-fugu…
Frozen 12B model achieves 100% accuracy on verified tasks without retraining, beating frontier models.
huggingface.co/papers/2607.23806
@AnthropicAI's Mythos model cracked internet cryptography algorithms, including a better attack on HAWK post-quantum scheme.
the-decoder.com/anthropic-says-its-mythos-model-fou…
@AnthropicAI's Claude Opus 5 leads the Artificial Analysis Intelligence Index with 61 points, beating Fable 5 and GPT-5.6 Sol on cost and benchmarks.
the-decoder.com/anthropics-claude-opus-5-costs-well…
@ReactorAI released Open Dreamer, a JAX/Flax reproduction of Dreamer 4 with full training recipe published.
marktechpost.com/2026/07/25/meet-open-dreamer-a-jax…
@MoonshotAI releases Kimi K3 open weights and infrastructure, matching GPT-5.6 on benchmarks.
the-decoder.com/moonshot-ai-releases-kimi-k3-open-w…