Any engineer fighting the use of LLM’s will ultimately find that their peers are achieving more in a shorter time then they are.
How do you measure that?
I am asking because I tried all the AI coding tools (copilot, cursor, claude code, etc.), even the paid versions, to find the utopia that doubles my productivity where lots of people are talking about. But everytime I find they take more time to generate code than me typing it, they rarely get things right the first time, they produce way more code than needed (which is much more to reason about) and all in all it takes longer to achieve something using “AI” instead of doing it on my own.
The worst results are the ones that seem to be correct, but can only be found incorrect on closer inspection. For instances, a part of my job is to write data collection applications for sensors; those sensors come with a datasheet (usually 100 to 200 pages) that explain how to “talk” to the sensor. Sounds perfect for an LLM - throw the datasheet in, get a small C application out. But either I get something that does not even compile, or something that seems to collect data, but configures the sensor completely wrong. Debugging that takes more time than just working through the datasheet myself and program it myself.
Yeah, my work is forcing us to use Claude for everything now. More than ever, I’m convinced that using AI is slower and worse than me just doing it by hand.
UNLESS, if I blindly accept the output, then Claude is faster. Sure, I can close the ticket in 15 minutes. Is it right? Idk. Is it good? Idk. Does it have bugs? Idk. Did it make shit up? Idk. Did anything even get compiled or tested? Idk. The ticket is closed though and that’s all that matters, apparently.
Well that’s because copilot and cursor are pure shit, and claude code is overpriced and over rated.
TBH I have been hopping between agent frameworks, but mostly I use locally run Qwen or Deepseek to draft individual function blocks or brainstorm in a notepad interface:
And manually insert them as I review the code… I like my stuff uncomplicated and unobstructed.
If I were to pick an API provider today, I’d buy one year of Xiaomi MiMo, because it is dirt cheap and good enough. GLM and Deepseek are great, but more expensive since they are popular now.
How do you measure that?
I am asking because I tried all the AI coding tools (copilot, cursor, claude code, etc.), even the paid versions, to find the utopia that doubles my productivity where lots of people are talking about. But everytime I find they take more time to generate code than me typing it, they rarely get things right the first time, they produce way more code than needed (which is much more to reason about) and all in all it takes longer to achieve something using “AI” instead of doing it on my own.
The worst results are the ones that seem to be correct, but can only be found incorrect on closer inspection. For instances, a part of my job is to write data collection applications for sensors; those sensors come with a datasheet (usually 100 to 200 pages) that explain how to “talk” to the sensor. Sounds perfect for an LLM - throw the datasheet in, get a small C application out. But either I get something that does not even compile, or something that seems to collect data, but configures the sensor completely wrong. Debugging that takes more time than just working through the datasheet myself and program it myself.
Yeah, my work is forcing us to use Claude for everything now. More than ever, I’m convinced that using AI is slower and worse than me just doing it by hand.
UNLESS, if I blindly accept the output, then Claude is faster. Sure, I can close the ticket in 15 minutes. Is it right? Idk. Is it good? Idk. Does it have bugs? Idk. Did it make shit up? Idk. Did anything even get compiled or tested? Idk. The ticket is closed though and that’s all that matters, apparently.
Well that’s because copilot and cursor are pure shit, and claude code is overpriced and over rated.
TBH I have been hopping between agent frameworks, but mostly I use locally run Qwen or Deepseek to draft individual function blocks or brainstorm in a notepad interface:
https://github.com/lmg-anon/mikupad
And manually insert them as I review the code… I like my stuff uncomplicated and unobstructed.
If I were to pick an API provider today, I’d buy one year of Xiaomi MiMo, because it is dirt cheap and good enough. GLM and Deepseek are great, but more expensive since they are popular now.