impce
@impce
SIMD enjoyer, tensor rotator | Software engineer & LLM inference optimizoor | Learning and building micro-SaaS
You might like
First impressions of @1x_tech neo robot, as a mechanical engineer that got to see it for the first time yesterday - It’s lighter than i expected. Only 66 pounds and @radbackwards even let me pick it up! (which was easy for me because i’m strong) - This short king can lift 150…
Seriously though, the bot problem on X needs to be fixed. These LLM response apes larping as humans are insanely widespread. Like half of the comments I read are LLM generated. It’s disgusting. I love this platform because it’s real, and shit like this ruins it.
> be me > exponential curve supervisor > get to work > check-in on the curve > it’s sigmoid
i secretly dream of an AI world where we never discovered RLHF. an entire decade of base models. the beauty we'd have woven already by now immense
I just created a landing page using @MeetGamma in 30 seconds - Prompted "A portfolio website for a freelance web developer" - Reviewed the text, selected a black & white theme for this one (you can make your own) - Hit publish I'm impressed with the result, check it out below:
I move so unexpected that even I can't predict my own actions. My business partner asks "are you joking? I can't tell if you're being serious right now" and I respond "I don't know either"
I've come to the conclusion that the AI risks movement is not just a little bit wrong, but that the whole field has gone off the rails and is operating as a pseudoscience AI risk theorizing that happened before LLMs has been so thoroughly falsified by reality that people have…
I have a 2½ hour podcast with @psychosort and the first 5 minutes are excellent and then the rest is rambling because Brian agrees with me too much How do I edit this?
ADHD hyperfocus rabbit holes should be pursued and prioritized at all costs, no matter how random. Every obsession adds another layer to your unique worldview. Another perspective to pull from and more data points to prop up your strongest skill (pattern recognition). Follow…
ADHD mfs at 3am doing detailed research on the most useless topics known to mankind
people seriously just ignore this they don't care. most are "too good" for manual data labeling. me Label data? But I'm a mathematician! Ideally Sundar Pichai himself should be labeling data. That's how important it is. It's just something that won't happen unless skin in game
google published the first paper I read about using LLMs for multi task, it was bbox and segmentation in 2021. Incredible
I figured out how to get Gemini Image Segmentation to draw little portraits for me with its bounding boxes. I'll explain how down-thread. Here's SpongeBob. (For whatever reason, it likes to draw these portraits sideways.)
genuinely shocking how much progress i've made. you really can create a billion dollar startup in a month using 3d printers, esp32s, cuda and C
alright; i've decided the programming language for my startup is going to be pytorch and C
after i reached a certain level of technical skill i found it easier to start from scratch rather than tear things down, even though tearing things down is more practical most of the time
LLMs die. end of life, no support. the tombstone is a backup in an amazon datacenter. S3 bucket - here lies LLM version 3. eventually, the tombstone gets blown away by some cost saving measure good bye thank you for all the code
qwen3-coder, running locally I had it set up testing infra using minunit and gcov and write some tests on a small ~5000 loc C project. Did it all. 2-3 months ago I tried this with codex, jules, cursor, etc. They all struggled at various parts but eventually did ok. Obviously…
United States Trends
- 1. Draymond N/A
- 2. Sixers N/A
- 3. Aaliyah Chavez N/A
- 4. #911onABC N/A
- 5. Fanone N/A
- 6. Embiid N/A
- 7. Oliver Moore N/A
- 8. City of Norman N/A
- 9. harry styles N/A
- 10. #RHOBH N/A
- 11. Jack Smith N/A
- 12. Sengun N/A
- 13. Jarry N/A
- 14. 76ers N/A
- 15. Jabari N/A
- 16. #Blackhawks N/A
- 17. #911Nashville N/A
- 18. Peyton Watson N/A
- 19. #YIAYhouse N/A
- 20. Paul George N/A
You might like
Something went wrong.
Something went wrong.