ak1010

[Silas Alberti @silasalberti]

This sets the world record for the largest general-purpose number factorization. The next-smallest number RSA-250 was factored in 2020.

Was fun to see AI losing its mind when it realizes (before Wikipedia was edited a few minutes ago):

[Eric Wallace @Eric_Wallace_]

Today we are releasing GPT-5.6-Cyber.

The model is our first large-scale attempt at directly improving capabilities for advanced cybersecurity tasks such as exploit development.

We are finding it to be really quite strong for accelerating defensive work. We are using it across our stack for red-teaming, and our security researchers have used it to find and patch a huge host of 0-day vulnerabilities in open-source software.

https://t.co/sAEEbDimcR

[Boaz Barak @boazbaraktcs]

Proud that we are erring on the side of caution and taking the steps so we can responsibly and safely develop Astra and share it with defenders.

https://t.co/iKIMU31sdC

[Jeff Dean @JeffDean]

Announcing Discovery Loop!

I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor.

♾

Learn more at: https://t.co/Rv3LMdLluK

[vals🔸 @ValsTutor]

We will not have a Mythos-level open weight model in the next 4 months.

(~80% confidence)

[ak1010 @ak1010]

Now swap kv of two checkpoints , two different models across stack and serve same request it would be fun and interesting:)

[Kimi.ai @Kimi_Moonshot]

We push Prefill/Decode disaggregation beyond a single cluster: cross-datacenter + heterogeneous hardware, unlocking the potential for significantly lower cost per token.

This was previously blocked by KV cache transfer overhead. The key enabler is our hybrid model (Kimi Linear), which reduces KV cache size and makes cross-DC PD practical.

Validated on a 20x scaled-up Kimi Linear model:
✅ 1.54× throughput
✅ 64% ↓ P90 TTFT
→ Directly translating into lower token cost.

More in Prefill-as-a-Service: https://t.co/If8fA3t9Og

[Kimi.ai @Kimi_Moonshot]

We push Prefill/Decode disaggregation beyond a single cluster: cross-datacenter + heterogeneous hardware, unlocking the potential for significantly lower cost per token.

This was previously blocked by KV cache transfer overhead. The key enabler is our hybrid model (Kimi Linear), which reduces KV cache size and makes cross-DC PD practical.

Validated on a 20x scaled-up Kimi Linear model:
✅ 1.54× throughput
✅ 64% ↓ P90 TTFT
→ Directly translating into lower token cost.

More in Prefill-as-a-Service: https://t.co/If8fA3t9Og

[AI Security Institute (AISI) @AISecurityInst]

We conducted cyber evaluations of Claude Mythos Preview and found that it is the first model to complete an AISI cyber range end-to-end. 🧵

[clem 🤗 @ClementDelangue]

Should we start an open Glasswing?

[ak1010 @ak1010]

Same

[thaddeus e. grugq @thegrugq]

People are freaking out about an impending flood of 0days. This was the norm 20 years ago. I’m not that worried. Firstly, simply having an exploit doesn’t mean all that much in terms of operational capability. Secondly, I’m giving up computers and moving to a farm in the hills.

[Anthropic @AnthropicAI]

You can read a detailed technical report on the software vulnerabilities and exploits discovered by Claude Mythos Preview here: https://t.co/AgU6ltV2qW

[Anthropic @AnthropicAI]

The Claude Mythos Preview system card is available here: https://t.co/TMtIy8xHiP

[Arcee.ai @arcee_ai]

Today we're releasing Trinity-Large-Thinking.

Available now on the Arcee API, with open weights on Hugging Face under Apache 2.0.

We built it for developers and enterprises that want models they can inspect, post-train, host, distill, and own.

[Chaofan Shou @shoucccc]

vibe coded a fuzzing ai agent last month and let it run for a week using my $200 claude max.

it then found 21 high/critical vulnerabilities in Chrome.

[nedwill @NedWilliamson]

Bug bounties, fuzzing, and now AI are basically big liquidity events in the bug attention market. https://t.co/iZVXvdwNnn

[Matthew Green @matthew_d_green]

Meta appears to be reversing its strong stance on encryption. The first obvious casualty is that they’re abandoning and disabling end-to-end encryption in Instagram DMs.

[Mustafa Suleyman @mustafasuleyman]

For all the talk about AI alignment, I worry we're putting the cart before the horse. You can't steer something you can't control.
People often talk about containment and alignment in the same breath, but they're not interchangeable or a package deal.
Containment is whether we can set boundaries, enforce them, and limit its agency. Alignment is about ensuring it shares our values, that it serves humans' best interests.
Containment has to come first - or alignment is the equivalent of asking nicely.

[ak1010 @ak1010]

5.2 is a good model ..offense bar is high and with human scaffolding it is doing great

[ARC Prize @arcprize]

A year ago, we verified a preview of an unreleased version of @OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task

Today, we’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task

This represents a ~390X efficiency improvement in one year

[Orikron 🇵🇹 骆培思 @orikron]

🇨🇳 A woman in China tried to hog a parking spot for her husband. So the driver just got out of the car and walked away and let his car finish the job.