-
Posts
32,168 -
Joined
-
Last visited
manikyath's Achievements
-
some months ago i spent some time vibecoding with an *actual* for-profit vibecoding tool built to make websites, that seemed to have a fairly decent understanding of what it is it was making..
on a friend's recommendation i tried gemini, mostly because i'm not stuck doing 5 free tokens per day...
holy hell it has no f*cking idea what it's doing... i thought the *actual* tool had issues with object permanence.. gemini's just doing crap like resizing stuff to weird sizes, straight up hallucinating things that dont exist, "for each" requests go trough most of them, but not all... menu buttons overlapping..
it's like developing by using an unmotivated junior developer with poor eyesight...
-
I whip chatgpt into making me shit that I am too lazy to do myself. I made myself a msi afterburner plugin to feed data into a 1602 display with esp8266 just today. It would have easily taken me a week to do so myself and it took less than hour to get it right with codex.
But yeah gemini is horrible, even my normie sister doesnt like it.
-
3 hours ago, Levent said:I whip chatgpt into making me shit that I am too lazy to do myself. I made myself a msi afterburner plugin to feed data into a 1602 display with esp8266 just today. It would have easily taken me a week to do so myself and it took less than hour to get it right with codex.
Codex, claude code, antigravity... (and their models), or open code (& something local like qwen3-coder-next q8),
they can all be pretty good for this sort of application:
- Low complexity
- No team
- No maintenance
I use them myself for stuff I don't really care about, and I'm starting from a blank slate.
Because, they rarely produce quality... unless I spend a considerable amount of time describing (or writing plan/docs) on how to achieve the goal, and correcting them along the way.
At which point, like @AbydosOne pointed out, I start questioning am I actually saving any time.
Anyhow, a couple suggestions for your 1602:
1)
It can store up to 8 custom characters in CGRAM.
So essentially you could make 8 icons, or 4 icons (if you want them bigger, by combining 2).
Would look nicer IMO, also would leave more space on the screen for additional metrics.
2)
Backlight is just an LED, you could PWM it for brightness control.
Bonus points if you add a light sensor, so it auto changes brightness.
PS I find it odd that frequency of whatever is at 200 MHz, while usage of whatever (else?) is close to 50%. -
7 hours ago, Biohazard777 said:It can store up to 8 custom characters in CGRAM.
wouldnt make sense as resolution is too low to make any legible custom characters, it makes sense to me as is
7 hours ago, Biohazard777 said:Backlight is just an LED, you could PWM it for brightness control.
if it doesnt receive any data, I just turn off the backlight automatically so i dont really need it.
7 hours ago, Biohazard777 said:open code (& something local like qwen3-coder-next q8),
I tried a whole lot of those, none can do as complex or hands-off programming as closed source ones.
7 hours ago, Biohazard777 said:PS I find it odd that frequency of whatever is at 200 MHz, while usage of whatever (else?) is close to 50%.
4070TiS paired with a 3440x1440p185hz display while watching video. Its not as demanding to trigger the next power stage
-
so.. i'm preparing to join a DnD campaign next week, and i got to say, even with 2024 rules supposedly being more noob friendly, the learning cliff here is seriously intense...
i'm at a point i might actually start putting stuff in excel, just to wrap my head around the workings of the character sheet...
or as Ginny Di on youtube put it: "why did i choose fantasy bookkeeping as a hobby?"

