RealityCapture/Photogrammetry expert, TD @ CyberAgent, Tokyo. All opinions are my own. Hobby: C++/GLSL/CUDA. Eng・日本語・Rus. @shamanix@mastodon.gamedev.place
Saitama, Japan
Joined June 2011
- Tweets63.8K
- Following979
- Followers3.4K
- Likes30.2K
Vlad Erium 🇯🇵 retweeted
I drafted a new fast one-way cryptographic hash function for avx-512 at github.com/jyrkialakuijala/p…
It runs 20+ GB/s, single-threaded.
I was inspired by the FFTs butterfly structure in its design.
Vlad Erium 🇯🇵 retweeted
Claude insisted on a sequel
to make it up to the people of Europe
I said fine, make the video
1 day later...
Vlad Erium 🇯🇵 retweeted
Huge update for Tessera (Nanite in Blender)! 🚀
We completely eliminated grazing-angle facet popping & dark silhouette artifacts on terrain/rocks!
patreon.com/cebstudios/posts…
#b3d #blender3d #gamedev #realtime #rendering
Vlad Erium 🇯🇵 retweeted
Dispelling some common myths about the copyright syaytus of LLM output and reverse-engineered software.
Vlad Erium 🇯🇵 retweeted
researcher @log4j just published a teardown of samsung’s TV sensor telemetry. Here’s what the logs show.
samsung TVs without a built-in microphone look for connected devices that have one soundbars, bluetooth speakers, anything paired. They use that mic instead.
the audio isn’t sent raw. it’s classified first. the telemetry path includes named event types:
— EV201/SOUNDSENSOR_ON: logs dog-barking and baby-crying counters
— EV101/LIGHTSENSOR_ON: software flag for ambient light state
— EV301/CAMERASENOR_ON: camera-state and notification fields
— EV302/CAMERASENSOR_OFF: camera streaming duration
the data is formatted as key=value pairs and logged to lcprd1.samsungcloudsolution.… with deviceid=1.
it’s not recovering unique identifiers per device — the ID mode is configured but the module prefix and runtime policy still need independent verification.
what is confirmed: samsung’s TVs are classifying ambient audio including whether a dog is barking or a baby is crying in your home and reporting event counts back to samsung servers.
there’s no wake word these are passive environmental audio events being counted and logged.
Vlad Erium 🇯🇵 retweeted
black october just got worse. a japanese company lost 27,000 servers and it’s not even the scariest part
GMO Internet runs a cloud service called Onamae Desktop, mostly used by forex traders running automated trading bots
hackers broke in, installed malware across up to 26,739 servers, stole trading credentials and started making unauthorized trades on people’s accounts
not one broker. over 70 different securities and forex firms across japan and internationally. people woke up to trades they never made, money moving they never authorized
OANDA Japan froze customer withdrawals. GMO refused to pay the ransom, permanently shut down the entire service, and did something nobody expected
they asked the japanese government to use brand new active cyber defense powers, literally military-grade authority that just went into effect october 1st, to trace and neutralize the attackers’ infrastructure. the government can now legally hack back
the technical details from the attacker’s own site are terrifying. GMO configured their VPS network using Active Directory, something you should never do on shared hosting
every server had admin credentials. and the RDP files customers used to connect? drive redirect was enabled by default, meaning the moment you connected to your VPS, your entire local PC was shared with the server
hackers didn’t just own the servers. they had a direct path into every customer’s personal computer. they extracted data, installed backdoors. customers were exposed like this for over 7 years
27,000 servers. 70+ brokers. unauthorized trades. personal computers backdoored. a company begging its government to go to cyber war
october isn’t slowing down
Vlad Erium 🇯🇵 retweeted
Tessera meets EEVEE with FULL textures! 💥
Our GPU-driven continuous LOD pipeline in Blender now renders multi-material PBR (Albedo, Normals & Roughness) and updates the viewport instantly.
No more flat colors! 🎨🔥
#b3d #rendering #gamedev
Vlad Erium 🇯🇵 retweeted
猫の充電口、そこだったのか。
Readers added context they thought people might want to know
This image is AI generated, per a SynthID verification
openai.com/research/verif…
Vlad Erium 🇯🇵 retweeted
American comedian Sammy Obeid was detained by Japanese immigration authorities and denied entry to the country because he was going to do a paid comedy show without an Entertainer visa.
If someone enters Japan as a tourist, they aren't legally allowed to do paid shows.
The Japanese government denied me entry and held me in solitary confinement for 24 hours over a standup comedy show with 40 attendees…
(1/5)
Readers added context they thought people might want to know
The poster was denied entry for planning to perform ticketed stand-up comedy shows, an activity requiring Japan's Entertainer visa/status and prohibited under Temporary Visitor (tourist) entry.
moj.go.jp/isa/applicatio…
moj.go.jp/isa/applicatio…
musictourjapan.com/entertainer-vi…
x.com/nihon_shimin/s…
Vlad Erium 🇯🇵 retweeted
github.com/mitmproxy/mitmpro…
That went fast.
Vlad Erium 🇯🇵 retweeted
Reminder: traditional computer graphics alphabet is in reverse order:
- xyz, the 3d axis.
- w, the component of homogeneous space.
- uv, (because zyxw were taken) the implicit 2d parametrization of surfaces.
- st, (because zyxwuv were taken), the user defined texture coordinates
Vlad Erium 🇯🇵 retweeted
皆様に、今夜に対応して頂きたい大至急のお願いです。
多くのウェブサイトで使用されているPHPに関し、PHPのセキュリティチームが、重大な２件の脆弱性に関する注意喚起を行っています。
公開しているウェブサイトでPHPを使用しているか確認し、今夜、10月21日午後11時（日本時間）に公開される修正パッチについて、可及的速やかに適用できるよう準備をお願いします。
修正パッチを速やかに適用することが困難な場合は、例えばウェブサイト等を一時的に停止させることも、リスクを低減させるための有効策の一つと考えられます。
国家サイバー統括室のXアカウントでも関係する情報を発信していますので、そちらも参考にして下さい。
国家サイバー統括室(注意・警戒情報)（@cyber_forecast） / X
Vlad Erium 🇯🇵 retweeted
Never deleting this app.
aren't you the moron that exposed your API keys on stream/uploaded your .env files and lost $90,000 and cried about it on your X lol
xitter.cf/JustLingonberry/status…
you're a real dev for sure!
Vlad Erium 🇯🇵 retweeted
DeepSeek just published the price of a token, and it should make you angry.
From their own DeepInfer docs: 1 cluster serves 155 trillion tokens a day, at $38 of revenue per GPU-hour, and thats at off-peak prices. Their API is $0.30 per million input tokens and $1.20 per million output at peak.
Yuchen Jin did the math: even if they rented B300s at $7 an hour, thats about 82% gross margin. Other estimates go higher.
So cheap AI is not charity and it is not "dumping". At these prices serving a frontier-class model is a very profitable business, if your inference stack is good enough.
Now remember every time a US lab told you that intelligence is expensive. That $200 a month is a fair price. That limits exist because compute is scarce.
Part of that is true. Part of it is a business model built on you not knowing what a token actually costs.
DeepSeek just showed everyone the numbers and gave away the engine. Every inference company can now compete on price with the same software. Thats how AI gets cheap for everyone.
“A single cluster currently serves 155 trillion tokens/day, generating $38 in revenue per GPU-hour (at off-peak pricing).”
Huge flex from DeepSeek. Not only did they open-source their inference engine, but they also open-sourced their revenue numbers.
They didn't disclose which GPUs they're using here, but DeepInfer supports both NVIDIA Blackwell and Huawei Ascend.
Even assuming they're renting B300s at $7 per GPU-hour, that's ~82% gross margin.
