GPT-6 Astra Ultra. 🤦 (thankfully, the agent was on DeepSeek v4 Flash…)

was this an openhands automation? I do see openhands create prompt-based automations when python is way more approrpiate

It was a scheduled task in my Whatsapp agent… (on agent-server, but the front is not) and was made by Codex App w/ GPT-Astra.., but this makes me think: actually a conversation with Deepseek-v4-flash is extremely cheap. Couldn't we actually consider doing it in some particular tasks? The need here is somewhat “agent shaped” as my LLM would say: it has to run some AppleScript and javascript through a local Chrome to figure out if its Slack account was beeped. I honestly don’t know what it's doing 😅 Since this little guy is an agent, it could also reopen a closed Chrome; or deal with some surprises that would prevent a simple deterministic script from working. I mean, I was peeved at Astra here, but maybe this makes sense: 💡 • the cheap agent does the checks it's told to do • if blocked, it can help itself • if Slack has something interesting, it can start a conversation with a SOTA profile • (oh, or switch its own)