AI Video Creator | AI Tools | Future Tech Exploring the latest AI innovations workflows and digital creativity
Joined August 2026
- Tweets1.1K
- Following11
- Followers2.4K
- Likes1K
Pinned Tweet
Tesla In INDIA 🤯
This AI-generated video was created entirely using AI tools and my own creative pipeline. From concept to execution, everything was handled by me.
Comment “Tesla” For the Prompt 🙌🏻
Rachel Woods retweeted
MiniCPM-OCR 1.0 (1.3B) from @OpenBMB is the first OCR-dedicated model in the MiniCPM family: SOTA 93.68 on Real5-OmniDocBench, solid on scanned/photographed/curved/tilted/recaptured docs, bilingual CN+EN, plus tables, formulas, charts, seals & KIE.
DSpark pushes end-to-end throughput to 1.6× (~55% more efficient): ~1.49× PaddleOCR-VL-1.6 and 3.62× MinerU2.5 Pro. Also SOTA 54.7 on OCRBench v2.
Paired with MiniCPM-OCR Jev (the reliability judge), you can flag only the results that need human review instead of checking every page.
That’s the practical unlock for high-volume document pipelines.
Day-one vLLM/SGLang/llama.cpp/Ollama + fine-tuning + SDK support makes it easy to deploy. The efficiency chart shows exactly where it sits on the quality-vs-speed frontier.
Hugging Face:
MiniCPM-OCR 1.0: huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev: huggingface.co/openbmb/MiniC…
GitHub: github.com/OpenBMB/MiniCPM-O…
@OpenBMB
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…
Rachel Woods retweeted
The offer should be easy to read, so here it is: Dreamina’s first month is $0.29. CapCut’s Black Friday sale is live Oct 20–30 on desktop only, with 60% off Pro and 40% off Ultra.
That same clarity is what I’d aim for in the ad. Create a clean background in Dreamina, bring it into CapCut PC, and leave room for the main message. Time the text to the narration, keep the product visible, and hold the final card long enough to read.
Dreamina and CapCut Pro normally cost $15/month each, with separate subscriptions and individual offer links. Both deals end October 30.
Give your next promotion a message viewers can follow.
CapCut: tinyurl.com/2v23rn22
Dreamina: tinyurl.com/5xbtv4x6
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
Rachel Woods retweeted
The cinematic idea in my notes might finally get its turn: Dreamina’s first month costs $0.29. CapCut’s Black Friday discounts are live Oct 20–30, desktop only, at 60% off Pro and 40% off Ultra.
I’d begin with one dramatic visual direction in Dreamina. Once the assets are in CapCut PC, the edit becomes about anticipation: hold the opening, bring in the sound, then cut to the reveal. A brief closing line can give the scene a purpose beyond looking good.
Normal pricing is $15/month for Dreamina and $15/month for CapCut Pro. They’re separate subscriptions, each accessed through its own offer link; both deals end October 30.
Take one ambitious shot idea through to a finished promo.
CapCut: tinyurl.com/2v23rn22
Dreamina: tinyurl.com/5xbtv4x6
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
🤝 Paid partnership
Rachel Woods retweeted
A little more creativity can make your next brand post feel completely different. ✨
Dreamina’s first month is only $0.29, and CapCut Desktop has a separate Black Friday deal: 60% OFF Pro or 40% OFF Ultra through Oct 30.
Use Dreamina to create AI visuals for your campaign, then continue refining the video with AI Edit in CapCut PC. They’re separate subscriptions and each offer needs to be purchased separately.
Dreamina offer: tinyurl.com/5xbtv4x6
CapCut Desktop sale: tinyurl.com/2v23rn22
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
🤝 Paid partnership
Rachel Woods retweeted
The creator of Claude Code at Anthropic uses 5 agents in his terminal and 5–10 more in his browser.
I took that idea and built a market research system with 12 agents, 1,284 news sources, and 3 trading signals.
Pick a stock, change the news, and watch how the signal changes.
You can also change the testing method and see the results and code next to the chart.
Each agent has a specific job:
→ Scouts collect news and stock prices
→ Parsers clean headlines and price data
→ Linkers find connections between stocks
→ Reviewers test every signal before showing it
A signal may look good but still be wrong.
Boris Cherny follows one rule: Claude tests every change before it goes live.
So the reviewers check three things:
→ Does it work with 3 years of data?
→ Does it stay stable when the news changes?
→ Does it perform better than doing nothing?
Every signal must show:
What is being tested
Which data is being used
How the results differ from the backtest
What still needs to be verified
In my demo, 28 stocks and 11 patterns were tested. 8 matched, while 3 were marked for review.
These are demo results, not financial advice. The testing method is what matters.
You can try the same idea with Claude Code using this prompt:
“Turn my watchlist into a market research lab.
Collect news and prices for each stock. Save the source for every claim.
Give each task to a separate agent and make them use the same data.
Test every pattern, record failures, and clearly mark unverified results.
Show the signal, test results, and errors side by side.”
Start with one stock and one news source before adding more agents.
Save the data and test results so you can run everything again tomorrow.
Watch the spider work below.
Giving This Free for 24 hours. To get it:
1. Comment the word ' DM '
2. Like and Retweet this post
3. Follow me
@techwith_fahim
Rachel Woods retweeted
My brother borrowed €300 from me.
He refused to explain why.
I trusted him and gave him the money.
Two weeks later, he came back asking for another €200.
“I promise I’ll pay you back.”
He never did.
So I made a decision: no more loans.
Yesterday, he showed up at my house without warning.
He walked into my room, left a bag on the bed, and walked out without saying a word.
Inside the bag were €500.
Every cent he owed me.
But there was something else.
A small key.
And a handwritten note.
Rachel Woods retweeted
🚨 You're using ChatGPT WRONG. And you're probably wasting hours because of it.
Sam Altman, CEO of OpenAI, says:
"You no longer need to write prompts."
In just 38 minutes, he explains how to use ChatGPT at a level most people haven't even discovered.
This isn't another random AI tutorial.
It's a talk he gave to Stanford students. A friend sent me the recording last night.
After watching it, I realized something uncomfortable:
I've been using only a fraction of what ChatGPT can actually do.
If you're still treating ChatGPT like a basic chatbot, you're missing out on its real potential.
Watch the full 38-minute talk.
Then follow me for more AI insights that can change the way you work.
Rachel Woods retweeted
A 4B model that achieves 82.9% mean pass@1 on AIME 2024–2026 while keeping memory usage constant, versus at most 18.8% for the evaluated bounded-state models, including 14B-class models.
72 MiB of attention KV per sequence. Same at 32K context, same at 256K.
Qwen3.5-4B uses 8 GiB of KV Cache at 256K. This uses just 72 MiB. That's 99.1% smaller.
Meet EdgeRock-4B 🧵
🤝 Paid partnership
Rachel Woods retweeted
Everyone’s debating which AI model is best.
Meanwhile, that PowerPoint deck still takes two hours.
Templafy Document Agents are tackling that problem right inside PowerPoint.
My take: AI should earn its place by saving us time on everyday work.
🤝 Paid partnership
AI that knows your company’s content and branding? 👀
With Templafy Team, build and share AI Agents that create editable PowerPoints using your company's knowledge and templates.
Get started 👉 templafy.com/templafy-teams-…
Rachel Woods retweeted
It’s easy to overlook your own deck after spending hours on the same slides.
Templafy Document Agents do more than just create decks:
→ they turn your content into a deck
→ they tweak it without messing up the layout
→ they give it a final look before sending it to the client
But, oh no, the review step is often skipped, and then everyone feels bad when it’s already out there.
Having an agent double-check the deck against your brand is like getting a second pair of eyes for a fraction of the cost on a Friday.
🤝 Paid partnership
AI that knows your company’s content and branding? 👀
With Templafy Team, build and share AI Agents that create editable PowerPoints using your company's knowledge and templates.
Get started 👉 templafy.com/templafy-teams-…
Rachel Woods retweeted
the moment a person replies to your agent is where most setups break. OpenMail sends the reply straight to your code over a webhook, in the same thread, so the agent can pick it up.
🤝 Paid partnership
Agents f*ck up.
When your SDR agent does, it shouldn't take the domain your invoices go out from with it.
OpenMail gives every agent its own inbox, its own sending limits and its own reputation.
openmail.sh
Rachel Woods retweeted
Finance teams are about to get their nights back
Hyphenate just raised a $45M Series A to put AI agents inside the office of the CFO
Not a chatbot. Agents that actually run the work:
• Order-to-cash
• Treasury
• GL accounting and close
• Procure-to-pay
• Reporting
Revenue grew 86x in the last 12 months. Customers start with one module and keep adding more
This is what AI agents doing real business work looks like
hyphenate.ai/press/hyphenate…
#hyphenatepartner
🤝 Paid partnership
@HyphenateAI just raised a $45M Series A at a $270M headline valuation.
We're bringing autonomous finance to every enterprise in the world. 🧵
Rachel Woods retweeted
🚨SHOCKING: I researched 30 Amazon products without opening a single product page.
I gave BrowserAct a simple prompt asking for the best gardening tools on Amazon US, and it built a browser automation bot that pulled everything into structured results:
→ product names, descriptions, and features
→ current prices and discounts
→ ratings and reviews
→ specs and availability
→ images, product links, and more
That's research that normally means hours of clicking through listings and copying everything into a spreadsheet.
You can also reuse the same bot, schedule it to run regularly, and have it track price or product changes automatically.
If you sell on Amazon or keep an eye on competitors, this one's worth trying.
browseract.com/?co-from=KOL-…
🤝 Paid partnership
Rachel Woods retweeted
Breaking: Your phone is sending data to Google every 4.5 minutes.
Screen off. Phone untouched.
Trinity College Dublin confirmed it in a peer reviewed study.
Not a theory. Not a conspiracy.
A published study with a 12 setting fix:
Rachel Woods retweeted
My faceless YouTube channels make me $12,997.80 +/month (long form only)
There is 𝗻𝗼 𝗹𝘂𝗰𝗸 involved.
Let me send you a free course on exactly how to do it.
To get: -
1. Follow (So I can DM you )
2. Retweet
3. Reply " Yt " Must follow me to get DM.
Free for 48 hours.
Rachel Woods retweeted
Creating content for a product is one thing.
Doing it consistently while managing everything else is a different story. 😅
I’ve been exploring ways AI can make content creation easier, and I came across Reachira.
What I like is the idea of starting with just your product link and turning it into short-form marketing content.
Here’s what you can do with it:
→ Generate videos for your product
→ Create UGC-style content and app demos
→ Get content ideas inspired by trends
→ Schedule posts across social platforms
→ Track campaign performance
Instead of spending hours figuring out what to create and when to post, you can bring more of your content workflow into one place.
I can see this being useful for SaaS founders, indie hackers, and small businesses that want to market their products without building a huge content team.
The best part? You can try it with 10 free AI credits, without a credit card.
Check it out: reachira.com/
#AI #ContentCreation #Marketing
🤝 Paid partnership
Rachel Woods retweeted
If you're in debt, never start by:
-Paying the minimum
-Taking out another loan
-Using one card to cover another
Instead, start by:
Rachel Woods retweeted
What's your biggest challenge right now?
A) Finding clients B) Growing on X C) Learning AI D) Making your first online income
Reply with one letter. 👇
Rachel Woods retweeted
Andrej Karpathy’s 1-hour Stanford lecture on AI engineering is one of the best explanations I’ve seen of how AI systems actually work.
The progression is simple:
10% → LLM
30% → Prompt
50% → Agent
70% → Loop
100% → Graph
The key takeaway:
AI engineering isn’t just about writing better prompts.
It’s about building systems around models — giving them context, memory, tools, feedback loops, and data flows.
“Delete everything, keep Graph.”
Definitely worth watching if you’re building with AI agents.
Watch → Bookmark it