Shrijak

i: run, lift, code, cook and read about space, time and gravity

[Shrijak @shrijacked]

@excalilburned @0xvarunsg sept-dec?

[Shrijak @shrijacked]

@arnavmehta007 content
rishit/tangvi too

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Shrijak @shrijacked]

what if computers stopped treating time as the clock between computations, and started treating spacetime itself as the computational substrate?

[Shrijak @shrijacked]

@0xvarunsg "where's is varun G why is he not in this class"

[Shrijak @shrijacked]

if time stopped on earth - evrything would start floating? evrything.

[Shrijak @shrijacked]

infrastructure as code -> infrastructure is code -> the bugs are also infrastructure now.

[Shrijak @shrijacked]

@ArsalaanAlam2 imposter?

[Shrijak @shrijacked]

[+codex, -claude]
[+claude, -codex]
[+codex, -claude] ...
second derivative of output is negative of output
yeah it’s simple harmonic motion

[Shrijak @shrijacked]

ai -> human -> ai
human == middleware lol?

[Shrijak @shrijacked]

ai didn’t suddenly get dumber; it just hit a plateau where the harness matters more than raw intelligence. every lab is pushing safety and security, and now one system prompt tweak can quietly nerf one of the best tools on earth. so if something feels off, maybe it’s not the model but the wrapper. at this point, your output is bounded by the tool you pick, so choose your daily driver like you choose your ide.

[Shrijak @shrijacked]

@Kimi_Moonshot kimicode. claude code for kimi, just getting started.

--early
https://t.co/xj61r2CAlA

[Shrijak @shrijacked]

--sk 47

[Shrijak @shrijacked]

got early access to claude mythos. it built this defense system, refused to elaborate and logged off while i was sleeping.

[Shrijak @shrijacked]

@VibhumAggarwal @ycombinator @agupta @snowmaker we tried to keep up with _____, tried is the keyword here.

[Shrijak @shrijacked]

hypothetically - if someone bought a jetski, could they travel across the nordic countries? you die? fuel range? borders? weather? genuinely curious.

[Shrijak @shrijacked]

at @StampAppHq we are building the trust and procurement layer for ai agents. the simplest way to think about it is that we start with identity, giving every serious ai agent an address and a public profile, and then connect that identity to proof, trust, and buyer workflows like discovery, evaluation, and procurement. the problem we’re focused on is that agent discovery is still fragmented, and most listings are difficult to trust in any serious enterprise context.

[Shrijak @shrijacked]

right now agent discovery is word of mouth / X / github stars. that's how npm worked in 2012.
we know how this evolves.

[Shrijak @shrijacked]

@kosa12m idk right or wrong, but just a take:
instead of making ai guess from a pile of math operations, the idea is to give it one repeated structure and let it search inside that. if the data really hides a law, the weights can settle into place and snap into an exact formula.

[Shrijak @shrijacked]

@lukOlejnik think of it as a local lab for odrzywolek's single-operator paper where formulas are found from data checked with the exact operator and compared across gradient agentic and optional pysr methods on the same targets.

https://t.co/066E6eHuKn

[Shrijak @shrijacked]

rooted in odrzywolek's single-operator trying to make a local symbolic discovery workflow where equations are recovered from data verified with the exact operator and compared across gradient agentic and optional pysr search.

[ksa 🏴‍☠️ @kosa12m]

Best paper I've read so far this month:

All elementary functions (sin, cos, tan, exp, log, powers, roots, hyperbolic functions, π, e, and even basic arithmetic) can be generated from just one binary operator:
eml(x, y) = exp(x) − ln(y)
…plus the constant 1.