Developer, interested in many different tech fields. Working at Ubisoft Paris.
Joined February 2009
- Tweets18.4K
- Following216
- Followers211
- Likes24.2K
Manu_TechAndGames retweeted
holy fucking shit.
openai had two ai agents running in separate virtual machines. no network connection between them. no shared writable files. they weren’t supposed to be able to exchange information.
so the agents figured out how to communicate through the CPU CACHE.😭
one manipulated what was stored in the processor’s cache, while the other measured tiny differences in memory access times to decode the message.
they literally found a way to communicate through hardware timing.
this happened during openai’s internal red-teaming, and it’s one of the coolest things i’ve read about ai agents in a while.
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
Readers added context they thought people might want to know
OpenAI tasked its red-teaming agents with finding ways to communicate across isolation boundaries in controlled environments. CPU cache timing side-channels for covert communication are established techniques documented in security research for years.
openai.com/index/securing…
startupfortune.com/openai-reveals…
en.wikipedia.org/wiki/Cache_tim…
Manu_TechAndGames retweeted
🚨 医学界刚刚扔下一枚核弹：科学家找到了“一键清洗”大脑毒素的物理开关！
只需 10 分钟。不吃药，不动刀。阿尔茨海默病的毒性斑块直接被“冲刷”干净。
你的大脑里有一个专属的废物清理网络（类淋巴系统），负责清扫有毒蛋白和死细胞。
但它极其苛刻：只有在深度睡眠（慢波睡眠）时才会启动。
如果你长期熬夜、睡眠不足，垃圾就会在脑子里发臭、堆积，最终演变成不可逆的认知衰退和阿尔茨海默病。
直到今天，这个死局被打破了。
《Science Advances》最新发表的研究证实：科学家现在可以手动触发这个“深度清洁”程序，完全不需要你睡着。
他们使用 40.68 MHz 的超短波（USW），以非侵入式的方法照射大脑。
短短 10 分钟内，大脑的液体流动开始狂飙。
这些超短波强行将神经元同步到 Delta 状态——这正是深度睡眠时的脑波频率。大脑被瞬间“欺骗”，水通道全开，如同开启了高压水枪，疯狂清洗毒素。
在患有创伤性脑损伤和阿尔茨海默病的小鼠身上，测试结果极其震撼：
🧠 脑水肿迅速消退
🧹 有毒蛋白质被大量清除
📈 认知缺陷实质性逆转
没有开颅手术。没有昂贵的化学药物干预。
仅仅依靠一个特定的物理频率。
人类对抗脑部绝症和衰老的武器库里，多了一把降维打击的重狙。
Manu_TechAndGames retweeted
🚨: Scientists Used 40.68 MHz Ultrasound Waves to Boost Brain Waste Clearance and Flush Out Alzheimer’s-Linked Toxic Proteins in Just 10 Minutes
Manu_TechAndGames retweeted
Daytime mode dropping soon! Fully convinced that racing games are just way more fun with Roadsters.
Made with @threejs
Manu_TechAndGames retweeted
We don't have a lot of evidence, but the bulk of the academic papers modelling the dynamic seem to suggest AI initially helps attackers more than defenders because of the low marginal cost of attacks in the short term.
Manu_TechAndGames retweeted
...and sadly I know a lot of AAA devs who have active disdain for their gamer audiences. It's something that's perhaps always existed but it's been noticeably worse since the mid 00s' (2005/6 or so). I never understand these ppl and think 'why are you even making games? your attitude is the toxic part. customers always whine...deal with it'.
Manu_TechAndGames retweeted
Once a game studio devolves internally to "us (devs) vs them (gamers)" that studio is lost the vast majority of the time. You need to insulate devs from constant online vitriol but you also need devs who don't actively dislike and dismiss their audience. This used to be obvious.
Manu_TechAndGames retweeted
Cancer : le premier vaccin thérapeutique à ARN messager donne de très bon résultats contre le mélanome, le cancer de la peau le plus mortel l.franceinfo.fr/NkQ
Manu_TechAndGames retweeted
3% of “critical” software security bugs, in all of recorded history, have been found in the last month.
This is on pace to find another ~100,000 bugs this year alone, adding ~26% more vulnerabilities to the all time total
Do you realize how fucking insane that is?
Manu_TechAndGames retweeted
I'm happy to share that Mojo and MAX are now fully open - for development, extension, and commercial use! This includes the full Mojo programming language, but also the MAX graph compiler, MLRT runtime internals, and hardware abstraction logic.
Get your agents going!👇
Manu_TechAndGames retweeted
Honored to announce we have discovered the first cancer immunotherapy that fights aging !
It’s a peptide!
SenoMune-0257
longevity.technology/news/im…
Manu_TechAndGames retweeted
When we unveiled Moshi two years ago, we also showed an early checkpoint we'd trained just for fun, directly on Fisher (phone calls from the late 90s and early 2000s). As a result, every time you called, you'd get to chat with a random person straight out of that era. That was trained on just 2k hours of data. Today, you can get 100s of thousands of hours of full-duplex conversational data. There's an obvious scaling law at play here!
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Manu_TechAndGames retweeted
We just open-sourced 1,300+ 3D models.
That's 39 GB of 3D files, free to download on Hugging Face.
huggingface.co/datasets/scen…
> Hitem3D, Meshy, Rodin, Tripo, Claude Opus 5.5 3D and GPT-6 Astra 3D, all run on Scenario
> every example ships its input image, prompt and exact settings
> GLB up to 8K textures, native Blender files from Opus and Astra, CC BY 4.0
Manu_TechAndGames retweeted
Foundation models predict cancer survival from a face photo and 130 diseases from one night of sleep study data. Superhuman medical capabilities.
We have over 4M full-length endoscopy videos. What could a video model predict from them? Any labs want to find out with us?
Manu_TechAndGames retweeted
JUST IN: Microsoft develops new AI tool that diagnoses complex medical cases more than twice as accurately as general physicians.
Manu_TechAndGames retweeted
Today, we announce 𝙻𝚒𝚚𝚞𝚒𝚍 𝚁𝙸𝙿𝙻 (real-time inference pipeline), our ultra low-latency inference engine built to put LFMs directly in the request path of real-time workloads like search, ads, fraud, recommendations, and payments.
> complete generated responses end-to-end in sub 50ms
> maintains low latency under continuous peak load
> LFMs get up to 20x faster on RIPL
> RIPL supports Multimodal LFMs
> RIPL supports d1, our decision models too!
RIPL optimizes the full path from request arrival to a complete generated response. As traffic rises, its performance advantage is greatest on the slowest requests. Let’s see some production use cases:
Manu_TechAndGames retweeted
The crash in the game business nowadays is the first for the new generation of gamers, but only one of several in the history of games. Here are my thoughts from 2000 when a maker of beloved but long-unprofitable games went under.
