@paulg "Which gene editing stocks should I buy?" sounds like a great question for Deep Research. If anyone else is interested, I'll post the result.

We tried .cursorrules at @pulley, but adding a long .cursorrules file negatively impacted the performance of Cursor Agent, so we rolled it back. We're now experimenting with shorter versions to see if that helps. I'll post about the results from those experiments when they come in.

@avichal @pulley

Sadly no! In this case the task was to do some research against a list of startups (about 100 in the list) and collect specific information about their funding status. In this case it was the same bad output as before, but in JSON form. The problem is that it ignores the provided list and comes back with information about 5 major publicly traded companies instead.

@rishdotblog @pulley @perplexity_ai Thanks for the tip, Rishabh, I'm testing it right now on one of the tasks. 🙏

@rishdotblog @pulley @perplexity_ai Oh interesting. So same kind of spreadsheet-style task, but when you ask for JSON output it works? Or different kinds of tasks that have a JSON output?

Einstein purportedly said that if he had an hour to save the world, he’d spend most of it defining the problem properly. While the attribution is debated, the principle stands: clearly defining the right problem is itself a high-leverage act.

@powerbottomdad1 The goal posts never stop shifting. This is the beginning of infinity.

I tried adding a cursorignore file, but no luck there. Is this a rare bug only affecting larger codebases?

furnace broke while it's -27 C outside and i need to buy a replacement for the inducer motor. my first thought was that this would be perfect for @OpenAI's new Deep Research agent! i've tried regular search and @perplexity_ai reasoning mode with r1/o3-mini to no avail. just my luck that Deep Research is down right now. 🤦

best unicorn poll

RT @OpenAI: OpenAI o3-mini is now available in ChatGPT and the API. Pro users will have unlimited access to o3-mini and Plus & Team users…

@MatthewBerman I tried the linked demo but definitely didn’t get 3,872 tokens per second. Seemed to be the same speed as the non-deepseek providers of R1 on openrouter.

ChatGPT has a new feature. So far I'm only seeing this in the web app, not in the mobile app or desktop app.