👉🏻👌🏻 (@pr0n) header
👉🏻👌🏻 (@pr0n) profile picture

Living in Truth 活得磊落真誠

Joined November 2007
Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)Photo by 👉🏻👌🏻 (@pr0n)
👉🏻👌🏻 retweeted
Same performance. New world. OTOY Studio VFX Tools: drop in your footage, describe the world you want, and run it on the canvas. Every movement, face and camera move stays exactly as you shot it. Only the world around it changes. Up to 4K, HDR and 4:4:4 color space, built for the highest quality film work.
7
20
2
171
10,678
👉🏻👌🏻 retweeted
Every driver needs their crew.
1,005
1,494
414
21,121
6,036,314
👉🏻👌🏻 retweeted
“Hong Kong” Single Artwork by : Christopher Doyle
386
1,643
233
12,725
554,791
👉🏻👌🏻 retweeted
Built by Gemini 4 Hydrogen purely in single-file code...no external libraries, no skills, no image generators, and zero embedded assets.
131
121
43
2,701
133,134
👉🏻👌🏻 retweeted
Meet Pika H3 Flex, debuting today as the #1 open weight model on DesignArena’s t2v board. It’s built for professional iteration, with exportable audio stems, resolutions up to 4k, and faster-than-realtime generation. It can produce a 15-second clip with soundtrack in as little as 7 seconds. So, yeah. It’s kind of a flex 💅
398
390
403
4,871
1,533,825
👉🏻👌🏻 retweeted
𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 brings video understanding, step-by-step reasoning, and spatial understanding to local assistants. Ask questions about a recorded lecture, a walkthrough filmed on your phone or a clip from a camera feed. It reasons through visual problems involving charts, mathematics and multiple steps. It can also locate objects and interpret their spatial relationships, providing a foundation for robotics applications. > Across 62 vision benchmarks, it averages 71.4, comparable with Gemma 4 12B’s 71.4, with just 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > Our new in-house vision encoder uses about one-third of the compute of the SigLIP2 encoder. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 model card: huggingface.co/LiquidAI/LFM3… 5/
3
11
5
251
44,635
👉🏻👌🏻 retweeted
Today, we’re introducing Ideogram's first vector model. Generate clean, editable SVGs with smooth gradients and sharp typography from a text prompt, or turn any image into a detailed, scalable vector. Live in Ideogram and the API: ideogram.ai/models/vector
303
328
302
1,833
614,464
👉🏻👌🏻 retweeted
If you have an existing Cursor, Grok, and or 𝕏 subscription, here's how to unify/combine all your accounts into the new 𝕏 Subscription: When you switch from your current plan, eligible monthly and annual plans are cancelled, and you get a prorated refund for unused time. The refund goes back to your original payment method separately from your checkout total, so your new plan is charged at full price. If you pay for Cursor through the Apple App Store, switching doesn't cancel that subscription, and 𝕏 can't refund it. Cancel it in your iPhone or iPad Settings so Apple stops billing you, and request any refund from Apple. Switching is optional. If you don’t switch, your current plan continues unless you change or cancel it. Existing Cursor, Grok, or X subscribers can switch to The X Subscription and unify eligible accounts at cursor.com/unify (Cursor), accounts.x.ai/unify (Grok), or on x.com (X) via the in-product banner. If you already pay for Cursor, Grok, or X, switch on the web at one of these links. For now, the mobile apps send you to the web to switch. If you don't pay for any of them yet, you can set up a new account and subscribe to The X Subscription in the Grok, X, Grok Bot, or Cursor mobile apps, if desired. These purchases are billed through the App Store or Google Play Store. Switching unifies eligible Cursor and Grok accounts into one account with one workspace, one plan, and one bill. Adding a personal X account gives you X benefits under The X Subscription. A completed unification is permanent. It can’t be undone, and Support can’t undo it for you. After you unify, you manage your plan, billing, and settings at grok.com/dashboard. When you switch, your old Cursor, Grok, and X Premium subscriptions are cancelled automatically, so you don’t need to cancel them yourself. Any refund for unused time is sent separately. If your X, Grok, and Cursor accounts use different email addresses, they don’t need to match. 𝕏 will connect the X account you sign in to, not an account that shares your email. When you switch, sign in to the exact Cursor, Grok, and X accounts you want to combine, and check each one before you confirm. Keep work and personal accounts separate unless you mean to combine them. If the wrong account appears, stop before checkout and contact Support. Learn more: docs.x.ai/x-subscription/uni…
SpaceX has just officially introduced a new "𝕏 subscription" that combines Grok, 𝕏 and Cursor into one plan. Here are the tiers & info: • Plus: $8/month • Premium: $30/month • Super: $100/month • Ultra: $200/month Plus: $8/month • Cursor access • Grok Chat, Voice, Imagine & Build • Grok API • Shared usage pool across products • Premium checkmark • Fewer 𝕏 timeline ads • Creator monetization • Edit posts • Longer posts & video uploads on 𝕏 • Articles & Media Studio • Advanced Analytics • Higher posting & DM limits • Bookmark folders • Download videos Premium: $30/month • Includes everything in Plus, and: • 4x more usage compared to Plus • Grok Bot • Latest Grok models • Faster Chat responses • Al images & videos • Ad-free scrolling on most surfaces on 𝕏 • Preferred @handle • Higher reply boost Super: $100/month • Includes everything in Premium, and: • 15x more usage compared to Plus • Priority model access at peak times • Faster Chat responses & generations Ultra: $200/month • Includes everything in Super, and: • 50x more usage compared to Plus • Earliest access to all products & models • Fastest Chat responses & generations Compare features: x.ai/pricing
239
152
50
1,978
205,004
Statement on Anthropic’s Report Revealing China's Use of Artificial Intelligence for “Stability Maintenance” and Public Opinion Surveillance On September 10, @AnthropicAI published "Detecting and Countering Misuse of AI: September 2026", which explicitly identified “Teacher Li Is Not Your Teacher” @whyyoutouzhele as one of the targets of a “stability maintenance” and public opinion surveillance operation. While reviewing Claude user activity, Anthropic discovered and banned a cluster of accounts it assessed to be linked to local public security and state security authorities in China. The users used Claude to collect information automatically, query government surveillance databases, and generate daily reports for their employers. Our account was included in the surveillance operation disclosed in the report. Following @OpenAI's disclosure of similar operations targeting dissidents earlier this year, Anthropic has now exposed these surveillance mechanisms from another angle: our news account, which speaks up for people inside China, is being treated as a surveillance target by the state security apparatus. The report also revealed that Chinese public security and state security authorities label and perceive information about social problems, labor protests, and news coverage as political threats requiring intervention. In another operation that Anthropic assessed to be conducted by a government contractor, users employed Claude to produce automated “public opinion briefings,” processing approximately 15 to more than 30 articles a day. They rated content for “political sensitivity” and reframed labor protests, student activism, minority communities’ demands, and foreign media coverage as threats to “political stability” and “ideological security.” Some documents also recommended government enforcement action. One of the most direct examples involved a user assessed to have ties to the public security system asking Claude to produce a “stability maintenance” report. The resulting report recommended “control” measures targeting ten Chinese petitioners, including intercepting petitioners before they could lodge official complaints, coercive questioning, and close monitoring of their movements and communications. Another local state security bureau used Claude to track overseas dissidents and human rights organizations, gathering information such as the assembly point and route of a pro-democracy march in Vancouver and the venues of Uyghur cultural events in Turkey, and write an internal operation manual for other personnel in the bureau. In a separate operation targeting Uyghurs in Syria, operators analyzed large volumes of group-chat messages to identify exploitable vulnerabilities, including financial hardship and family separation. They paid particular attention to people with relatives still in Xinjiang and conducted covert recruitment outreach to individuals who might have access to local Uyghur armed groups. The report did not confirm whether recruitment succeeded. The operation also involved locating civilians and planning the suppression of Uyghur journalists. Religious communities were also subjected to surveillance. The report covered Catholics, Tibetan Buddhist communities, Falun Gong practitioners, and the Presbyterian Church in Taiwan, among others. The dossiers collected personal histories, social relationships, and exploitable “points of leverage,” and even included floor plans of religious venues. Beyond the surveillance operations targeting dissidents, the report also said that some Chinese AI companies exposed users’ sensitive information while conducting unauthorized model distillation. It named seven Chinese AI companies, including @Alibaba_Qwen, @Moonshot, @deepseek_ai, @ZhipuAI @jietang, and @XiaomiMiMo, alleging their involvement in unauthorized model distillation or related data collection. In some of these activities, sensitive data submitted by users was also forwarded to the US company Anthropic. According to the report, Moonshot forwarded some Kimi users’ requests to Claude, served Claude’s answers back to those users, and saved some of the exchanges for training. The material included surveillance records submitted by a user assessed to be potentially affiliated with the People’s Liberation Army, drawn from hundreds of cameras in Chengdu and used to track a specific individual. It also included internal code and active access credentials submitted by engineers working for Chinese state-owned enterprises. The report also stated that Xiaomi submitted saved user conversations and coding sessions to Claude to generate training material. The information included users’ names, contact details, and corporate data. Separately, engineers developing a case management system for a municipal Public Security Bureau in China used DeepSeek, and their requests were forwarded to Claude to help develop a tool that uses citizens’ national ID numbers to compare their movements against police records. These disclosures raise serious questions about user privacy and AI safety standards. The report warned that unauthorized distillation could enhance dangerous cyber or biological capabilities, while Claude’s existing safety standards would not transfer along with them. In our view, there is an unavoidable issue at the heart of US–China AI competition: the institutional differences between democracy and authoritarianism will profoundly shape how this technology is used, and whether ordinary people can refuse, question, and exercise oversight over that use. We have repeatedly seen AI being used to make China’s domestic surveillance and “stability maintenance” systems more efficient, with some operations already extend overseas. As these tools mature, the possibility that these methods will be replicated in other countries and adopted by more authoritarian governments is a risk we must confront. The democratic AI we support should protect people’s expression, privacy, and freedom to make their own choices. It should also allow the public to hold developers accountable. Every company must demonstrate this through its actions. Publicly exposing such abuses and banning accounts involved in them are responsibilities companies should fulfill. We thank Anthropic for its disclosures and hope they ultimately translate into protection for the people affected. In the face of continuing repression, we remain committed to the principle of open information and will continue working to ensure that more people can see the realities of Chinese society. Team Teacher Li
We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them. We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies. These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve. We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop. Read the report: anthropic.com/threat-intelli…
25
49
1
365
115,082
👉🏻👌🏻 retweeted
It seems they didn't release the RVQ tokenizer encoder. There is no way to turn real audio into the codes the model speaks, so you can't teacher force it on your own music, making full finetuning impossible. Unless they release the tokenizer.... Or someone distills a clone... 😉
🎵MiniMax-Music3 Next-Generation Open-Weights Production-Ready & Versatile Music Model huggingface.co/MiniMaxAI/Min…
25
19
4
264
24,556
👉🏻👌🏻 retweeted
Never forget.
New broadcasting guidelines for women's athletics have been released to crack down on camera angles and slow-motion replays that ‘sexualize’ athletes The guidelines highlight shots broadcasters are urged to avoid
810
2,185
152
64,616
6,867,232
👉🏻👌🏻 retweeted
I'm excited to share that Qualcomm is acquiring Modular: this will accelerate our path to unifying accelerated compute with an open platform. This will also mark a new era in open software development for Qualcomm. 👇
Hardware plus software defines leadership in the AI era. Today we announced an agreement to acquire @Modular, advancing our evolution as a developer-first AI solutions company delivering generative and agentic AI from edge to cloud. bit.ly/44rBpe5
84
104
34
1,268
165,931
👉🏻👌🏻 retweeted
I want to say a final thing about my Fable first reaction: I dedicated my life to programming and I'll use every innovation in the field, also to extract value and bring it to the local inference world, to Redis, and so forth. But:
70
219
23
4,001
1,101,547
👉🏻👌🏻 retweeted
I asked Fable to invent a new color, and I got my first "chat paused". It did however decide to pursue a strategy of shining lasers in your eyes to trigger otherwise impossible cone activations 🤯
16
9
4
251
11,343
👉🏻👌🏻 retweeted
What's crazy to me is that Fable is blocked from life sciences broadly, nerfed even if you get passed the classifiers and filter level blocks. The whole point of AGI/ASI is to cure all diseases. Everything else is just nice to haves. But Anthropic wants to close off that path. I think Anthropic might be the worst company on the planet.
400
367
103
5,172
237,006
👉🏻👌🏻 retweeted
We have been working closely with @nvidia to ensure Hermes Agent works smoothly on their new @NVIDIARTXSpark superchip and integrates with the new OpenShell runtime, which connects Hermes to @Microsoft's security primitives. Watch our feature in the big announcement at Computex:
306
600
162
6,568
5,988,698
👉🏻👌🏻 retweeted
More than a decade of work coming to an end. I was in my twenties when we started, now in my forties. But we did it. We finished the story as it was meant to be finished.
Introducing The Talos Principle 3. Coming 2027 to PC and PS5. Are you ready for one last journey?
66
102
2
1,403
55,873
👉🏻👌🏻 retweeted
the 3.44% is your context window btw 😅😍 Grok's official answer on rate limits: There are none!
3
1
41
2,568
👉🏻👌🏻 retweeted
In the age of AI slop, some of us are still making things by hand, I promise ❤️ Octocat for the new GitHub Copilot App was modeled, rigged, and animated in Blender...
Cooking up something new 🧑‍🍳 Join the waitlist for early access to technical preview of the GitHub Copilot app 👇 gh.io/github-copilot-app?utm…
203
1,622
104
38,853
2,415,237