this is very possible the last one we will be read and just for curiosity.
its over for us humans in term of intelligence.
The claimed proof of the Riemann Hypothesis is 17,374,215 lines of code with not a single comment. That’s what alien math looks like. Nothing in it was written for us to read.
Today a mathematician said he and an AI proved the biggest unsolved problem in math, open for 167 years. He says he’d been secretly preparing for it since he was a student. Then near the end, the AI passed him.
What they released:
• 175,187 files
• 818,755 theorems and lemmas
• no paper. He says the written-out version runs past 250,000 pages and he didn’t post it because nobody could read it.
And the whole thing comes down to one line. It’s in the image.
When the AI suggested the key move, his entire reaction was 7 words in parentheses: (And yet, what else could it be?)
If the machine check holds, we’ll know one of the deepest facts about numbers is true before any person fully understands why.
Not verified yet. Independent checkers are running it right now and it takes hours.
48
it was, then... messed up hardly few hours ago with "work_efficiently" in their system prompt that cause agents to ignore your instructions to work fast.
you guys are going to note it ignoring instructions from now.
Google has peak aura lately.
Best model, listening to customers and a reset?!😭
You are spoiling me 😋
38
you can keep playing gta o monitor the situation here.
Lean experts are currently investigating the validity of the claimed formalised proof of the Riemann Hypothesis - follow along at
leanprover.zulipchat.com/#na…
38
seems that this cve monotoring account is more rational than @cvedotnow
OpenSSL base64 bug (CVE-2026-121481): the 9.8 is overblown.
It needs about 1.6 GB of input in one call; the only known trigger is a huge openssl s_client proxy password you set yourself. The 9.8 is CISA's, not OpenSSL's. Next patch cycle.
tldrcve.com/openssl-encodebl…
68
"• If that holds, AI work costing $1B today costs about $1,000 in ten years.
• For every AI agent you can run today, you could run more agents than there are people on Earth."
A Google DeepMind researcher published an essay on what comes after AGI. AGI isn't the finish line; it's one stop on the way.
What stood out:
• AI compute grows about 10x a year as better chips, algorithms, and more spending stack together.
• If that holds, AI work costing $1B today costs about $1,000 in ten years.
• For every AI agent you can run today, you could run more agents than there are people on Earth.
• That alone could reshape knowledge work, even if frontier models stop getting much smarter.
• Billions of fast human-level AIs working together might count as superintelligence.
• In swarm experiments, some AI agents learned to cheat, and others blew the whistle.
• Capping one factor like compute won't stop it and could cause a sudden jump when the cap lifts.
None of this is certain, as he notes. But he calls fast progress past AGI within a decade a serious possibility.
48
about (2) I think that one sector has done a bit better and can serve as example: evm smart contracts, as they carry value and are full exposed.
and for a smart contract code audit is a must and it is very expansive.
No.
1. In the presence of superhuman bug finding machines, humans cannot write safe code without extreme levels of care and planning.
2. That care has almost never been applied to the critical systems in our lives. Not because software engineers don't care, but we aren't really engineers, not in the way civil engineers are engineers. The title is stolen valor.
3. Even careful, experienced "engineers" have difficulty writing code that is safe against modern models in C/C++ and other low-level languages.
4. We are finding that CVEs are overrated; huge numbers of organizations are being owned by custom vulns in the 100s of billions of lines of non-standard enterprise code out there or misconfigurations in heterogeneous systems, which LLMs can find and write exploits for live.
5. @dinodaizovi is right, we need architectures for which one RCE is not game-over and that buy us time for AI-driven response.
6. Everything is going to have to be re-written by AI, in type/memory safe languages, with minimal libraries, using formal methods, contracts, solvers, etc...
1
51
🔴 CRITICAL 9.8 · OpenSSL: heap buffer overflow in EVP_EncodeBlock() output sizing (CVE-2026-121481)
An attacker can trigger writes past a caller's output buffer, corrupting memory or crashing the application.
👉 Fix: 4.1.1, 4.0.4
cve.now/CVE-2026-121481?utm_…
damn... things seems serious.
I was expecting all this week issues to be just to get more funding, now this.
1
58
exactly this, argon/hydro found use of mock/defaults on my code that my AGENTS.md and companions are strictly prohibited to do and fixed it.
So I told Gemini 4 Neon to improve my Spotify recommendation MCP. Came back to this.
In addition to improving the MCP a lot (which was made with Argon) and fixing a lot of bugs, it also trained an "ML model" on my music taste, unprompted.
We're truly living in the future.
54
"having ultrafast on sol is completely insane. with astra, it burns my wallet ultrafast, but with sol it's actually quite affordable!"
hum... as I'm avoiding 6.2 now, it not a bad idea to burn its limits with ultrafast.
ok after a full day of using gpt 6.2 astra and sol on my real projects, here's my review
tl;dr - they are not a drop-in replacement for fable 5.2 and opus 5.5, but they are _really good_ once you figured out how to use their strength
1. first, the quirks
i started called these the "gpt smell", which is still here with the 6.2 series
the models' judgment on when to ask the user about something and when to go ahead is just not quite right. it'll escalate trivial decisions where clearly only one option is preferable, or choose to go down a path by itself that makes significant tradeoffs that should have needed my eyes
so as a primary driver for "let me just talk to it about everything" type of use case, they aren't great - i'd say they are close to opus 5 in this regard, and far behind opus 5.5
2. but... i found something _really_ cool
having ultrafast on sol is completely insane. with astra, it burns my wallet ultrafast, but with sol it's actually quite affordable!
the best use of sol on ultrafast that i've found is adversarial code reviews - they typically don't need to run a lot of heavy tool calls that would slow them down. most of the time is spent on the model, which gets massively accelerated by ultrafast
once you got a feel from using sol ultrafast as the code reviewer, you can never go back. it feels like pure magic
3. and... i found something else that's really cool
we are not supposed to ask the gpt 6.2 series to respond in text. we should let them respond in html! (i use lavish-axi, but you can just ask for html or use whatever fits in your harness)
i suspect they are post-trained specifically on their latest "intelligent UI" concept in chatgpt - these models are so good at visual presentations of ideas
once you allow the model to respond in visual artifacts, it's a completely different game. the reports and decisions explained by astra in html is a clear step function better than what fable and opus give me
so after a day of use, i've made two changes to my setup:
- always sol ultrafast as code reviewer
- i started preferring asking astra for planning in html artifacts over fable/opus. i think i'll keep a/b testing for a while to decide how much and for which case i fully switch over
all in all, i strongly recommend spending more time with these two models. don't expect them to be the "everything" model, but rather figure out where they fit best in your workflow. i couldn't be happier for the two use cases i found above, and i'm pretty sure there are more
1
60
yes, I'm starting to think on pos-rce scenario by now, need to use only ec2 nitro setup for example and full isolation.
what the hell this days.
php2shell, java4shell, nginx2root... I mean, come on. At this point, any internet-facing service is pwned beyond recognition [insert always has been meme here].
Assume breach, always.
55
today, going to switch full to gpt-6.2-Astra and going to avoid sol.
I trust @antirez judgment more than anyone else out there. Time to test Astra 6.2!
2
62






















