Parham (@par_ham) header
Parham (@par_ham) profile picture

Hit me with that fire 🔥 bored of “professional” Tweets... except those coding Tweets, they icy

Amsterdam
Joined March 2013
Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)Photo by Parham (@par_ham)
I'm putting a few of my future predictions here. Just as a reference, but feel free to comment :) #predictions #future #decade
2
In the age of cheap typing; pen and paper thinking is critical
36
My theory on @AnthropicAI Mythos... It's not that it's too good it's that it's another "finisher" model good at a narrow task. It's not a "worker" model that follows agentic flows and so @AnthropicAI is embarrassed to release it so they're trying to make some marketing content
38
Hide your secret sauce and make your UI simple
29
LLMs automated two things; assuming they're the same thing is what gets you in trouble: 1. Automated typing 2. Automated copy/paste, run loop
35
Tonight we're eating koobideh and tahdig. 🇮🇷
2,365
1,555
230
18,110
771,634
Update the 🇮🇷 emoji @nikitabier @elonmusk
124
300
26
4,188
437,271
Replying to @2irl4u
Give me a few hours
997
1,148
322
11,889
1,421,425
Is Nvidia... same as Intel in the 2000s? It would be soooo funny if yes
35
Why are all the new cars lately so fuqin ugly da fuq
26
Django + LLM = 🌶️
44
Why the hell does cursor keep putting back random files from other chats? If anyone familiar with VSCode shenanigans could let me know I'd appreciate. Context: I only use Cursor for it's agent mode and do the rest in Vim.
53
GPT 5.1... so far pretty good
52
Calling LLMs AI is a disservice to it's automation powers of being a bridge of Natural Language and the Machine. Calling it AI means the focus of it's training is to be faking intelligence rather than following the best probable path for it's instructions.
1
71
I think that's why GPT-5 is now pretty trash and Sonnet 4.5 is winning in enterprise.
1
57
I rarely open OpenAI anymore for anything other than a Google replacement. I've actually started Googling a little more lately. I no longer use it for any sort of "task" when I need output to be sculpted around my instructions.
47
LLMs are punch card machines and the prompts are the punchcards. They are not AI until the punch card machine can self update the punch card machine. P.S. Don't tell me that's what prompting is... it's not. k thanx bye
1
1
109
* Probabilistic Punch Card Machines
51
Mobile "Apps" (except messaging) will become redundant as the world revolts against the disease that they are within 10 years. Web will be dominant again as we revolt against attention demanding phone. #future #predictions
1
LLMs are punch card machines and the prompts are the punchcards. They are not AI until the punch card machine can self update the punch card machine. P.S. Don't tell me that's what prompting is... it's not. k thanx bye
40