Anish

647 posts

Anish banner
Anish

Anish

@nebulaanish

https://t.co/qCQEo5VePN

Kathmandu, Nepal Katılım Temmuz 2022
362 Takip Edilen109 Takipçiler
Sabitlenmiş Tweet
Anish
Anish@nebulaanish·
I feel the need to learn History and legislation of my country. I realized that I know nothing to have an opinion of my own.
English
19
120
737
38.9K
Anish
Anish@nebulaanish·
blue tick is too attention grabbing for the type of shit I post. sub cancel garda pani hatdaina ta ajhai.
English
0
0
0
29
Anish
Anish@nebulaanish·
Cool site. Hopefully I get to travel some districts of eastern Nepal by the end of this year.
Anish tweet media
English
1
0
2
103
K
K@iiamkrshn·
ZXX
8
79
2.6K
54.6K
Anish
Anish@nebulaanish·
@N_ambush lol true, I have downgraded cc to keep using 4.6. Slightly better.
English
1
0
1
28
jheulayo
jheulayo@N_ambush·
DAMN, the token burns way faster than i have assumed man, gotta wait for next 4 hours. fml
English
1
0
1
74
Anish
Anish@nebulaanish·
@N_ambush Thursday ho sir. I'll most probably skip this one.
English
0
0
0
17
Anish
Anish@nebulaanish·
Hamstring injury, arggggh. I guess nomore futsal this week. 😞
English
1
0
1
124
Anish
Anish@nebulaanish·
Most of the software engineering infrastructure is built to solve the fundamental problem of search. I can't unsee it now.
English
0
0
0
54
Anish retweetledi
Aakanchya 🔮
Aakanchya 🔮@scarabwitchh·
the lack of empathy in people is astounding yes the clearing of the sukumbasis was inevitable, the problem is not the clearing itself but the government's failure to manage them before bulldozing their houses
English
6
15
106
2.5K
'(Bibek Panthi)
'(Bibek Panthi)@bpanthi977·
Today I learned about Post Training LLMs: Here we take a base model and improve it for conversation, reasoning and domain tasks using supervised learning and RL. 1. Mid-training trains base model on a mix of domain specific and general dataset before moving to SFT. 1/3
English
2
0
4
181
Anish
Anish@nebulaanish·
What weird timing is this game happening in. I didn't know aaja Nepal ko game thiyo.
English
1
0
2
111
Anish retweetledi
Tyler
Tyler@rezoundous·
Opus 4.7 is insane guys. It one shotted my session usage limit.
English
601
1.3K
31.5K
1.1M
jheulayo
jheulayo@N_ambush·
Missing bounties :(
English
1
0
3
105
Anish
Anish@nebulaanish·
If corporate world is RL environment, middle managers are reward hacking. Because they get high scores without actually getting any work done.
English
0
0
1
111
Anish
Anish@nebulaanish·
This opus downgrade feels like a setup to launch opus 4.7.
English
0
0
2
141
Anish
Anish@nebulaanish·
@N_ambush Why are you rage baiting me
English
0
0
1
37
jheulayo
jheulayo@N_ambush·
@nebulaanish yeah but i think i hardly Belong to yellow, or strongly yellow
English
1
0
0
102
Anish
Anish@nebulaanish·
The net worth of ministers was made public and it reminded me of the Maslow's law. You need a level of privilege to start worrying about nation's problems or say the top hierarchy of problems according to Maslow. Most people can't afford to work actively to solve them. But having said so, it becomes counter intuitive to think of that some of the most radical changes have been brought about by people who couldn't afford not to solve those.
English
1
0
5
153
Anish
Anish@nebulaanish·
I tried qwen-code free version(with coder-model Qwen 3.6 plus) and it did a good work for the most part. It is quite capable. I would say it's comparable to gemini-cli but definitely behind claude code. Some of it's decisions were dumb and hence interesting to look at. 1. For instance, there were some build issues with this next js app and it asked permissions to re-do the whole project instead of fixing the actual issue. 2. A more interesting one was when I asked it to kill process running in 3000 port and re-start the server. It decided to kill every node processes and killed itself. And this suicide happened more than once.
English
0
0
2
192