Humanity’s Sixth Sense is a new benchmark for intuitive visual reasoning. It includes 522 open‑ended tasks covering images and video. The tasks test spatial, causal, and social understanding.
Key points
Benchmark name: Humanity’s Sixth Sense
Task count: 522 open‑ended tasks spanning images and video
Sources
Rive_app Balloon-Popping Game – New Version Released
The balloon-popping game from @rive_app has a new version. It now includes six arrow types and three bosses. A link to the game is provided for testing and feedback.
Key points
Feature: Six arrow types added.
Feature: Three bosses added.
Sources
- Original post
- Linked resource - Linked in the post
MiniMax AI M3 Prompt Caching Live on SambaCloud
MiniMax AI M3 now supports prompt caching on SambaCloud. Cached prefixes reduce time‑to‑first‑token by 35–88 % and can speed up processing up to 4.7×. Input cost drops about 90 % and cached tokens cost $0.06 per million. No code changes are needed.
Key points
Speed: TTFT improves 35–88 % and overall speed up to 4.7×.
Cost: Input cost reduces 90 % and cached tokens cost $0.06 per million.
Sources
- Original post
- Linked resource - Linked in the post
CoreWeave Agent Lens – full‑corpus trace logging
CoreWeave Agent Lens logs all agent traces and reads the entire production corpus, not a sample. It groups conversations that fail in the same way. The post cites 40,000 overnight conversations with 230 failures that received no alarm.
Key points
40,000 conversations processed overnight, 230 failed.
Agent Lens reads the full production corpus and groups similar failures.
Sources
Monad Private Settlement announced – privacy on public blockchain
Monad Private Settlement is announced. It claims confidentiality on a public blockchain natively. It lists speed, finality, composability and real‑world access control for privacy and compliance.
Key points
Announcement: Monad Private Settlement is introduced.
Claim: Provides native confidentiality on a public blockchain.
Sources
Replit builds and runs apps locally on Windows with sandbox containers
Replit announced that on Windows it builds and runs applications locally. Each build runs in its own sandbox. The sandbox uses Microsoft Execution Containers and Nvidia OpenShell.
Key points
Build environment: Windows local builds run in isolated sandboxes.
Sandbox technology: Sandboxes are powered by Microsoft Execution Containers and Nvidia OpenShell.
Sources
Hebbia launches Headless Hebbia API and MCP server
Hebbia announced Headless Hebbia on 1 October. It is an API and MCP server. It brings Hebbia’s retrieval, orchestration and analysis into custom applications and existing AI assistants.
Key points
Announcement: Headless Hebbia announced on 1 October.
Offering: API and MCP server for retrieval, orchestration and analysis.
Sources
Liquid AI releases Open d1 decision models
Liquid AI released two open-weight decision models called Open d1. The d1-3B model handles text and vision, while the d1-omni-600M handles text plus image or text plus audio. Both read a state plus typed questions and return calibrated probabilities in a single forward pass.
Key points
Model: d1-3B (text + vision)
Model: d1-omni-600M (text + image or text + audio)
