Sharing insights on AI, No Code, Tech Tools & Prompts | AI & Tech Content Creator | 350K+ LinkedIn & Instagram Community | ✉️ shahmanishkumar1@gmail.com
Joined November 2023
- Tweets56.9K
- Following998
- Followers52.9K
- Likes58.6K
wait... I got a referee-style report on my paper before I submitted it 🧵
I used WisPaper to map the field first, then to check my own draft.
Here’s what happened 👇
🤝 Paid partnership
Manish Kumar Shah retweeted
Sound designers, you need to see this.
Mirelo generates SFX directly from video footage.
One wrong number in a document can lead an AI agent to make the wrong decision.
Drex by @NaceAI helps prevent this by checking extracted data against the original document and fixing errors before they reach your agents.
It ranks #1 on Nace's parsing leaderboard and reports 38% faster corrections on RealDocBench.
It also supports 30 file types, including PDF, Word, Excel, and images, with options to run it in the cloud or on your own infrastructure.
A useful tool for anyone building AI agents that work with documents.
🤝 Paid partnership
Introducing Drex for Document Processing 📑
Decision-powered parsing that catches errors before they reach your agents.
→ #1 on our parsing leaderboard
→ Beats GPT-6 Astra, Claude Opus 4.7 & every parser tested
→ 38% faster corrections on RealDocBench
Try it: console.nace.ai
What if you could handle research, content creation, and app building all in one place?
That's what makes KOOKO by Baidu interesting.
It's an all-in-one AI workspace that brings research, writing, editing, and app development together, and Baidu just unveiled some major upgrades at Baidu World 2026.
One demo shows KOOKO turning a stock market research request into:
- A detailed financial research report.
- An Excel portfolio analysis with charts.
- An interactive investment dashboard.
All within a single workflow.
KOOKO also supports documents, spreadsheets, presentations, websites, code, videos, and marketing materials.
Its latest upgrades make it easier to build and edit websites and lightweight apps just by describing what you need.
It also remembers your preferences and project context, making things easier next time.
KOOKO has already reached over 12 million international users across more than 200 countries and regions in its first year, which just goes to show that people really do love having everything in one AI workspace.
Explore: kooko.ai
AI agents shouldn't just say they completed a task. They should prove it.
That's what makes Plurai interesting.
It verifies what actually happens in an agent's environment. If an agent claims it processed a refund, Plurai checks whether it really happened.
This helps developers catch failures, replay issues, and fix problems before deploying AI agents to production.
Less guessing. More proof. More reliable AI agents.
🤝 Paid partnership
Big day for us at @pluraiAI ! Plurai Simulation is live and free for every AI engineer.
AI agents look great in demos. Until real users break them.
So we built a world where your agent can fail, learn, and get fixed before your users ever see it.
→ Digital twin: Your real agent, its tools, and environment recreated from your code
→ Simulated users & tasks: Realistic users, tasks, and workflows that push your agent into the edge cases that matter
→ Auto evals: Check what your agent actually did, not just what it said
→ Verified fixes: Every failure comes with the run that exposed it and a fix you can verify
→ Replays: Turn any production failure into a simulation and replay it against every new version of your agent
All inside the coding agent you already use: @claudeai code, @OpenAICodexCli , or @GitHubCopilot .
Start simulating for free: simulation.plurai.ai/connect
What happens when you bring the most wanted Black Friday product into a café? Apparently, nobody gets to finish their coffee.
I used CapCut PC to explore this product-grabbing ad with a product image and original video references.
For anyone editing promotional videos with several layers, keeping everything organised can quickly become a headache.
Multi-Track Editing helps separate the main footage, background music, captions and sound effects, so each element can be adjusted independently.
See the editing process: tinyurl.com/2v23rn22
#capcutpc #capcutmarketing
#capcutblackfriday #CapCutDidThat #chatwithcapcutagent
🤝 Paid partnership · 🤖 Made with AI
Baidu just launched ERNIE 5.5 Flash Preview, and its 3D recreation of the historic Hōryū-ji Temple complex shows what's possible with AI-driven modeling.
And it didn't just create the scene. It handled almost the entire process itself.
It researched the temple's architecture, used the Blender Python API to build the structures, found Buddhist statue scans and vegetation assets, and created gardens, a lotus pond, and an arched bridge.
It also added realistic materials, rendered the scene, and used FFmpeg to turn the frames into a finished video.
Beyond 3D modeling, ERNIE 5.5 Flash Preview brings advanced coding, AI agent capabilities, and a 1M-token context window, with 280B total parameters and 14B active per token.
It also delivers generation speeds exceeding 200 tokens per second.
And it's making progress on the coding benchmarks too.
In the latest CodeArena: WebDev rankings, the model scored 1,668 points, ranking 2nd among open models and 4th among Chinese models, behind larger models like Kimi K3.1 Pro and Qwen3.8-Max.
The API is already available on Qianfan, with model weights scheduled to be open-sourced before November 25.
I find this direction really interesting.
The ability to go from architectural references to a detailed, rendered 3D environment could change how complex scenes are built.
If one shared AI model could learn from a whole team’s work, would every company end up training its own?
Genuinely interesting to think about.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
Manish Kumar Shah retweeted
Binocular shots. Secret signals. Product close-ups. Every detail builds the mission.
I use CapCut PC's multi-track timeline to layer the main footage, B-roll, text, and effects separately.
More control over every shot. More room for creative storytelling.
Check it out: tinyurl.com/2v23rn22
#capcutpc #capcutmarketing #capcutblackfriday #CapCutDidThat
🤝 Paid partnership · 🤖 Made with AI
Manish Kumar Shah retweeted
What if you could build a blockchain app just by describing your idea?
No complicated infrastructure setup. No endless API integrations.
Just one prompt.
That's what @Tatum_io is bringing with Tatum AI Builder. 🧵👇
🤝 Paid partnership
Manish Kumar Shah retweeted
I want my ad side hustle to have a production process. Currently, it has tabs.
My affordable pair is Dreamina + CapCut PC. I upload a product photo to Dreamina Marketing Studio, generate images and videos, then import them into CapCut PC’s multitrack timeline.
AI Edit fixes a product detail; AI Extend lengthens the ending.
I box-select the talking clips and tell CapCut Agent EditPilot to tighten long pauses throughout. One plain-English sentence replaces repeated manual cleanup. I can preview and fine-tune the actual edits on the timeline, then layer music, sound effects, and transitions before exporting.
Try this workflow for your next product-ad offer.
Dreamina: tinyurl.com/376c8r3x
CapCut PC: tinyurl.com/2v23rn22
#CapCutPC #CapCutDidThat #Dreamina #Smallbusinessmarketing #chatwithcapcutagent
🤝 Paid partnership · 🤖 Made with AI
Manish Kumar Shah retweeted
Dreamina teamed up with CapCut to give CapCut Pro members more to play with.
Every eligible user gets rewards: up to 10 free Seedance 2.0 generations, bonus credits, and a month of Basic for as little as $0.29 if you're new to Dreamina.
dreamina.capcut.com/ai-tool/…
#DreaminaPartner
🤝 Paid partnership
Most AI tools tell you what to do.
What if one could actually do the work for you?
Meet Nod by @n8n_io.
It has its own computer and browser, so it can handle real tasks across your apps and websites, while checking with you before taking action.
Less time managing tasks. More time getting things done.
Worth exploring 👀
Try it: nod.n8n.io
The real test of an AI agent isn't how well it performs in a demo, but how reliably it handles chaos.
Simulation-driven evaluation could become the new standard for production-ready Agentic AI.
🤝 Paid partnership
Big day for us at @pluraiAI ! Plurai Simulation is live and free for every AI engineer.
AI agents look great in demos. Until real users break them.
So we built a world where your agent can fail, learn, and get fixed before your users ever see it.
→ Digital twin: Your real agent, its tools, and environment recreated from your code
→ Simulated users & tasks: Realistic users, tasks, and workflows that push your agent into the edge cases that matter
→ Auto evals: Check what your agent actually did, not just what it said
→ Verified fixes: Every failure comes with the run that exposed it and a fix you can verify
→ Replays: Turn any production failure into a simulation and replay it against every new version of your agent
All inside the coding agent you already use: @claudeai code, @OpenAICodexCli , or @GitHubCopilot .
Start simulating for free: simulation.plurai.ai/connect
Open-source agents have grown a lot this year, but their memory is still mostly a black box. I'd like more of them to let you open it and check what gets used.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
Product people know a feature only counts if people use it.
AI memory should be judged the same way, by whether the agent actually uses what it saved.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
Finding information is easy.
Connecting the dots across the internet is the hard part.
That's what Ploid is solving.
Deep Research brings a much wider view of someone's digital presence without spending hours jumping between platforms.
Big move from the team.
🤝 Paid partnership
Introducing Ploid Deep Research
You can now see everything
> Ask about anyone and get their full digital footprint in minutes
> Photos, videos, every social account, even the ones you'd never find (like WeChat)
> Works inside Claude, Codex, or any agent
Try it today at ploid.com
Claude's new Whiteboards, Sheets, and Data are getting attention, but there's another option worth looking at.
puppyone has open-sourced its own whiteboard, spreadsheet, and dashboard.
They're free, work with any AI model, and are available now.
go.puppyone.ai/4iBgoWr
🤝 Paid partnership
Great topic for anyone new to AI agents! Whether they improve at a job the more they do it is worth understanding early.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.