Holy crap, I get phishing emails almost everyday but this is a MASTERCLASS in phishing.
Absolutely brilliant play. Didn't fool me for a second but the thought that went into this one is excellent.
Emotional trigger and urgency trigger.
They're hoping to trigger an emotional response ("wtf are they thinking?!") that will temporarily blind your rational side and cause you to login to the fake site immediately.
1
26
Of all the wearables I use, Whoop has the best AI.
Oura doesn’t have an AI at all.
Of course, I also pull data from all wearables via their APIs for use in my own personal AI health agent.
WHOOP is making it easier to turn your daily habits into real insights with the help of AI. Expect two exciting updates coming to the app soon. Keep reading to find out more. 👇
1
29
This pretty much echoes my own experience with implementing AI.
It's mainly workflows that use LLMs within them for areas which require judgement and decisioning and the rest of the workflow is plain old regular programming.
This produces reliable output. Safe, deterministic output that businesses require.
Handing over the reigns completely to an agentic workflow, at the moment, produces output that can break horrifically at times.
20
If you skip past the reply bots, you can see all the real users are pretty much saying the same thing.
Every other update ships something that breaks. Sometimes catastrophically.
2026.4.29 broke the tui for me. The update before broke Telegram topics. The update before that...
You kinda get the idea.
If this is the future of vibecoding, Zuckerberg won: move fast, break everything.
OpenClaw 2026.4.29 🦞
💬 Group chats feel much better now
📌 Follow-up commitments from context
🔐 Safer exec, pairing, and owner controls
🟩 NVIDIA provider + model catalogs
⚡ Faster startup + plugin/channel fixes
Group chat finally feels agent-native.
github.com/openclaw/openclaw…
55
A Chinese company has restricted their use of their API to coding tools only. That means you can't use OpenClaw or other agentic harnesses on it.
Other Chinese AI model suppliers have quietly raised their prices. Alibaba's coding plans are sold out.
What this means: China is running out of compute. The demand is so high that they have to start restricting use (as much as possible) and increasing prices.
An interesting note: OpenClaw has made clear the issues with "true" agents: the context costs for true agents are far too high.
To perform as general purpose agents they require a LOT of context. They have to hold many things and tools in memory and this increases the token costs of running even a simple general purpose agent.
Coding only agents only need to hold the current project in memory and while they still burn tokens, they don't burn as much as general purpose agents.
Agentic context optimisation is the next progress layer in AI.
33












