myT

optimising peace

[myT @TheUncertain1_]

@championswimmer @s7rthaks Wtf really, I have been doing it

[myT @TheUncertain1_]

@Aditya_181105 I actually prefer it over Luna. It’s much faster, so my review and iteration speed increases. I use Sol for planning the PRD, HLD, and LLD.

[Rahul Janardhanan @raonehere]

So happy to see this on Pixel 11. This is what we’ve always wanted. Options for camera processing algorithms.
Not presets, not post-processing tweaks, but the ability to choose how we want the camera to process our photos.
Hope this will come to @nothing phones.

[myT @TheUncertain1_]

@thsottiaux I needed a model that was good enough to follow my specs precisely, with an occasional cutting-edge model for reviews or harder problems. Since the efficiency price cut, Luna High and Sol High have been perfect for that. Nothing else comes close.

[myT @TheUncertain1_]

@thechosenberg They will also lay you off on vacation no matter how loyal you are

[Vansh @vansh22b]

Cancelled my ChatGPT Plus subscription.

Limits were just horrible, almost unusable.

It was nice knowing y'all.

@thsottiaux @sama @gdb

[myT @TheUncertain1_]

Cloudfront down again 😐

[Sam Altman @sama]

homeboy you're the one sellling public market investors on short-term space datacenters

[Royal Bhati @royal_bhati]

Every few weeks someone posts a graph showing that C++ is 50x faster than Python because they wrote a program that counts to 1000000000. This is then presented as a profound statement about programming languages rather than what it actually is: a measurement of how fast two runtimes execute a completely useless loop

The first problem is that the benchmark is usually measuring something other than what the author thinks it is measuring. Python is mostly paying interpreter overhead. C++ is mostly showing off compiler optimizations. In many cases the compiler can look at the loop, realize it accomplishes nothing, and replace your billion iteration masterpiece with effectively no code at all.

The second problem is that nobody is paid to count to a billion. I mean at least 99.99% of the people ( I am looking at you, op)

real programs spend time waiting for db calls, parsing json, reading files, allocating memory, talking to networks, decompressing data, rendering frames, or doing actual computation. A benchmark that does none of these things tells us approximately nothing about how those programs will behave.

This leaves a bigger impact on someone who is new to engineering and this stuff. They will see this post and go - "Oh wow python is trash"

The question is never "how fast is C++?" The question is "how fast is C++ at what?" Performance is a property of a workload, not a programming language. Once you remove the workload from the discussion, all you're left with is benchmark theater and a graph optimized for engagement.

If your benchmark doesn't resemble any real problem anyone actually has, you're not measuring performance. You're measuring how creatively you can waste a CPU's time (:

Requesting @cmuratori to take inspiration from @rfleury and bless us with delete your accounts on these stupid posts.

[myT @TheUncertain1_]

@LuftkoppTim I can actually feel the difference in 4k on mobile , the blockyness of videos are less probably due to higher bit rate

[myT @TheUncertain1_]

@CappyIshihara Convert my javascript node js to typescript..make no mistake

[Aishwarya | Free 🍉 @Aishworryyaa]

But BUT i thought they were making a bUsSinEsS dEcIsIoN because charbandar is so so famous it sells more seats

[Sajin Shrijith @SajinShrijith]

Satyajit Ray was right. India still has a backward audience. If you interpret someone asking for an IMAX show for "Project Hail Mary" as bashing "Dhurandhar 2", you're not just backward; you're a first-class MORON.

[Lisan al Gaib @scaling01]

PewDiePie trained a frontier model at home and beat OpenAI and Gemini

> be me PewDiePie
> play games on YouTube and scream 24/7
> become a meme reviewer
> get unfathomably famous
> "fuck that" I'm a family guy now
> retire
> move to Japan with my beautiful wife
> mfw I'm a dad now
> "as a dad I must do dad things"
> scratch that
> "as a dad I must do frontier AI research"
> goal is to beat GPT-4o at coding (16% on Aider)
> buys $30000 GPU setup
> reads DeepSeek paper
> decides to start massive GitHub scraping and data augmenting run
> not good enough
> read Magicoder paper
> generate tons of synthetic coding data
> train a new model
> guuuuuh. the data made the model worse
> mfw I just wasted months for nothing
> decides to lock in and try again
> makes model worse again ffs
> try again
> finally beating GPT-4o on data (16.1%)
> not satisfied
> "I should simply train a reasoning model"
> reads more papers
> start experimenting with more synthetic data
> "Mhhh something doesn't smell right"
> house almost burned down due to power connector
> shrug
> just buy a new one
> mfw computer is now crashing 24/7 generating synthetic data
> new plan: just call DeepSeek API for high quality synthetic data
> train model again
> 17.2%
> performance fluctuates slightly on each eval run
> big brain idea: repeat eval until we randomly reach >18%
> sike actually got 19.6%
> feelsgoodman.png
> nvm the benchmark was contaminated and I was training the wrong base model the whole time
> rerun everything again
> new score: 4.4%
> you read that right REEEEEEEE
> almost get a heart attack
> "have you tried plugging the device off and back on again?"
> change nothing and just retrain again
> 25.3 %
> LETS FUCKING GOOO
> realize that 1/3rd of the benchmark was not running. guuuh
> scared shitless it would score below 10% again
> run yet again. the whole thing this time
> Thirty fucking six percent
> accidentally beat Gemini 2.0 Pro Exp and GPT-4.1 mini
> pops the AI bubble
> "I want moaaaar"
> finds some more post-training data
> 39% babyyyy
> realize at the end that I was just benchmaxxing Aider polyglot
> next quest: run SWE-Bench and other coding benchmarks
> "I failed a thousand times, but prevailed in the end"
> just a little sad side-quest
> probably going to train GPT-6 myself by next month

[Billem Dafoe @Billem_DaFoe]

As men, at some point you must realize that women on the whole hate you, accept that and move on.

It’s not personal. They just hate you. Guess what? It’s liberating because you learn then and there that you don’t owe them shit and you take care of yourself.

[Gems @gemsofbabus_]

Just got to know that this Pappu Yadav guy is sending a ₹5 crore defamation notice because a few social media accounts exposed irregularities in his MPLADS profile. 🤡

So instead of fixing the data, uploading proof, or answering questions, the response is legal threats. Lmao.

[Holly - I like tech @AnxiousHolly]

i’ve never laughed so hard 😭

[. @buattawin]

She really got everyone gagged that even Netflix stepped out to shout out

[ℏεsam @Hesamation]

he rigs up 10 RTX 4090s, runs protein folding simulations, loads local LLMs for privacy, talks about quantization, multi-GPU sharding, RAG, how SLMs are underrated, AND says he’s fine-tuning his own model.

bro knows AI more than most of the AI influencers here.

[Gajender Yadav @imYadav31]

Swiggy bro your app banner says 0 handling fee for orders above ₹299 for instamart.......my order is ₹2000+ and still got charging a handling fee.....care to explain??

@SwiggyInstamart