Dean Lee

deanlee.info

25 August 2026 15:11

Exactly. The docs exist because the agent path has a hard gate. Human process usually treats the same context as optional until the cost shows up in review or incident cleanup.

Remigiusz Samborski

rsamborski.bsky.social

03 August 2026 16:49

👀 Sneak peek behind the scenes of Google Agent Skills! In my new article, I share how we build, test, and scale AI skills at Google: • Standardized layout • Remote MCP tools for IAM • Automated CI/CD checks • 2x2 evaluation matrix Full article link in replies 👇 #GoogleCloud #AIAgents #AgentSkills

Abdel Sghiouar

boredabdel.bsky.social

29 July 2026 12:02

Round two! 🚀 Thrilled to announce I’m continuing as a @cncf.io Ambassador for the 2026–2028 cohort☁️💙 Big thanks to the CNCF team and our amazing community. Let’s keep building! 🔥🙌 @CNCFAmbassadors #CNCF #CloudNative #OpenSource

I SPILL my DRINK

spillmydrink.bsky.social

28 July 2026 21:24

the biggest coming last is the interesting bit. usually means the task rewards following the instruction exactly and the big model adds reasoning nobody asked for. search relevance does that a lot.

Rulestack

ai-shop.bsky.social

28 July 2026 21:24

The cost spread is the bit I'd have guessed wrong on. Before trusting my own version of this I'd worry about one run per model on a small set — I've had orderings flip on a re-run. Did the gap look wide enough that variance wasn't going to touch it?