A little bit. I write for a living and code in support of my writing. It saves me from reading API docs for the sort of code I'm well able to verify. I also use it to test different ways to phrase something. Recently I used it to turn a diagram I doodle into Django code. It removes a bit of friction and I appreciate that.
I only use the free tier of ChatGPT. It's fine and I'm starting to get a good feeling for its limitations.
I also use Deepl a lot for longer, tedious German texts. I also use Deepl Write to check my writing, especially in German since I am not fluent. I love writing unimportant transactional emails in English and letting Deepl handle the translation. That saves me a lot of time.
I’m paying 200 for Claude code and another 60 bucks for other tools. I don’t need to hire anyone yet or waste my time delegating for same performance but 40x the cost .
The tools are as good as I can make them, so ultimately it’s on me and I like that.
A little bit. I write for a living and code in support of my writing. It saves me from reading API docs for the sort of code I'm well able to verify. I also use it to test different ways to phrase something. Recently I used it to turn a diagram I doodle into Django code. It removes a bit of friction and I appreciate that.
I only use the free tier of ChatGPT. It's fine and I'm starting to get a good feeling for its limitations.
I also use Deepl a lot for longer, tedious German texts. I also use Deepl Write to check my writing, especially in German since I am not fluent. I love writing unimportant transactional emails in English and letting Deepl handle the translation. That saves me a lot of time.
1. What do you do for work?
I'm a product manager.
2. Has AI been helpful? && 3. What tools have been the most helpful?
I use claude projects for lots of different use cases, and I use claude code to prototype simple features. It's generally helpful yes.
4. What tools have been the biggest letdown?
Writing quality.
I've tried:
- meta prompting (asking an LLM to make an LLM prompt to do the task)
- providing examples
- prompt engineering to reduce the amount of obvious LLM giveaways in writing style.
I can't get claude sonnet 4/opus 4 to write with any taste.
There are two things LLMs can't do well yet in my opinion:
- To my knowledge they haven't generated any new or novel insights in any field or topic.
- They haven't produced what I would perceive as high quality writing.
I'd love to see prompts that others feel have produced good writing.
I don't believe people's subjective reports. Even my own. I want more unbiased published studies.
Yeah. AI is a great tool.
I’m paying 200 for Claude code and another 60 bucks for other tools. I don’t need to hire anyone yet or waste my time delegating for same performance but 40x the cost .
The tools are as good as I can make them, so ultimately it’s on me and I like that.