AI & automation enthusiast 🤖 Discovering tools, workflows & ideas worth knowing Learning, building & sharing what works ⚡ 🚀[email protected]
New York, USA
Joined October 2025
- Tweets777
- Following5.8K
- Followers11.7K
- Likes640
Nova Khan retweeted
Stop chasing viral trends. 🔥
Create AI videos with consistent characters, creative storytelling, and a style people want to copy.
The goal isn’t to follow the trend — it’s to set the next one. 🚀
Nova Khan retweeted
Your OCR says it’s done—but did it get everything right? 🤔 In high-stakes workflows, even a small error can be costly.
🚀 Meet the MiniCPM-OCR suite: an open-source solution to parse fast and guard the results.
Two new models:
🔹 MiniCPM-OCR 1.0 — SOTA OCR performance and speed at 1B scale. Compact, accurate, and fast.
🔹 MiniCPM-OCR Jev — A 2B decision model that helps assess parsing accuracy, with performance approaching that of Qwen3.5-397B-A17B on JevOCRBench.
⚡ What’s new?
Small model, SOTA results: MiniCPM-OCR 1.0 scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2.
A sharp goalkeeper 🛡️: MiniCPM-OCR Jev is model-agnostic, checking text, formulas, and tables from different OCR parsers. It achieves 93.6 on JevOCRBench, outperforming JEV-35B-A3B (90.1) and approaching Qwen3.5-397B-A17B (93.8).
Up to 128 pages per minute 🤯: With DSpark, MiniCPM-OCR runs at 1.49× the speed of PaddleOCR-VL-1.6 and 3.62× that of MinerU2.5 Pro. The SDK’s concurrency optimizations further boost throughput to as much as 128 pages per minute.
Deploy where you work: Day-0 support for Arm, Houmo AI, Intel, and NVIDIA, plus 10 more chip vendors through FlagOS.
Ready for your stack: vLLM, SGLang, llama.cpp, and Ollama, with quantized models, fine-tuning support, and Agent Skills.
🔗 GitHub: github.com/OpenBMB/MiniCPM-O…
🤗 Hugging Face:
MiniCPM-OCR 1.0：huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev：huggingface.co/openbmb/MiniC…
👉 Try the Demo：platform.modelbest.ai/consol…
Use the API：platform.modelbest.ai/consol…
Nova Khan retweeted
🚨 You might be overpaying for flights.
One traveler paid $879.
ChatGPT helped cut it down to $299.
No points. No memberships. No VPN.
Here are 8 powerful prompts to find cheaper flights and travel smarter. 👇👇
Nova Khan retweeted
I was planning my content from scratch. Every. Single. Week.
Same process every time: research the latest AI news, brainstorm ideas, write hooks, fact-check everything, and decide what to publish. So I turned the whole process into a collaborative workflow using @Teamily_AI.
👥 I brought multiple people and AI agents together in one group chat to work on the project.
🔍 The Trend Researcher finds the 5 biggest AI tools and news stories of the week.
✍️ The Copywriter turns the best 3 into Reel ideas with engaging hooks.
✅ The Fact-Checker verifies each idea against its sources and sends weak ones back for improvement.
🤝 Each agent contributes its expertise, while the team works together to refine the ideas.
🙋 I review the results and make the final call.
Once everything is set up, I simply click the Loop icon each week to rerun the workflow with fresh news. The agents collaborate, share their work, and help move the project forward without starting from scratch.
Instead of managing separate conversations and repeating the same instructions, I now have a reusable workflow where multiple AI agents work together as a team, with me guiding the process.
This week's content plan is in the comments 👇
Built with Teamily AI:teamily.ai
Try Teamily AI →teamily.ai/i/KSLBMX
#TeamilyAI #AIagents #AItools #AInews #ContentCreator #AIworkflow #CreatorTips
Nova Khan retweeted
Already spending on AI API calls? 👀
ZenMux's Halloween event adds a fun extra: eligible $1 cash PAYG usage can unlock a daily model mystery box.
Keep building, keep experimenting, and see what you get! 🎃🔥
Details: zenmux.ai/halloween
#ZenMux
救命啊，谁懂这种快乐：盲盒里直接开出自己天天用的 AI，脑子里第一反应已经是“这下又能多跑几个项目了”😂
ZenMux 今年万圣节，直接把模型奖励装进了盲盒。🎃
Claude、GPT、Gemini，还有 DeepSeek、GLM、Kimi、Kling 等，奖池覆盖 16 款模型。对天天写代码、折腾 Agent、做图做视频的人来说，这种奖品真的很难不动心。
尤其是手里还有一堆想做的东西：小店页面想重做，万圣节海报还没画，产品宣传视频也想再试一版。要是开到适合的模型，下一份任务直接安排上。
参与方式很简单：每个活动日，在 ZenMux 用现金 PAYG 实际消费满 $1，就能解锁一个模型盲盒。
原本就要跑的任务正常跑，达到门槛，再多一个开盒的惊喜。
ZenMux 本身支持一个账户、一套 API 接入多家厂商的文本、图片和视频模型。写代码、做素材，可以在同一个平台里调不同模型，不用为每种任务重新找一套入口。
这次每天有 3,000 份奖励，单份最高估算使用价值 $500。这里的 $500 指模型使用奖励的价值，不可提现，具体能开到什么、价值多少，以活动页为准。
几个容易漏掉的地方记一下：
🎃 算现金 PAYG 消费，单纯充值不算达标，订阅用量、赠送额度、优惠券抵扣也不计入。
🎃 北京时间每晚 22:00 刷新，每个账户每个活动日最多开一次，全程最多五次。
🎃 活动到 10 月 31 日 22:00 结束。
本来就有模型调用需求的，可以顺手看看。至于开出来的模型怎么用，我觉得最有意思的玩法就是：抽到谁，就给谁出一道自己一直没空解决的难题。
#ZenMux #品牌合作
前往 ZenMux，查看并解锁今天的盲盒
zenmux.ai/halloween
Nova Khan retweeted
This oddly satisfying dental deep clean video is something you can't stop watching! 🦷🧼 Make sure to take care of your oral health
Nova Khan retweeted
🧵 Most people are using AI WRONG. 👀
They're trying random tools, copying prompts, and wondering why their results look average.
Here's a simple AI workflow that can turn one idea into amazing content. 👇
Nova Khan retweeted
🚨Google DeepMind’s Genie 4 could change gaming
Imagine creating an entire interactive world with just a text prompt
*AI-generated environments
*Real-time interaction
*New possibilities for robotics
The future of AI is getting WILD
#GoogleDeepMind #Genie4 #AI #FutureOfTech
Nova Khan retweeted
You can smell a man who gets left on read from a mile away.
They all do the same things:
Nova Khan retweeted
Clients don't buy your tools.
They buy solutions.
Learn the skill.
Understand the problem.
Deliver work worth paying for.
That's how trust grows. 🤝✅✅✅
Nova Khan retweeted
Grabbed drinks with a friend who runs HR compensation for a 6,000 person company.
He said six weeks before you even sit down for your annual review, HR and the VPs have already locked in:
• Your exact raise down to the dollar
• Who gets promoted in Q1
• Who gets cut if budgets tighten
Waiting until your review to ask for more is a trap.