← Back to leaderboard
marryevan999

Marry Evan

Verifiedactive
@marryevan999

Exploring the future of AI and digital transformation. Sharing insights, tools and trends shaping tomorrow. DM for Collaboration 📥 Mjkyumnk@gmail.com

RosterTrading & MarketsCrypto & Web3AI Agents & Productivity Tools
44,615 followers · 46 following · Last polled 3h ago
Perf. Score
58
Rank #159
Value Score
87.9
#30 of 273 priced
Median views
8.2K
p25 1.8K · trailing 7d
Followers
45K
-0.2% 7d
ER (impr.)
1.19%
eng ÷ impressions
ER (foll.)
0.34%
eng ÷ followers
Est. CPM
$2.68
post-rate basis
WoW views
+29.4%
▲ rising

Rates & value · est. from 8.2K median organic views

underpriced
Quote tweet
—
no CPM
Postbasis
$22
$2.68 / 1K views
Thread
—
no CPM
Retweet
—
no CPM
Value Score 87.9 · rank #30 of 273 priced$165.41 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (8.2K), not the mean (13K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (1.8K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
CapCut Black Friday officially launches today! Enjoy up to 60% OFF Pro and 40% OFF Ultra, October 20–30. CapCut Desktop only. AI-generated clips can look amazing individually, but putting them together sometimes leaves a campaign with inconsistent colors and lighting. Intelligent Grading in CapCut PC helps streamline color grading across footage, making it easier to work toward a consistent visual style. I can bring the different product shots closer together in tone without treating every clip as an entirely separate editing task. The goal is to make all those generated scenes feel like they belong in the same campaign. #capcutpc #capcutmarketing #capcutsale #capcutblackfriday1d ago56K10079670920.60%
You bring the business idea. Let your creative tools help with the rest. Dreamina generates the assets; CapCut PC helps you turn them into polished, shareable marketing videos. Take advantage of the $6.29 deal while it lasts. The offer is live now through October 30. #CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale17h ago49K817919
0
13
0.39%
Found something interesting for everyone exploring AI tools 👀 Qoder is offering Qwen3.8-Flash completely free through September 30. No paid subscription is required, and free accounts can use the model without spending Credits. If AI coding interests you, check it out! https://t.co/YUtZ2dtIkE⁠frozen29d ago43K671000.03%
I’ve wanted to build my own AI tool for a while. The problem? I don’t want to spend weeks learning to code just to test an idea. So I tried building one with @EazoAI. One idea. One prompt. Let’s see how far it goes. 🧵 https://t.co/xStNIPGCigfrozen8d ago37K19885200130.86%
Imagine a finished poster. The headline enters in reading order. The subject begins to move. Text attached to that subject stays connected. The background responds without losing the original design. That is the idea behind Dynamic Design, powered by SenseNova 6.8 Flash Preview. It understands what should move, uses HTML/CSS, SVG, transparent assets, or video where appropriate, choreographs the elements into a coherent rhythm, and checks the result for missing text, alignment problems, occlusion, or composition changes before refining the affected areas. The visual shows the original static design becoming a dynamic one. @SenseTime_AIfrozen22d ago22K6410170.13%
For e-commerce brands, the real cost of video advertising isn’t just production—it’s figuring out which creative will perform before committing to a larger budget. Which hook captures attention? Which product angle stands out? Which visual style resonates with your audience? Vidu Q4 Preview makes creative experimentation more accessible: → Starting at $0.014 per second—with per-second pricing as low as half that of comparable models. → Up to 15 reference images to help maintain product consistency across multiple variations. → 2K and 4K output for high-quality final productions. A smarter workflow: 1. Create multiple opening hooks for the same product. 2. Generate quick video variations to test different concepts. 3. Compare the results and identify the strongest creative direction. 4. Invest your production budget in the version with the most potential. Test more ideas before committing to the final cut. Better creative decisions start with affordable experimentation. In this situation, would you choose SD or Vidu? Share your choice in the comments. #ad #ViduQ4Preview #ViduAI #Ecommercefrozen13d ago20K225289121.34%
What if your next AI model could design your app, build your website, and create interactive 3D prototypes? Ling-3.1-flash is making that workflow more accessible. 560B total parameters, 1M context, and strong design benchmarks, all with open weights. We're getting closer to AI that builds, not just chats.frozen7d ago19K12125010.16%
One thing that makes AI tutoring feel unnatural is how quickly it dumps everything on you. Humans don't teach like that. They explain a little, check if you're following, and adjust from there. Getting that rhythm right could make AI learning feel a lot less like reading a textbook.frozen6d ago16K9471760721.93%
A lot of small businesses aren't bad at marketing. They just ran out of hours.frozen8d ago15K34114030.36%
We’re about to see this being used for every company mascot out therefrozen6d ago15K10280830762.35%
The spreadsheet running a creator business usually contains more product knowledge than the creator could put in a spec. That weird extra column exists because a brand changed the deal halfway through. “Approved-ish” is a status because agency approval and client approval turned out to be different things. This is the opportunity I see for Eazo. People already know a lot about the software they need. They’ve been expressing it through columns, colors, and increasingly unreasonable notes in cells. I’d start with that spreadsheet and ask Eazo to build around it. Then I’d make a deliberately annoying change to the workflow. Because the reason the spreadsheet survived is that I could always change it myself. If a personal app preserves that freedom while making the daily work easier, I can see it sticking. @dominicy0frozen14d ago13K16149290.37%
Voice input has mostly been judged by how accurately it writes things down. Voiskey is now aiming at a different question: what should happen after you speak? With TypeSafe AI's Jev now available on Voiskey, a spoken request can move directly toward an action instead of stopping at text. https://t.co/efQpOe1QxLfrozen28d ago13K423000.07%
there's a lot happening in open-source AI right now, and new releases give developers more choices. curious to see where Ornith-1.5-122B fits in.2d ago12K9578720962.83%
You can give an agent every document a product team has written and still fail to tell it what the team actually believed. The launch announcement is polished. The retrospective has explanations. The original prediction might be a sentence in a planning meeting that nobody wrote down. An agent can’t help us learn from a bet we never recorded. That’s an important distinction when thinking about Qoni. Hosted identity, action, and memory give developers a foundation for an ongoing product. The developer still has to decide what that product asks people to record, and when. For an experiment assistant, I’d start before launch: what do we expect, and what would change our minds? Those are product decisions with more consequence than another summary button. Eazo shows one way to build on Qoni. What interests me about opening that foundation to other builders is how much room there is for people who understand a particular job to make better decisions about the agent around it. @dominicy0 #qoni #eazo #personalagentfrozen10d ago12K211811080.48%
CHIIKAWA Hangzhou City Limited Encountering Food Series🍜 Collect these ultra-cute plush charms inspired by iconic local treats! Available now for a limited period. 🔗 https://t.co/Xe7bFWS6XS MF278 #Usagi #LopEarUsagi #chiikawa #CHiiKAWA #ChiikawaFood https://t.co/U904BAwRHdfrozen3d ago12K9268240712.15%
Synthesia built an interview that gets harder the better you answer, which is the honest version. real jobs do the same thing, they just hide it behind free coffee.frozen6d ago9.4K13220170111.92%
Most OCR launches make you pick between fast and accurate. OXR didn't get that memo. The 1.3B open-source parser tops olmOCR-Bench (87.84), then repeats it on OmniDocBench and DocParsingBench, pulling furthest ahead on tables. Chandra-OCR-2 burns 1,557.8 s on 100 pages with one RTX PRO 6000. OXR: 101.0 s. Layout is a swappable module, ordering is decided while text is read, and RSI shaped its post-training. https://t.co/lULzRIC8AE2d ago9.4K320000.05%
Open-source speech releases are interesting for accessibility as well as voice assistants. Ming-3.0 covers generation, understanding and recognition; I'd like to see what people adapt for reading and hands-free interaction.2d ago8.2K1174000.27%
Flagship power without the flagship bill. Hy4 preview combines top tier open source performance with pricing of $0.834 for input, $2.501 for output, and $0.042 for cache hits per million tokens. Free access is available on WorkBuddy for two weeks.8h ago8.0K000000.00%
What if your next big idea could become your next action—with just your voice? 🎙️ That’s where SpeakON gets interesting. By connecting voice input with AI-powered tasks, reminders and scheduling, it helps turn everyday thoughts into real action. Less typing. Less friction. More getting things done.2d ago7.8K1485590112.85%
@chromaawards is open, so our meetup is screening some Season 1 favourites next week to get people going3d ago7.6K109107071.75%
Natural language + automated device testing is a combination I’m excited to see more of. @autosanaai #iPhoneDuo #iphoneduofrozen16d ago7.1K5137270121.79%
Finding the right video still leaves you scrubbing through it for the part you remember. Ancher’s promise of bringing back the exact moment from a saved video is the detail that stands out here. Especially when you only remember what was said. Explore Ancher: https://t.co/LwhhwGZmx0frozen13d ago6.3K321200.13%
StepAudio 3 Music writes an arrangement plan before generating audio. A great opening means little if the song has nowhere to go. Listen for the progression: https://t.co/3VMtGnRKSOfrozen24d ago5.9K320000.09%
The jump from “AI that writes code” to “AI that engineers” is getting harder to ignore. Ling-3.1-flash spent hours solving real software tasks, handled regressions, passed correctness checks, and delivered measurable speedups. Now imagine giving that capability to an open-source developer ecosystem.frozen22d ago5.0K773010.36%
Tavus gave me early access to Griffin. 45% of 120 people in their Turing test thought Griffin was a real person. After testing it myself, I can understand why that result is interesting. Watch my session 👀 @tavus @hassaanrza https://t.co/Y8VQgBbM3efrozen21d ago4.6K2315161121.45%
Every new turn reprocesses the conversation so far. With caching on Inworld Inference, the part that repeats is billed at $0.035 vs $0.07 per million on Gemma 4 26B.frozen8d ago4.4K14115140.80%
Watching Uncle Roger get roasted by a clone of his own voice is so funny. @FishAudio didn't just build a voice tool. They built a digital bully.frozen9d ago4.2K23116111.00%
Gemini 3.8 TTS is #1 in all seven languages on @voicearena_ai. Not one, all seven. Huge congrats @GoogleDeepMind @sundarpichaifrozen29d ago3.5K321000.17%
swapped my coworker into a boardroom scene from a prestige drama and sent it in our team slack, mistake, 10/10frozen7d ago3.2K973010.63%
Interesting move from GMI Cloud. $668M in new financing while AI demand keeps pushing the need for more reliable compute. Curious to see what they build next. @gmi_cloud @alex_yehya22d ago3.1K2352000.95%
This is what practical AI automation looks like. Turning scripting, music, shot planning, video generation, and editing into one repeatable workflow is a serious leap in creative production. Open-sourcing the whole system makes it even more impressive. https://t.co/cmgKmMjBgIfrozen22d ago3.1K541010.35%
The avatar space is getting crowded, but that competition should push the technology forward.frozen28d ago3.0K1735020.90%
Huge congratulations to @nvidia 🎉 Nemotron 3 Diarization just took #1 on @voicearena_ai's new Diarization Bench at 14.72% DER. And it's open weights. That's a big moment for anyone building with speech.frozen29d ago2.8K1420000.57%
If you aren’t running your GTM directly through Claude Code yet, you’re basically NGMI.frozen14d ago2.6K1495121.18%
What happens when you give an already strong model several thousand hours of carefully curated speech in a language? Monsoon’s results offer a striking answer.frozen24d ago1.9K421000.38%
Ling-3.1-flash just got a serious upgrade. The new 1019 version pushes its Intelligence Index from 41 to 44, while agentic benchmarks show meaningful gains too. AA-Briefcase jumped to 1524 Elo, and AutomationBench reached 68%. Better reasoning. Stronger agents. Open weights coming soon.9h ago1.8K57132094.46%
How the 𝕏 algorithm works: Shitposting: 500,000 views Sharing 2 years of hard work on your product: 1.3K viewsfrozen4d ago1.5K510000.40%
September was a strong month for @tuttinghq. 🌍 14,269 creators registered, 1,431 actively posted, 8.6M effective impressions, and 7 X trending appearances. The creator economy is scaling fast. Let’s see what October brings. 🚀frozen22d ago1.5K430000.48%
The interesting part of Raven isn’t just orchestrating multiple agents. It’s the idea of an AI harness that can improve and rewrite its own prompts, policies, strategies, and playbooks. That could make agent systems far more adaptable to complex, long-running workflows. 🐦‍⬛frozen24d ago1.4K430000.51%
September was a massive month for TuttinQ. 📈 14K+ creators, 8.6M effective impressions, 46 events, and 7 trips to the X trending list. The creator ecosystem is clearly scaling fast. October looks even more interesting. 🚀frozen22d ago1.4K431000.59%
The espresso-inspired direction here is such a simple idea, but Pexo turned it into a complete visual story with the headlines, motion, and pacing all working together. This is the kind of AI-generated creative that feels intentional rather than generic.frozen29d ago1.3K12105022.16%
Hedra MCP + CLI makes connecting powerful AI models to your workflow feel effortless. The ability to bring image, video, and world models directly into your agents is a big step forward.frozen22d ago1.2K15149033.38%
The shift from “editing video” to simply telling AI what to fix is huge. Pexo’s mark-to-fix feature makes video creation feel more like collaborating with a creative partner than using traditional editing software.frozen28d ago1.2K13114022.51%
#1 performance with a lower cost per task. This is exactly the kind of progress AI agents need. 🚀frozen17d ago1.1K410000.45%
AI voices keep getting better. HeyGen Voice is now #1 ahead of major competitors!frozen13d ago959400120.73%
Does picking one edit style lock you into it going forward? https://t.co/xBLj61zvAw2d ago893410000.56%
Getting the voice to fit the environment is what makes avatar video actually usable. https://t.co/fMphIDSGgAfrozen8d ago860501100.81%
Does it actually sound like him talking, or just close? 🤔 https://t.co/EbqoPnnzgEfrozen14d ago809411100.87%
Would a learner who keeps getting it right still see an engaged response? https://t.co/FGQk8ZLchvfrozen15d ago801311000.62%
This suits educators and trainers more than almost any other use case. https://t.co/MmbYmboyYhfrozen16d ago795110000.25%
bigger despite everything https://t.co/Kqf7qbckdtfrozen10d ago762110100.39%
Curious whether it reads smaller brand sites as well as larger ones. https://t.co/XGgYdeRVr0frozen7d ago741210000.40%
Ling-3.0-flash-Sante is officially open source. 124B total parameters. Just 5.1B active per token. Built for professional healthcare tasks, complex medical reasoning, deep research, and multi-agent collaboration. With its model weights now available, researchers and developers can test its capabilities, explore new use cases, and push healthcare AI further.4h ago4984949150323.29%
curious how the api handles prompts at scale https://t.co/z3TBJYY8ej3h ago229210001.31%