← Back to leaderboard
krishnasagrawal

Krishna Agrawal

Verifiedactive
@krishnasagrawal

Sharing cool things in AI, tech & web development AI Agents • LLMs • No-Code • Automation

RosterAI Model News & ReleasesAI Agents & Productivity ToolsTech Career & Coding Education
48,597 followers · 577 following · Last polled 2h ago
Perf. Score
65.5
Rank #100
Value Score
91.4
#17 of 273 priced
Median views
35K
p25 24K · trailing 7d
Followers
49K
-0.1% 7d
ER (impr.)
0.35%
eng ÷ impressions
ER (foll.)
0.37%
eng ÷ followers
Est. CPM
$0.99
QT-rate basis
WoW views
+133.7%
▲ rising

Rates & value · est. from 35K median organic views

underpriced
Quote tweetbasis
$35
$0.99 / 1K views
Post
—
no CPM
Thread
$95
$2.69 / 1K views
Retweet
$20
no CPM
Value Score 91.4 · rank #17 of 273 priced$186.17 / 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 (35K), not the mean (52K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (24K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
The opening day of CapCut Black Friday is today: save 60% on Pro or 40% on Ultra. Get the offer October 20–30, exclusively on CapCut Desktop. The campaign draft is taking shape, and my visual review has narrowed down to one area of one shot. With Local Inpainting in CapCut PC, I can give that selected area another pass and assess the updated version within the edit. Reviewing it in context helps me decide whether the change supports the surrounding footage. Try this approach when your next revision has a clear visual target and the overall creative direction already feels right. #capcutpc #capcutmarketing #capcutsale #capcutblackfriday2d ago112K17834160110.21%
AI can get your visuals started. The edit is where you make them yours. 🎬 Create in Dreamina, shape the final video in CapCut PC, and grab the $6.29 starting deal before Oct 30. #CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale https://t.co/w27sCYOx0f1d ago95K1533121011
0.23%
Jean Phil shows how a character can become the center of a visual trend. Now imagine telling a whole brand story with that same creative spirit. Use Codex to coordinate your concept and AI-generated visuals, then use CapCut Web’s short-film workflow to bring the story together into a marketing video. #capcutweb #capcutchatgptplugin #capcutcodex #capcutmarketingfrozen6d ago80K10340110130.21%
My little side project now has a login, a database and a checkout. Half the small businesses on my feed still take orders through DMs and a notebook. So I asked Eazo @eazoai to build a proper order page for a home bakery. No code. One prompt. Try Here 👇 https://t.co/aWWFAMhEeJfrozen6d ago63K13137100100.30%
AI startups can move incredibly fast, but infrastructure still has to keep up. GMI Cloud just raised $668M to expand GPU capacity across the U.S. and Asia-Pacific for training, inference, and production workloads. The next phase of AI infrastructure should be interesting to watch. @gmi_cloud @alex_yehyafrozen22d ago54K30476160150.76%
Adani group’s play here is pure vertical integration. Gautam Adani already committed $100 billion to build 5 GW of green-energy AI data centres by 2035. Now, they are moving to own the hardware layer. By partnering with Jabil, they combine their energy footprint with precision manufacturing. Generating the green power, building the data centres, and manufacturing the liquid-cooled AI racks that go inside them creates a highly defensible ecosystem. This is a definitive milestone for India's push to become a global AI hardware exporter rather than just a consumer. https://t.co/er7SQGBfrwfrozen9d ago54K2671288080.77%
One thing that slows product teams down: A PM spots a small UI issue. A designer wants to change some copy. QA finds a bug. Marketing needs a small update. But every change ends up waiting in the engineering queue. @base44 takes a different approach with Base Code. You connect the GitHub repository your team already uses, and anyone on the team can work on the existing codebase directly from the browser. Just describe the change in plain language. Base Code builds it on an isolated branch, lets you preview the real product with the change, and then the change comes back to engineering as a standard GitHub pull request. No local setup. No rebuilding the app somewhere else. No separate prototype disconnected from the real product. Engineering still stays in control of what ships. More people can build. Engineering still owns what ships. I also have a coupon code for the first 50 users. DM me if you want the code.frozen24d ago46K21852171150.65%
Creating content usually means planning, scripting, recording, editing, and then publishing. But what if you could go from a simple prompt to a ready-to-post video in minutes? I tested this workflow with VEED. https://t.co/hZ7qJwOYOkfrozen22d ago46K11132160180.39%
Building business software usually takes time, especially when you need something tailored to your workflow. https://t.co/xhh6br9m3T and Proto are exploring a different approach: describe what you need, build the app, and let AI agents take care of tasks like finding overdue invoices and drafting reminders. The interesting part is seeing how it works in practice and where human approval still matters.frozen7d ago42K16839160120.56%
I’ve been testing Dolphin AI recently. They gave me the code XX-WSVKFB to share if you want free credits to experiment with it too. AI video has become remarkably good at generating individual shots. But generating the same world twice is still a much harder problem. Same character. Same face. Same outfit. Same environment. Then carrying all of that into the next scene. Dolphin’s approach to continuity is what caught my attention. Once characters and settings can persist across scenes, AI video starts becoming much more useful for actual storytelling. Especially for short films, episodic content and brand narratives. Try it with the code XX-WSVKFB: https://t.co/9i4fV18FTIfrozen22d ago38K20654340170.83%
AI agents need fresh data, not just existing knowledge. Oxylabs Web API helps your agents search the live web and turn web pages into clean, structured data through a single API. It reliably extracts from any public site at scale, with CAPTCHA solving and JS rendering handled for you, so your agents get results even on tough targets. With MCP support and token-efficient output, it fits naturally into AI agent workflows. Want to give your agents access to what’s happening now? Start for free and get $5 towards your next project 👇 https://t.co/Hj6t2jDjXU @Oxylabs #adfrozen6d ago35K129286090.49%
Have an idea for an app but haven’t started building it yet? Qoder is making Qwen3.8-Flash free for everyone until September 30. You don’t need a paid plan, and the model won’t consume Credits during this offer. Could be a good time to finally tackle that side project. https://t.co/F3ALtrRByR⁠frozen29d ago31K39125060.20%
A Slack message goes in. Ads, SEO, funnel work, social posts, and customer emails come out. Big moves still wait for your yes.2d ago28K1312412050.62%
Voice agents need to understand real conversations, not just clean audio in a studio. Luel Multi-Speaker Separation separates messy conversations into individual voice tracks at up to 48 kHz, trained on real conversations from 900K+ contributors. This could be useful for building better voice agents.1d ago24K14530110110.83%
There's a big difference between a voiceover that sounds fine and one that says your product name right. @OnepinAI checks every take for naturalness and pronunciation, and fixes the misspoken words directly instead of regenerating the whole line.frozen21d ago21K118392090.79%
Behind better AI infrastructure is a lot of research most people never see. Hongyang Zhang, a machine learning researcher and Assistant Professor at the University of Waterloo, is connected to the EAGLE research journey. With EAGLE-4, the team is pushing adaptive speculative decoding further, with a focus on different batch sizes and hardware. Worth a look if you're interested in what makes LLM inference faster 👇 https://t.co/2V4o5SmL2q2d ago20K13828120110.95%
AI that can see, hear, and understand human interaction more naturally feels like a big step forward. Griffin is definitely an interesting one to watch.frozen20d ago15K79152060.70%
most companies when a new device drops: “we’re exploring support.” @autosanaai when a new device drops: “already live for all customers.” that’s it. that’s the tweet. the whole GTM strategy in one sentence.frozen16d ago15K106115070.89%
Enterprise teams, here's the part of the launch for you. Fish Audio covers 3 months when you switch providers. Plenty of time to test Drama 3 properly.frozen9d ago14K2011030.17%
This is a pretty interesting shift. A codebase that lives in the cloud means the whole team can work with it, not just the person who has the local setup. Base Code makes that possible with any GitHub repo.frozen24d ago13K1394720071.63%
The real test for any agent: Can it sit in a group chat without being annoying. Teamily built theirs to jump in when it's useful and stay quiet otherwise. Dots hasn't had to pass that test yetfrozen23d ago12K3437020.37%
Same price for a better model is not something you see much in voice AI.frozen22d ago12K3363090.42%
Most ad teams know iteration is where the wins come from. Very few have the time to actually do it. Fastlane v2 takes that part off your plate. It creates ads, launches them on Meta, tracks what performs and builds new variations of the winners automatically. 80% of the results, 20% of the time. That's the whole idea behind it 🏎️11h ago12K7782050.76%
Visual coding is a lot harder than it looks. PPTBench tested 18,000 reconstructions from 500 real scientific diagrams, and only 2.57% passed every gate with zero recorded defects. The gap between “looks right” and “is actually right” is huge. https://t.co/hRbMmu0XIdfrozen23d ago12K43410150.52%
Video creation went from expensive production to fast iteration being able to fix things with a prompt makes that even betterfrozen28d ago12K4347030.48%
Sabi's roadmap: > Thought-to-prompt > AI tuned to your neural preferences > Intent prediction > Intent-to-action It goes from what you mean to what an agent does. Backed by $50M.frozen13d ago12K2512040.27%
What surprised me most is how little you have to “tell” Griffin what to do. You can show it something, pause, react, or just keep talking, and it actually follows along. That makes the whole experience feel much more natural. I tried it myself and it’s honestly addictive to play with 😄frozen21d ago12K136292041.47%
There’s a big difference between automating research tasks and expanding the space of research.frozen24d ago11K2902030.31%
The interesting part of xLLM isn’t just the throughput. Being able to change things like the tokenizer, data mixture or parallelism without rewriting the whole training setup could save a lot of rework at scale.frozen24d ago10K3419030.46%
The camera stops rolling. The acting doesn't have to be final anymore. 🎭frozen15d ago9.4K3402020.41%
I've spent time tweaking profiles, rewriting summaries, and trying to guess which keywords matter. Seeing a recruiter-style ranking doesn't answer every question, but it gives me a more concrete place to start.frozen9d ago8.1K4952060.77%
Building an AI agent that passes tests is one thing. Building one that understands a constantly changing real-world environment is a whole different challenge. That's what makes NewEyes AI an interesting direction for personal agents.frozen10d ago3.1K6162032.32%