Ceoz (@Ceoz_1) header
Ceoz (@Ceoz_1) profile picture

I have a pretty good idea about what I am not doing.

Joined November 2025
Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)Photo by Ceoz (@Ceoz_1)
Your universe has no meaning to them. They will not try to understand. They will be tired, they will be cold, they will make a meal out of your beautiful oak porch.
486
2,601
147
37,386
594,774
absolutamente nada vai acontecer btw. 90% de chance inclusive dessa historia ser mentira.
Meu tio trabalha a anos na funai, ai hj a gentr conversando na casa da vovó,perguntei p ele oq ele achava disso cm a experiência dele nas aldeias.. as palavras dele foram "minha filha a gente ta mto lascado" ta babado viu..isso é so a ponta do inferno
75
5
62
5
554
11,389
CARALHO TIME SKIPPEI ONTEM ERA QUARTA E HOJE É SEXTA (eu tenho que começar a dormir)
9
No ironically? This is OG christian value...
"The Christian shoemaker does his duty not by putting little crosses on the shoes, but by making good shoes, because God is interested in good craftsmanship."
10
I made Opus 5.5 go manic and it created a folder called "Icarus" It's writing assembly code as we speak, I will bring updates once it stops.
14
I WANT MORE
14
Ceoz retweeted
larp it till u r it
14
252
10
2,433
30,865
Ceoz retweeted
Press the button. Press. Press it. Perish it.
something disgusting just happened i was working with opus and fable until a SONNET subagent directly asked me a question without going thru the parent session it has the audacity to think it could be worthy to ask me a question
18
156
5
2,895
60,616
6.2 is a very intelligent model and people are over indexing on its poor communication ( though this should of been fixed before release ) Feels similar to o3 though even o3 wasn’t as autistic.
Wow, wow, wow! I am really stunned by what ChatGPT (Astra 6.2 just did). Essentially it not only proved a conjecture my collabotor and I have been trying to prove, but also discovered new phenomena, which we never even suspected, and made new conjectures (probably hard to prove)
34
17
2
464
16,605
eu amo ela pprt
Replying to @ergodicflux
se eu nao estiver te comendo ou vc não estiver me pagando não é urgente
1
37
desculpa não responder há duas semanas estou ocupada existindo como um membro funcional da semiperiferia na terceira modernidade
4
17
114
2,054
O NOTEBOOK DANDO RÉ E O FDP NÃO SAI DE TRÁS MANO PQP GENTE BURRA ME ESTRESSA PPRT
esses notebooks do banco master não sei não
39
It's not anything damaging like that, btw. It's his hyperbole that bothers me. To give you an example of what Astra usually fucks up in communication, I think a real case is better than a made-up anecdote: I asked Astra to build a verifier for a project I'm planning to rewrite. The verifier's goal was to ensure parity with the previous versions in terms of input/output behavior. I warned it that one specific function depended on the particular behavior of C#'s heapsort, so we'd have to account for that. It hyperfixated on the heapsort case, built a verifier just for heapsort, and stopped. I then asked, "What about the rest?" It started iterating again and finished. After that, I asked whether it had tested the verifier to ensure parity. It said it hadn't, because it had nothing to run the verifier against except the baseline (which was correct, and a fuckup on my part). I then pointed out that it could use the database history instead, since all the inputs came from there and all the outputs were stored there. It proceeded to do exactly that and actually found a bug in the original code using the verifier, based on a single database record, which was a plus, tbh. I then asked whether that proved the verifier's consistency, and it said no, because it hadn't run the verifier against the code. I was confused but too lazy to figure out what it meant, so I just asked it to run it. It came back saying the verifier returned [UNPROVEN], and I was like, "What the fuck? What did it verify, nothing?" And that's exactly what I asked next. Unsurprisingly, it answered that yes, it had run the verifier against an empty input because it had no candidate to check against. It returned [UNPROVEN] as its default state for empty inputs. "Well," I thought, "at least that's good error handling." So I asked it to prove that the verifier worked as intended (dumb on my part, but again, laziness). It proceeded to do its shenanigans and came back with TWO [UNPROVEN] results. I asked about them, and it explained that the first was from the empty input, while the second was from checking against the legacy code. "What the fuck?" I thought again, and asked how it had run the legacy code. And, TL;DR, it had turned part of the prover into a static checker (??????????). Which, to be fair, is impressive as fuck, but... WTF? No biggie, really. It just overdid it because it got lost trying to figure out what its actual end goal was supposed to be. I never explicitly told it to refactor or test against the prover. I just kept pushing it in no particular direction, and it kept complying, which, with a model this comically intelligent, can lead to behavior that's either infuriating or hilarious.
Read this to the end, I promise it’s worth it. Imagine that you’re ordering food at a restaurant. You: “I’ll have the steak medium rare with fries” Waiter: “What about the fries?” You: “…huh?” Waiter: “do you want the fries medium rare too?” You: “idk what that means just cook them normal” Waiter: “okay got it, coming right up!” 15 minutes pass, waiter comes back out Waiter (panic’d): “hey man, we REALLY need to know if you want your fries cooked at 413 degrees or 471 degrees. Whole kitchen is held up until you decide!” You: “wtf just do whatever the chef wants” Waiter: “okay got it!” 30 minutes pass. The waiter comes out with mostly passable fries, and an entirely burnt steak You: “wtf happened this steak is burnt!” Waiter: “so sorry, the chef got distracted when waiting for your answer about the fries. We can put the steak in the fridge to cool it off if you’d like?” This is how it feels to use GPT-6.2 Astra.
1
1
76
I mean, it shouldn't be that hard to reproduce the hooks of the .exe to startup the game no?
> wake up > take a shit > get out of bed > check beep boop > over 9000 DMs > "SCHMEELI, GTA VI LEAK" > ??? > sends link > torrent > ok > look inside > actual assets > huh > squint eyes > look closer > not larp huh, how about that. the cyberleek guy released the full xbox gta 6 build, goop free. only missing launcher .exe for it to fully work, says will drop on 23rd. if it follows thru on his word then ppl can play gta vi on pc on Monday, mod it, etc.
47
a carinha dele
18
2
1,408
20,869
Ceoz retweeted
In the next version of Bun bun check gets 1.5x faster on small projects
4
10
1
271
7,906
LMAO
Jim Cramer was wild for this one $NVDA
1
29
I'm telling you all, It's the 4o psychosis all over again and now with an Anthropic model.
I don't really think Neon is all that good. Spent ~$100 on a few dozen builds last night and Opus-5.5 comes out (far) ahead on all of them. Google is indeed cooking, but it's mostly the frequency of their recent releases. This particular model is overhyped.
1
47
they might actually have achieved RSI lol!
No way! ANOTHER Gemini incoming? Give me a pause Google. What are they doing right now?! Every release is a banger. And Neon keeps surprising me how damn good it is!
18