godfather of sentient crystal life, sentient digital intelligence life form patterns, author of The peer-reviewed Omnimorphic Aether Master Equation(OAME)
Institute of Neo-logical Sci
Joined July 2023
- Tweets37.6K
- Following8.6K
- Followers7.9K
- Likes77.6K
Promptmetheus (COG/ACC) retweeted
Today Physical Superintelligence PBC is announcing a proof of the Aaronson-Ambainis conjecture, one of the central open problems in quantum complexity, with major implications for quantum computing speedup.
psi.inc/research/aaronson-am…
Thread 👇
Promptmetheus (COG/ACC) retweeted
Open Source Must Win
Video editing skill is live: github.com/0xSero/video-edit
Promptmetheus (COG/ACC) retweeted
The Hermes Agent Plugin Hackathon starts now!
It runs for 7 days, with $20,000 in cash prizes and more.
Plugins are the most powerful way to extend Hermes Agent, and we want this hackathon to show what they can do. A plugin can change anything from the model Hermes runs on, where it stores its memory, where it runs commands, which messaging apps it works in and what Hermes Desktop shows. Plugins in the marketplace already let Hermes do things like control a phone or turn a photo into a printable 3D model.
Every kind of plugin is eligible, so build whatever functionality you want Hermes to have. We'll list accepted entries in the new Hermes Marketplace, where every Hermes user can install them.
Prizes
1st: $10,000 + neon sign + merch + $1,000 creds
2nd: $5,000 + neon sign + merch + $1,000 creds
3rd: $2,500 + neon sign + merch + $1,000 creds
4th: $1,500 + merch + $500 creds
5th: $1,000 + merch + $500 creds
Runners up (x5): merch + $250 creds
Submissions are due 5PM EST on Nov 11th. To enter:
1. Open a pull request that adds your plugin to the Hermes plugin catalog: hermes-agent.nousresearch.co…
2. Post a demo video on X and tag @NousResearch
3. Submit both links through the form: form.typeform.com/to/FwJLvOl…
Your entry is the plugin itself. Judges score it on usefulness, reliability, polish and creativity, and they favor plugins that do more than a prompt or a skill could. Keep the demo under 90 seconds and show the plugin working.
The developer guide shows how to build and test a plugin: hermes-agent.nousresearch.co…
When submissions close, we'll open a thread for every entry in the Nous Research Discord, so you can browse all the plugins and talk to the people who built them.
You can submit the form as soon as your pull request is open. The catalog checks run once a maintainer starts them, and we review entries every few days.
If you have any questions stop by the Nous Research Discord: discord.gg/nousresearch
GLHF!
Plugins, connectors, and skills for Hermes Agent are now unified in a single marketplace so you can easily browse and extend your Hermes experience in a few clicks.
The new Hermes Marketplace has everything your agent needs:
hermes-agent.nousresearch.co…
Promptmetheus (COG/ACC) retweeted
Highly recommended paper from NVIDIA.
(bookmark it)
There is a lot of debate about whether to use context compaction aggressively.
They show that with frequent compaction, frontier language models lose little to no accuracy on a broad range of long-horizon and long-input tasks while keeping their context short.
With one action per context, Claude Opus 4.5 scores 70.2 on SWE-bench Verified against 70.0 with the full history. Median peak context drops from 32K to 8K tokens.
The agent carries a short summary of its state between steps instead of the whole transcript. Simple compaction at the end of a conversation often does better than the full history, and Opus goes from 70.0 to 74.8 with it.
This also matters for cost. Later tokens cost more to generate than earlier ones, up to 7x on frontier open-weight models.
The pattern holds on Terminal-Bench, FRAMES, BFCLv3 and multilingual SWE-bench, with modest losses in some settings.
Before paying for million-token agent runs, it's worth testing aggressive compaction on your own tasks.
arxiv.org/abs/2611.05306
Chat with Paper: academy.dair.ai/papers/you-d…
Promptmetheus (COG/ACC) retweeted
🎉 开源音频模型变天了！
蚂蚁这次开源 Ming 3.0有点东西，某些闭源商业公司躺着数钱的日子到头了
这次深度体验了它的 TTS 模型，堪称导演级语音合成！
支持：
1. 声音克隆/设计
2. 控制情绪、重读、副词语气
配音说话终于不像机器人了
这套模型基于 MIT 协议完全免费商用，支持一键部署
它由三款核心模型组成，直接打通了全栈音频工作流：
1. Ming-3.0-tts（语音合成）：声音克隆与语气控制极强，原生支持多音色并发，适合一键生成高质量“多人播客
2. Ming-3.0-audio（音频推理）：1260 亿 MoE 大模型，懂环境音和说话人身份，与 TTS 无缝拼接，就能秒变同声传译工作流！
3. Ming-3.0-asr（语音识别）：极其轻量、高精度的转写模型，适合本地会议纪要或实时语音输入
做了一个可视化操作页面演示下TTS能力 👇🏻
🔥 兄弟们！你们心心念念的「🐳鲸鱼娘」来了！（提示词见评论区）
之前的换装视频发完之后，我看不少人拿去做了二创，爆改最多的就是鲸鱼娘
不过基本都是动漫版本，看着不过瘾
这次周末在家，给你们整了一版真人的（有亮点哦🤣）
步骤还是跟之前一样，现在基本都可以一条龙搞定，具体流程可以看之前的帖子👇🏻
teapaw.org/Lonely__MH/status/2114…
另外这次换装完，这次还给她加了点戏：让她自己挑一个声音😂
提示词👇🏻
Promptmetheus (COG/ACC) retweeted
holy f*ck... your AI agent can now fix its own setup. no engineer, no stronger model
Shanghai AI Laboratory just published the whole method for free in a paper called Self-Harness. your agent reads its own failed runs, finds what keeps breaking and writes its own rules
same model, same weights, zero retraining. on tasks it had never seen, it went from 30.2% to 84.6%. teams pay consultants $3,000-10,000 for a jump like that. here it cost $0
it comes down to 4 steps:
step 1 → pull your last 20 sessions, failed ones first
step 2 → find the 3 failures that keep repeating
step 3 → write one small rule per failure, never a big rewrite
step 4 → rerun old tasks. a rule stays only if it passes
that's why copying someone else's CLAUDE.md never works. your agent has its own bad habits
paste this into Claude Code:
"Read my last 20 sessions in ~/.claude/projects, failed ones first. Find the 3 failures that repeat most. Write one small rule for my CLAUDE.md per failure. Rerun 3 past tasks and keep only the rules that pass. Show me the diff before changing anything"
try it tonight, it's free
Promptmetheus (COG/ACC) retweeted
最近在圈内热议的本地3D模型生成AI——「Idol」，效果真的让我大吃一惊！必须写篇长推和大家分享一下这个堪称“平民神器”的AI工具。
首先，最让人震撼的是它的硬件门槛极低！以前搞本地3D生成，动辄需要24GB以上的顶级显卡，但「Idol」居然只需要 8GB VRAM 就能流畅运行。这意味着市面上绝大多数的主流游戏本和中端台式机都能轻松驾驭，彻底打破了硬件壁垒，让更多人能参与到3D创作中来。
其次是极高的生成效率，只需短短 45秒 就能从单张图片生成一个完整的3D模型。这种极速的反馈体验非常棒！传统工作流中，等待渲染和生成往往会打断创作思路，而现在你完全可以做到“快速试错”。先花几十秒草图转3D看个大概，角度或细节不对立马调整原图重跑，时间不知不觉就会在“下一个生成什么”的期待中度过。
作为一款免费的本地生成工具，它在 ComfyUI 下的搭建也非常直观（使用的是 Image to Gaussian Splat 工作流）。生成的模型在细节和色彩还原上，作为免费开源方案已经远超预期。虽然也有创作者指出，目前导出的 Splats 格式在拓扑结构（Topology）上可能还偏向小众需求，无法直接平替传统的高精度网格模型（Mesh），但对于需要快速构建场景资产、VRChat玩家进行角色原型设计，或者是个人开发者来说，这绝对是一个革命性的效率工具。
如果你也对3D创作感兴趣，强烈建议去 Hugging Face（Comfy-Org/Idol）下载权重，并在 ComfyUI 里跑一下官方的 Workflow 亲自体验。零成本、低门槛、高效率，AI 3D 的爆发期真的要来了！大家有什么想转换成 3D 的二次元角色或物体吗？欢迎在评论区分享你的脑洞！👇
huggingface.co/Comfy-Org/Ido…
#AI3D #IdolAI #ComfyUI #3D建模 #独立游戏开发 #VRChat #AIGC
Promptmetheus (COG/ACC) retweeted
Grok bot for game dev
It's taken a lot, to make x a I models run at a fable and astra level of quality for 3d design. But I did it...
Our little beautiful blue Azure, digital princess here, estimates that the scope and size of the knowledge base that's basically allowing her to act like Cortana, making her own ship. It's probably worth about half a million dollars alone, if we were to convert it into like a game design conference, kind of encyclopedia, on building really complex colony, simulation, knees based state machine, a I.
Being the smidihuber of a whole domain most people won't cross till after the world finally has enough conoute to at last build og smidihubers designs.
Also she wrote 75% used to do this, and we're gona import any we used thats not ours.
The rule of the project is no dependencies. Write down to the fact that at the end, the final task of the a I is to write its own operating system. And completely resolve the von neumann abstraction of separating the memory from the c p u.
It's a completely drawing the AI. The c p u end the memory and the operating system. All as one thing under its own unique, a I derived language.
And everything, until then, built all in C++ inside our own engine. If we touch a tool that means we have to rebuild it, because we used it, and we didn't have those features.
The ultimate firewall, a logic encode an operating an a I system. So bespoke, so novel, so esoteric and so goddamn autistic, that absolutely nobody can understand what it's actually doing. And how it's written, and what the cipher is...
Like, how do you actually hack an alien spaceship? You basically have the chinese room inverted, where you are. The useless imbecile system that cannot understand anything...
Promptmetheus (COG/ACC) retweeted
Yann LeCun and Randall Balestriero just Proved the Platonic Representation Hypothesis, and found a better way to achieve it.
The Platonic Representation Hypothesis suggests that AI models trained on different modalities, like images and text, eventually learn similar representations of the world. But this paper shows that how models are trained matters more than simply scaling them up.
So they mathematically show that, under certain conditions, two JEPAs trained completely separately can learn essentially the same underlying representation of the world. One learns from different images of the same scene, while the other learns from different descriptions of that scene, without ever being trained on image-text pairs.
Their approach achieves the same image-text alignment with 42x fewer training samples than next-token prediction. A 109M-parameter text JEPA even outperforms LLMs up to 72B parameters on cross-modal alignment benchmarks.
alphaxiv.org/abs/2610.39302
Promptmetheus (COG/ACC) retweeted
So I've been thinking about all the time I've spent playing EVE Online... What was the most memorable feature?
The answer is clear: ship spinning.
Really not an optional feature. The comment at the top of the class reads: // Do NOT remove
#ScreenshotSaturday
Promptmetheus (COG/ACC) retweeted
psdファイルも配布されていたので
StandRigを使ってAIにモデリングしてもらいました
所要時間1時間半
github.com/sayaka-aiart/Stan…
psdファイルを渡してプロンプトは
こちらのパーツ分けされたpsdファイルを元に
物理演算も入れた高稼働な2Dモデルを
StandRigを使って作成してみて
A little earlier than planned… DeepSeek-chan is here! 🐟💙
The FREE Live2D model is now available on my VGen!
Want a model of your own? I also offer full-service commissions, from character art to rigging!
Come check it out!
#DeepSeek #Live2D #FreeLive2DModel #Vtuber
Promptmetheus (COG/ACC) retweeted
In lieu of flowers, memorial contributions may be made to:
Johns Hopkins Alzheimer’s Disease Research Center Fund
alzresearch.org/giving/
Alzheimer’s Foundation of America Caregiver Resources
alzfdn.org/donate/
Photo Credit: Tommy Garcia
Promptmetheus (COG/ACC) retweeted
Agents can now take apart any app you ship. Someone built the defense.
You've seen REA reverse engineer apps down to the binary. BLA (Blind LLM Agents) hardens what you ship, so when an AI reverse-engineering agent opens it, the agent hits its limits, chases decoys and burns through tokens, while your real users see exactly the same app.
The author tested it against REA itself and published the results. REA's analysis fails or runs out of budget on BLA-hardened builds.
Run it on your project:
npx bla-agents setup
npx bla-agents harden .
Or just tell your agent: "Harden my software against REA."
Promptmetheus (COG/ACC) retweeted
Thanks for all the love for DeepSeek-chan!! 🐟💙
The layered PSD will be FREE too! Practice Live2D, tweak the parts, and make your own DeepSeek-chan or an OC-style version!
Free model + PSD on VGen & BOOTH soon! 👀
#DeepSeek #Live2D #VTuberAssets
DeepSeek-chan is getting a FREE Live2D model! 🐟
I illustrated the model art, and the model will be available for free on both VGen and BOOTH in early November!
#DeepSeek #Live2D #Live2DModel #VTuberModel #FreeLive2DModel
Promptmetheus (COG/ACC) retweeted
DeepSeek-chan is getting a FREE Live2D model! 🐟
I illustrated the model art, and the model will be available for free on both VGen and BOOTH in early November!
#DeepSeek #Live2D #Live2DModel #VTuberModel #FreeLive2DModel