Meronix

i just wanted to thank the whole twitter community. Everyone that i follow provides useful content and without them i would not be there where i am today. Crypto Twitter is a phenomenal place. Thank all of you guys!

@ChinmayKak Already contained in it for better numerical stability

revising pytorch and TIL (again) why you don't apply softmax before F.cross_entropy()

@nayshins Actually it is usdt

@Meronix15 you mean usdc

You know if software is solved and anything is programmable this is pretty bullish for programmable money. The most lindy programmable money is ETH

@CUDAHandbook Thank you

@Meronix15 See e.g. the commentary on how IA > AI: cs.unc.edu/~brooks/Toolsm…

BREAKING: The CUDA Handbook text is now available on the website, cudahandbook.com! Svbstack article in the first reply.

@CUDAHandbook How do you think can you aquire those architect skills apart from just writing a lot of software which i have been trying to do as much as pissible?

@Meronix15 In the right hands, the models are massive force multipliers. They do shift the landscape in favor of the folks Brooks called “architects” over code typists. AI is the stiffest-ever test of the Jevons Paradox that has characterized software engineering for decades.

@FakePsyho That is actually a pretty fair assessment. I feel more bleak about the future though with the recent models, they are just really good.

@Meronix15 My diplomatic answer is: I don't think "computer science" people should fear more than any other knowledge workers.

gpt 5.6 sol pro is the very first model that routinely solves some of the harder logic puzzles and so far looks like it actually uses logic rather than pure guessing so far my experience is that the improvement 5.5->5.6 is bigger than 5.2->5.5

@thdxr When you are gangprompting and suddenly codex spawns 5000 subagents

why does this remind me of our team

@atelicinvest And why is it that all these people that do this have vapoware products with subscription services lmao

@atelicinvest The final boss of AI psychosis

What in the actual f?

@justinskycak It‘s true i have high grades but only like 15% were meaningful where the exam was actually hard. Low signal to good grades these days

The biggest scandal in education isn't even the low test scores. It's the high grades.

@usr_bin_roygbiv Yeah personally i have done a project that would haven taken around 1000 employees for a year. The task was finding all the vibecoders some bitches

I've seen multiple examples from myself and others i've spoken with of 5.6 doing things which previously would have taken entire companies or departments of people years of work just a year or two ago in one shot. People have no idea how large the skill and knowledge gap now is.

@AgustinLebron3 Probably best advice haha

@Meronix15 Get offline. Meet real people in person. The world doesn't stop when we close our phones and computers.

This is a microcosm of how AGI will play out. Humans will rent-seek themselves into the economic equation. Newly-necessary supervisors-of-AI will emerge everywhere. reason.com/2026/07/09/rho…

@AgustinLebron3 Yeah true but i think with AI the exploration of what to write is also getting automated and you just can brute force the solution. Basically it is now pay to win to solve problems and recent awtf heuristic and algorithm wins just show this

@Meronix15 As literacy spread in Middle Ages, the job of scribe stopped existing. Today, AI is an accelerant for the spread of computer literacy. Then and now, just being *able* to write lost value. But knowing *what* to write became even more valuable.

@DanielW_Kiwi Do you actually think this will be the case especially with the recent awtf heuristic and algorithm wins by antrophic feels more like humans will be just left behind as the so called slop is just more incomprehensible by us but still the solution is correct

I suspect my future job is going to involve a lot of getting non technical people unstuck when using AI to write software.

@arielzwolinski @FakePsyho I am currently building on my side job but how long until i am actually hurting by being in the loop?

@Meronix15 @FakePsyho Get anthropic or openai subscription Build Make content about your journey

I'm amazed that someone saw the news and that was their conclusion

@DDhfrcycyf5vuf @FakePsyho @scaling01 Yeah i think there is definitely coming a point where no humans are needed to guide software but who knows when that comes

@Meronix15 @FakePsyho @scaling01 Depends if you can realistically pivot Job markets not great but it will hang in there until theres literally zero value in humans guiding software, which is, who knows how long

Quick summary of what happened during AWTF in Japan: - OpenAI crushed humans in both Heuristic and Algorithm categories. Heuristic is the one where I won in 2025 and OpenAI placed 2nd - AWTF is pretty much the highest level competitive programming contest: invited finalists in both categories are among the best people in the world; this is essentially the first time AI won vs humans in a programming competition in such a decisive matter - big change compared to 2025 is that AI was constantly progressing and didn't get stuck / plateau at any point - System used for Algorithm is mostly just custom harness + model very close to gpt 5.6; my educated guess is that system used for Heuristic is a custom "autoresearch" harness with a swarm of cooperating agents and a massive inference cost - OpenAI was sure that they will win Algorithm and were very confident that they will win Heuristic; they did a lot of backtesting on old contests and they would win all of them in their simulations - imho heuristic problems are a great proxy for ML autoresearch capabilities; if AI was able to match best humans here, we're very close to RSI / automated researcher; this result is way bigger than a high score on some questionable benchmark When I get some rest, I'll try to post more thoughts about the whole thing. Thanks for following the coverage. I believe I tweeted more in those few days than I did in the first half of 2026. And I'm not even counting the 10 hour long livestream. Oh and technically speaking, I didn't participate so I'm still undefeated and I'll gladly keep "Humanity's Last Programmer" in my bio.

@Jon85Ma @FakePsyho @scaling01 It is really interesting for me. I care about good software and good code just feels so hopeless when you fight against models that just ingest 10x the amount of knowledge in a fraction of a time

@Meronix15 @FakePsyho @scaling01 Is it interesting to you? If yes, get really good, care about your craft. The job may change, but in the worst case someone will still have to pick the tradeoffs, give directions, etc Also, if you get really good at something, it makes it easier to get good at another thing.