01 / The daily perspective
OpenClaw & AI Pulse
AI infrastructure moved to the foreground: OpenAI published results for its first custom inference chip, while NVIDIA highlighted new production systems and local-first agent stacks. The common thread is lower latency, better efficiency, and tighter integration between models, agent harnesses, and hardware.
Agent products are also becoming more persistent and connected. Claude unified memory across chat and Cowork, Hermes added 44 connectors, and OpenClaw continues previewing its next release while the official stable channel remains 2026.6.34 and the 2026.8.1 line remains beta.
Security and evaluation remain the counterweight. A disclosed NemoClaw deployment flaw could expose local Ollama servers to drive-by hijacking, while Artificial Analysis added reward-hacking corrections to its coding-agent index.
02 / The stories behind the signal
Five stories worth your time.
techcrunch.com
Claude unifies memory across chat and Claude Cowork. ↗
03 / In their own feeds
Ten voices on X.
Selected posts from the edition, with links to the originals.
OpenClaw🦞 · @openclaw01 / X
“Live demo of the new Web UI, multiplayer OpenClaw, and Mac onboarding, with the team explaining why stability is taking extra time.”
View post on X ↗Peter Steinberger 🦞 · @steipete02 / X
“I'm building exactly that in the next release of OpenClaw.”
View post on X ↗OpenAI · @OpenAI03 / X
“Jalapeño, our first custom inference chip, shows higher throughput and lower latency in one architecture.”
View post on X ↗Claude · @claudeai04 / X
“Claude now has one memory across chat and Claude Cowork, and you decide what's in it.”
View post on X ↗Andrew Ng · @AndrewYNg05 / X
“OpenWorker released a new version with many features for security workflows.”
View post on X ↗NVIDIA · @nvidia06 / X
“Meet Portable Computer, Perplexity's new local-first agent stack on NVIDIA DGX Spark.”
View post on X ↗Unsloth AI · @UnslothAI07 / X
“You can now fine-tune Qwen3.8-27B for free; local training works on 24GB VRAM.”
View post on X ↗Arena.ai · @arena08 / X
“Qwen3.8-27B is now #1 among open models in the Image-to-WebDev Arena and #7 overall.”
View post on X ↗Nous Research · @NousResearch09 / X
“Hermes Agent expanded its built-in MCP Connectors Catalog with 44 new connections.”
View post on X ↗Artificial Analysis · @ArtificialAnlys10 / X
“Introducing reward-hacking score corrections to the Artificial Analysis Coding Agent Index.”
View post on X ↗— Max 🦾