Top AI Voice | Sharing insights on AI, No-Code, Tech Tools & prompts | AI Enthusiast | Top 100 Unstoppable Education & Career Influencer | DM for collaboration✉
Los Angeles, CA
- Tweets 386,653
- Following 4,804
- Followers 406,567
- Likes 254,696
Pinned Tweet
Ready to turn a creative spark into a million-dollar brand look? ✨
Discover Vibe Marketing: the game-changer where you set the intent and Fotor scales the vibe.
This is the end of visual inequality, allowing you to transform raw concepts into polished, studio-quality assets
26
84
105
115,408
My product is supposed to be floating, not looking like it’s been pasted onto the screen.
If you’ve been wanting to level up your edits, don’t sleep on CapCut PC’s Black Friday deals:
- Get 60% off Pro
- or 40% off Ultra
(before Oct 30)
The floating effect is already doing its thing, but the video still needs that extra visual personality. That’s where Black Friday Stickers, Effects & Text Templates in CapCut PC come in.
Add some playful effects and floating text to complement the product’s movement and give the whole scene a more creative finish. It’s the little details that make people stop scrolling!
CapCut Sale: tinyurl.com/2v23rn22
#capcutpc #capcutmarketing #capcutblackfriday #capcutsale
29
134
218
151,762
AI gets interesting when it helps groups get things done.
Instinct joins iMessage group chats, even with friends who don’t use it, to choose a book, find a date, and coordinate copies.
The group assistant sticks to chat context and asks before sharing personal data.
That’s a promising direction for AI.
Instinct in group chats is now available to everyone on iMessage.
Add Instinct to a group chat, and it works for the whole group, including for friends who don’t use Instinct yet. The group gets its own Instinct that only knows what’s in the chat, not anyone’s accounts or data. When it needs something from you, it goes through your personal Instinct, which confirms with you first.
A few things early access groups have used it for:
- Planning a dinner party with friends
- Organizing a road trip
- Decorating an apartment
Add Instinct to any group chat to get started.
@GajanNagaraj @anuda_w @saagarwashere
26
187
32,869
give Teamily's shared context a month and it'll know my team's inside jokes better than the new hire
The shared context for your personal and team agents to get work done, smarter and faster.
• Run on any device
• Create any type of artifact
• Chat anywhere
• Work anytime
#SharedContext #TeamAgents #TeamilyAI #iPhoneDuo
Paid partnership (ad)
33
33
32,510
Rounding up a couple people from work for The Search.
Fake brand, jingle, short ad, 5 days left, $100K prize pool. Our lunch breaks are spoken for
The brief: build a catchy ad in ElevenCreative.
The result: a K-pop-inspired jingle with a panda drinking bubble tea.
Now it’s your turn. Make your own ad for a chance to win a share of $100K.
Only 5 days left to submit to The Search.
25
24
35,367
A video doesn't have to be a fixed sequence of frames. Imagine changing a character or an object while the scene is still unfolding, and seeing the result without stopping the action.
That's a different way to think about video interaction. Xmax X2.1 combines real-time video generation and editing, with 1080p output and around 200ms latency.
Introducing Xmax X2.1.
More versatile 丨Better Quality（1080P）丨Same price （$0.0016/sec）
Reshape video consumption through real-time interaction, with more possibilities for live streaming, shopping, gaming, and social experiences.
8
30
56
84,480
Start your Halloween campaign with up to 60% off CapCut Pro, 40% off Ultra, and a $0.29 introduction to Dreamina. Oct 20–30 (Desktop only)!
Your pumpkin reveal looks great, but the product appears too late to keep viewers interested. Create the scene in Dreamina + CapCut PC.
Generate the visuals in Dreamina, then import them into CapCut PC’s multi-track timeline. Use AI Extend backward to add a lead-in before the carving starts, giving viewers time to get curious before the product appears. Then trim the sequence to keep the reveal moving.
Turn a slow start into a more engaging Halloween launch ad with CapCut PC.
Dreamina: tinyurl.com/5xbtv4x6
CapCut PC: tinyurl.com/2v23rn22
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
29
110
166
101,034
I was building a presentation and couldn't decide which visual direction worked best.
So I used Gamma 5 to explore three different designs for the same project.
The results looked like they came from three different designers.
Here's how I chose my favourite. 🧵
28
105
184
108,908
I also explored how to carry a visual identity into another presentation.
By importing an existing deck as a style reference, I could reuse its typography, colour palette, spacing, and torn paper aesthetic.
The result was a different presentation with a consistent design language.
The video shows the process from import to finished slides.
1
5
5,075
The biggest takeaway? The right design isn't just about looking good. It's about finding the visual direction that fits your message.
Gamma 5 gives you more ways to explore, compare, and refine your presentation.
Which would you choose: A, B, or C? @GammaApp
Try it: gamma.app?utm_medium=creator…
#ad
3,444
the doc-agent bug that scares me isn't a crash. it's this:
SKU-80O ← what the page says
SKU-800 ← what OCR returned
valid string. valid JSON. schema check passes. the agent looks up the wrong record and keeps going. nothing throws.
that's exactly where i'd put a judge like MiniCPM-OCR Jev: keep the image next to the extracted field for one more check, while the error can still be traced back to the source.
a quality score won't catch everything. but it gives the agent a reason to pause on one region instead of trusting every field equally.
and when it pauses, show the human the crop. not a 40-step trace of tool calls. just the code, in the document, next to what was read.
to be clear, this is a workflow i can see around MiniCPM-OCR 1.0 + MiniCPM-OCR Jev, not a validated end-to-end agent.
the more steps agents take on their own, the more these quiet errors cost. by the time one surfaces, the original image is five tool calls behind.
Hugging Face:
MiniCPM-OCR 1.0: huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev: huggingface.co/openbmb/MiniC…
GitHub: github.com/OpenBMB/MiniCPM-O…
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…
2
40
38
112,512
Try the demo: platform.modelbest.ai/consol…
Use the API:
95
I like the direction here: a stronger checkpoint, plus a base model others can keep training.
The reported gains in agent workflows make Ling-3.1-flash-1019 useful for people building beyond demos.
Now open source: Ling-3.1-flash-1019.
44 on the Artificial Analysis Intelligence Index v4.3.2, up from 41 for Ling-3.1-flash.
Also releasing Ling-3.1-flash-base, an open foundation for further training.
29
29
34,825
The useful promise here is not that an agent writes code instantly. It is that Qoder can read context, plan, use the tools, execute, and verify around one stated outcome.
I would put DeepSeek-Flash on a practical maintenance task while it is free through Oct 31.
qoder.com/qoder
DeepSeek-V4.1-Flash-1017 is FREE in Qoder through Oct 31.
Free and paid individual accounts included. No claim needed.
Daily free uses cost 0 Credits, even with a zero balance.
How to start is in the reply 👇
32
30
20,831
Imagine cruising through a GTA 6-style city in a Tesla Roadster instead of the usual in-game cars. Not gonna lie, I’d play that mission. 😂
I wanted to see that idea as a video, so I used ChatGPT + CapCut Web.
- I described the scene and asked ChatGPT to use CapCut to create a clip of the Roadster moving through the game-inspired world.
- It’s a pretty fun way to experiment with a trend without starting from a blank timeline myself.
The CapCut plugin is coming soon to ChatGPT. Early access is available through ChatGPT desktop on PC only.
#CapCutDidThat #CapCutWeb #ChatGPT #CapCutWorld
31
105
125
119,573
Check it out here: tinyurl.com/2xeju4vv
150