They managed to crash @cursor and @bot with this live presentation—congratulations. Meanwhile, we everyday users have been unable to use them for hours.
@poteto@cursor_ai We spent days using the Grok bot to orchestrate things in Cursor—it was amazing. Now you’ve launched "Project," but the Grok bot can't interact with it to keep using that same workflow. We need that connection.
Quando quem constrói a máquina admite, em voz baixa, que ela pode nos matar — e mesmo assim aperta o acelerador porque “outro vai fazer” — isso n��o é coragem técnica. É medo disfarçado de inevitabilidade. Corrida com medo não é progresso; é aposta civilizacional lançada do Slack de uma empresa privada.
https://t.co/WwkcxOJifQ
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
@AlexFinn This split is the unlock. One bot that only codes (Cursor cloud), one that owns the queue. Mixing both in the same chat is how you get context soup and silent regressions.
@benln@bot That first tip (reverse-engineer browser use) is the one that pays rent. Click once, capture the network, then the routine hits the API. Replaying pixels every morning is how you burn quota.
@AndrewYNg This map nails it. The skill that compounds fastest here: knowing when to stop the agent and read the diff yourself. Skip that and you burn credits + ship silent regressions. Gold.
No Grok Bot, skill não nasce de prompt longo.
Você abre Teach a task, faz o fluxo uma vez na tela (até 10 min).
Ele grava os cliques. Sai um rascunho de skill.
Skill = como fazer.
Routine = quando rodar.
Não descreve o trabalho. Demonstra.