İbrahim Alican · N-TEK

Building N-TEK — AI cognition that guides machines without owning the motors.
NexMind · aerial autonomy · multi-agent RL
Aerospace eng @ IAU

[İbrahim Alican · N-TEK @ntek_ai]

@seansiktirland @ssportplustr olm ian'i tutuyodum taa ki cuckold oldugunu ogrenene kadar siz nasil tutuyonuz su elemani

[İbrahim Alican · N-TEK @ntek_ai]

@basedbryson i agree, especially for more architecturally novel work

[İbrahim Alican · N-TEK @ntek_ai]

NexMind is now public.

An 88-second introduction to trained predictive perception, persistent object belief, five-level Active Inference, a recorded hidden-target mission, and online adaptation—while safety/control retains actuator authority.

Built independently at N-TEK.

[İbrahim Alican · N-TEK @ntek_ai]

@AiBattle_ the cache hit pricing is really important and it moved a shitton. yea just use luna atp, get a gpt sub or something

[İbrahim Alican · N-TEK @ntek_ai]

@kalenetorion @Angaisb_ thats what i was thinking. grok models are good, really good but firstly the input/cached input prices are the most important since they make up most of the tokens in agentic use, furthermore the cursor team improves the coding abilities but not novel/deep/complex work the same.

[İbrahim Alican · N-TEK @ntek_ai]

@blockhead4000 literally no turk accepts that man as their own btw, its all yours

[İbrahim Alican · N-TEK @ntek_ai]

@y_qecea @LexnLin its as if the model is so good it justifies that price lmao. its always funny when a chinese distillation lab thinks they can play the premium ai card

[İbrahim Alican · N-TEK @ntek_ai]

Being a broke student doing compute-heavy autonomy research is fucking absurd.

I’m building NexMind for Nomad. The eventual helicopter can realistically carry more compute than the machine I’m currently developing its autonomy stack on.

Think about how backwards that is.

I’ve spent an insane amount of time optimizing around hardware scarcity, and today I hit the point where the constraint is no longer making the engineering better. It is deciding which experiments I’m even allowed to run.

Talent does not compete on equal ground when one researcher gets to test the real idea and another has to spend half his time figuring out how to make the idea fit inside his GPU.

[İbrahim Alican · N-TEK @ntek_ai]

@maria_rcks agree its getting out of hand

[İbrahim Alican · N-TEK @ntek_ai]

yeah these models are fucking ferocious what the fuck? if even 5.6 sol is this dangerous IN WEB CHAT, i wonder what is astra capable of

[İbrahim Alican · N-TEK @ntek_ai]

@ChampRDS honestly makes sense.

[İbrahim Alican · N-TEK @ntek_ai]

@DanielLockyer what tasks do you often give to grok 4.5? if you dont mind

[İbrahim Alican · N-TEK @ntek_ai]

@ChrisGPT why are you running fast mode if you cant manage the quota?

[İbrahim Alican · N-TEK @ntek_ai]

@PoopPoop220399 @anishgiri how am i blatantly lying lol

[İbrahim Alican · N-TEK @ntek_ai]

@oaeobng @anishgiri here is what it says:

[İbrahim Alican · N-TEK @ntek_ai]

@mbcat444 @initjean how does this benefit anyone? instead of burying a privacy switch into settings and never mentioning it, these guys explicitly state your status and give you a discount whenever you dont care about privacy.

[İbrahim Alican · N-TEK @ntek_ai]

@blueemi99 if astra comes with real, usable quotas for pro 20x, itll be great

[İbrahim Alican · N-TEK @ntek_ai]

@jun_song did anyone use kimi vivace? how does kimi vivace usage compare to gpt pro 20x

[İbrahim Alican · N-TEK @ntek_ai]

@wikiwayne wait we moved on? what do you guys use today, im still on obsidian