← Back to leaderboard
learnwithsubhan

Subhan Qureshi

Verifiedactive
@learnwithsubhan

Sharing insights on AI, Tech, & Empowering driven Founders and CEOs to build impactful Personal Brands on Twitter |📩 DM For Collaboration Subi2023@gmail.com

RosterTech Career & Coding EducationAI Agents & Productivity Tools
43,477 followers · 747 following · Last polled 3h ago
Perf. Score
66.9
Rank #91
Value Score
92.9
#12 of 273 priced
Median views
24K
p25 6.3K · trailing 7d
Followers
43K
+0.9% 7d
ER (impr.)
0.82%
eng ÷ impressions
ER (foll.)
0.84%
eng ÷ followers
Est. CPM
$1.18
QT-rate basis
WoW views
+39.7%
▲ rising

Rates & value · est. from 24K median organic views

underpriced
Quote tweetbasis
$28
$1.18 / 1K views
Post
$48
$2.03 / 1K views
Thread
$66
$2.79 / 1K views
Retweet
$14
no CPM
Value Score 92.9 · rank #12 of 273 priced$141.41 / 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 (24K), not the mean (45K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (6.3K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.26d ago371K1.1K227169555190.57%
🚨 BREAKING: Claude can now BUILD a YouTube channel from scratch—and hit monetization in just 90 days. 100% free. Here are 8 prompts to make it happen:12d ago230K9032284032.3K1.49%
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
6d ago
187K
929
222
126
30
347
0.89%
A discount gets attention. A creative video earns it. CapCut Pro is 60% OFF and Ultra is 40% OFF, Oct 20–30, only on CapCut Desktop. I used CapCut PC to turn a promotional concept into a more engaging Black Friday ad, combining AI-generated visuals with timeline editing, layered effects, and audio adjustments. From the first visual to the final polish, CapCut PC gives creators the tools to build marketing videos that feel like content, not just commercials. #capcutpc #capcutmarketing #capcutsale #capcutblackfriday3d ago115K775324050.14%
The distance between imagining a great product video and actually making one can feel surprisingly long. This Tesla Roadster concept explores a more direct route. With CapCut Web × Codex, start by describing the video you have in mind, including its overall style and marketing purpose. Codex interprets those instructions and calls CapCut Web to help turn them into a video. The idea is simple: let natural language communicate the creative direction, then let the connected workflow handle more of the execution. #capcutweb #capcutmarketing #capcutcodex #capcutchatgptplugin2d ago111K139102340160.26%
Get CapCut Black Friday savings today with 60% OFF Pro and 40% OFF Ultra. Available October 20–30 on CapCut Desktop only. A screen recording, a voiceover, and a few useful labels can become a much clearer demonstration. Bring them together on CapCut PC's multi-track timeline, trim the waiting time, and align the explanation with the action. Use captions and text to emphasize the steps, then balance the audio so the instructions remain comfortable to follow. Give your next demo a practical edit that helps viewers understand exactly what happens on screen. #capcutpc #capcutblackfriday #capcutsale #capcutmarketing1d ago108K5952210100.13%
Nvidia CEO Jensen Huang said: “AI will create more millionaires in 5 years than the internet did in 20.” So I looked at the people already getting rich from AI. Turns out they all do these 4 things most people never think about:27d ago87K229603023030.71%
CapCut Pro at 60% off? That’s a steal. Get 40% off Ultra, too. Oct 20–30, only on CapCut Desktop. The Jean Phil trend showed how one AI character can become endless content. I tested a new AI video editing workflow: Codex controls CapCut PC through prompts, automatically creating the character, generating the video, and handling the first editing steps. Then I jump back into CapCut PC to refine the timeline and polish the final marketing video. This is what Codex video editing looks like: AI creates the workflow. Creators shape the final story. #capcutpc #capcutchatgptplugin #capcutcodex #capcutmarketing #capcutsale9d ago84K694922020.17%
Nvidia CEO Jensen Huang said: “AI will create more millionaires in 5 years than the internet did in 20.” So I looked at the people already getting rich from AI. Turns out they all do these 4 things most people never think about:11d ago65K247874012921.02%
Some rumors I've heard. I'll start with Google, then Anthropic and OpenAI. Google: Insiders say Google has either cracked RSI (recursive self-improvement) or is very close. Their pace has picked up fast in recent months. They went from GPT-5 level to Opus 5.5 level in months, and speculation is that RSI made that possible. Anthropic: About a week ago, Anthropic started a new pretraining run. It is not Fable 5.5 but what would become Fable 6, or something bigger, to catch up to OpenAI's Bel. OpenAI: Bel (it may have another codename) is the model behind the math results OpenAI published a few days ago, and RSI apparently helped there too. From what I'm hearing, OpenAI, Anthropic, and Google are all very close to cracking RSI or already have. It will get scary and exciting at the same time.6d ago62K27093462840.80%
What if the biggest limitation of AI assistants isn’t what they can do but what they need to know before they can help? I recently came across Littlebird, and that’s the part that grabbed me right away. Most AI assistants require a ton of setup: you explain the project, paste prior threads, upload files, recap context—then finally ask the real question. It gets old fast, especially when work stretches across days, and the details are scattered across apps and conversations. That’s where @LittlebirdAI takes a different approach. It’s built to learn from the work you’re already doing and use that context when you need help. Your computer activity, conversations, meetings, and other signals that actually matter. What I’m most curious about is how well this holds up in real life: If I come back to a project after a few days, can I continue without rebuilding the entire context? If I need something from an earlier meeting, can I pull it up without digging through threads? Can routines handle the recurring tasks I normally have to keep in my head? Here’s what I’ll be exploring in littlebird: → Chat to work with information from my existing activity → Meeting notes to capture and organize key takeaways → Routines for recurring tasks and updates → Hummingbird for help while I’m working without constant app switching I’m not making big claims yet. I’m just getting started, and I want to see what it does with real work. I’ll test the features, figure out where it fits in my workflow, and share what I learn along the way. That’s what interests me most about littlebird: not whether AI can remember everything, but whether it can understand enough of the right things to be genuinely useful. Explore littlebird: https://t.co/P3OEFMa3nC #ad16d ago60K1399945090.49%
AMAZON PRIME COSTS $139 A YEAR, AND MOST MEMBERS USE TWO FEATURES: Free shipping and Prime Video. That's it. But Prime is a giant bundle, and some of the perks you're already paying for quietly expire every month. Here are 12 most members never touch 🧵28d ago54K180904212190.98%
Claude Opus 5.5 makes “I have a game idea” feel a lot more dangerous in the best way. A distinctive world, a mechanic worth figuring out, and a game that actually feels complete. This is DEAD STATION - Last Signal, from the Combos Fun community. With Combos Fun @CombosFun_AI, you can take an idea from an Opus 5.5 prompt, put your own spin on it, and turn it into something people can actually play and share.29d ago54K1118839070.45%
🚨 BREAKING: Claude can now BUILD a YouTube channel from scratch—and hit monetization in just 90 days. 100% free. Here are 8 prompts to make it happen:5d ago53K4361253548382.72%
I already dropped my work handbook into @LittlebirdAI Chat. How I run a new project. How I find and vet KOLs. Growth, KOL management, BD. That’s the work—not “I make content.” I didn’t start from zero today. I asked it to write a one-page note I can send for a project this week—using only what’s already in the handbook. Don’t rewrite my bio. Don’t invent rates or deal terms. It came back with the exact process I already use: Learn the product first. Define the audience before you talk to creators. Vet talent against the brief—not a random list. Send a shortlist. Onboard. Follow up when posts go live. No “tell me about your business.” No long-form. That’s why I keep using it. It already has the context. I ask for the next thing, and I get work I can send. The video is the chat. Not a website. Not a fake screen. If you’re still spending half your time explaining your job to the AI—that’s the difference. https://t.co/9hNRwbDG0o22d ago47K12887360100.55%
❌ iPhones that AREN'T WORTH A DAMN: ・iPhone 16e ・iPhone Air ・iPhone 17 Pro ・iPhone 14 Pro ・iPhone 17 Pro Max ✅ iPhones you SHOULD buy ↓24d ago44K14282340520.71%
A 19-year-old Japanese student built a trading bot with Claude Code in 2 days. Used his iPad as a second monitor. First night: $6,732 profit. Starting capital: $68. Total profit so far: $750,000. Here's how it works: The bot scans over 50 markets simultaneously. Syncs live BTC data from Binance every second. Spots price errors before humans even notice. The edge is pure speed + pattern recognition. While traders stare at charts trying to predict the next move, his bot is already executing on mispricing across dozens of markets. No guessing. No emotions. No hesitation. Just Claude Code logic finding gaps that close in seconds. He built the entire system in 48 hours: → Claude Code handles the trading logic → Binance API feeds real-time BTC data → iPad displays multi-market monitoring → Executes trades when arbitrage windows open The system runs 24/7. Every price dislocation = profit opportunity. Most people are still trading manually, refreshing charts, second-guessing entries. Meanwhile this 19-year-old engineering student turned $68 into $750K by letting Claude Code do what humans can't: process 50 markets instantly and execute without fear. Why are people still trading manually? 💡 I'm giving away the exact Claude Code setup for free. 24 hours only. To get it: 1️⃣ Comment "Fable" 2️⃣ Like and Repost 3️⃣ Follow @LearnWithSubhan I'll DM you the complete setup.10d ago42K37316221021902.21%
Anthropic has just released a 37-minute guide explaining how to create AI Agents capable of automating an entire business. ( THREAD🧵 The rest of the video is in the comments. ) It's free and it comes straight from the engineers who created Claude. Agents that actually work, delegate tasks, and complete their work autonomously. 🔖 Bookmark this guide. You might need it sooner than you think.25d ago37K3331415426793.26%
imagine watching the final cut and thinking: “the scene is perfect, just change the emotion.” now you can.15d ago34K1107530030.63%
One more AI coding option to add to your toolkit. DeepSeek-V4.1-Flash-1017 is available on Qoder at 0.0x Credits through Oct 24. Both free and paid accounts can use it during the offer. Download Qoder, sign in, and select it from the model picker. https://t.co/7zcTdYInfV3d ago31K856231040.58%
My sister sent out 47 job applications. Not a single callback. Not one interview. Then I ran her resume through ChatGPT with 7 targeted prompts. 13 replies in 7 days. Dream job offer accepted. Here's exactly what I typed:28d ago31K223751823932.30%
🚨 BREAKING: Claude can now BUILD a YouTube channel from scratch—and hit monetization in just 90 days. 100% free. Here are 8 prompts to make it happen:27d ago31K245973523642.43%
Stop telling ChatGPT “Make a graphic.” Stop telling ChatGPT “Design a flyer.” Stop telling ChatGPT “Create a poster.” Weak prompts = Poor designs and AI slop. Try these 7 powerful commands to get much better results from ChatGPT: 👇4d ago27K2631103213252.67%
Ling-3.1-flash-1019 is raising the bar. Intelligence Index: 44. AA-Briefcase: 1524 Elo. AutomationBench: 68%. The gains in agentic tasks are what stand out here. Now imagine developers getting access to the API and open weights. That could make this release worth watching closely.9h ago26K746125060.63%
Apple sold you an iPhone with its top camera settings switched off. On purpose. $1,200 for a camera you've never actually used. Six settings are off by default, and they're the ones that can make your photos look dramatically better. Fix these 6:20d ago26K172924111861.87%
Free accounts + Qwen3.8-Flash + 0 Credits? 👀 Qoder is making that combo available through Sept 30. If you’ve been curious about AI coding agents, this is a nice opportunity to actually test one instead of just reading about it. Download Qoder, sign in, and select Qwen3.8-Flash: https://t.co/jnrSnEbSKP29d ago26K644517060.51%
“Good morning. Please continue on page 212.” Now that's a more interesting kind of AI memory than just another chat history. NewEyes AI is bringing personal agents into the world around you, not solely confined to the screen.10d ago24K644924030.59%
AI assistants are useful, but they often miss the context behind your work. So you end up re-explaining the project, re-sharing files, pasting links, and rebuilding the background before they can help. @littlebirdai is built on a different idea: your AI should understand the work happening around you. It can read what’s on your screen, take notes in meetings, remember what matters across your projects, and pull context from the apps you already use when it’s needed. So instead of starting every chat with a long explanation, you can ask for what you need. The more you work, the more useful it gets. Littlebird is available on web ➜ https://t.co/P3OEFMa3nC #ad15d ago23K513419070.48%
It's incredible! We now have an open-weight model release from @OpenBMB: MiniCPM-V 4.7 35B-A3B. This model is a replacement for Qwen3.6-35B and performs better across various benchmarks. The weights are available for download on Hugging Face. I'm really looking forward to trying this out.9d ago22K34236010.29%
Okay, I knew MoE was going to be big for local inference, but running this stuff on actual phones?? That's a different level of crazy lol.9d ago22K271811050.28%
It's interesting how I use these two tools every day, yet I needed an article to realize I could combine them.6d ago21K34199070.33%
Open-source AI is getting seriously powerful. Ling-3.1-flash brings 560B parameters, 1M context, and the ability to turn rough ideas into websites, mobile screens, and interactive prototypes. Even better, optimized weights can cut local serving requirements from 8 GPUs to 4. This is the kind of progress developers need.7d ago19K2159020.19%
Hot take 🔥: for vibe-coded games 🎮, rigging by hand is done 👋. Viggle's JST-2 turns one image 🖼️ into an animation-ready character ✨ with no rigging or skinning step 🙌, and your agent can call it through the PINOC MCP server 🤖. Why model anything by hand 🤔? https://t.co/i7gJUFFAC13d ago19K956023050.98%
Inworld Inference processes tens of trillions of tokens every month, with a significant majority originating from applications that individuals incorporate into their daily routines. It appears that open models are quite capable of managing consumer applications effectively.8d ago15K1124010.12%
The World Models’ Last Exam in Physics has arrived and might serve as the SWE-bench for evaluating physical consistency in video world models. It features 8 state-of-the-art models, 40 controlled tasks, and 9 physics categories, all assessed through direct programmatic measurements.. The highest score recorded was merely 57.76 out of 100, whereas videos of free-fall averaged 26.61 out of 100. There is a substantial disparity that needs to be addressed, and the individual who successfully resolves it has the potential to influence the development of the subsequent era of trustworthy world models.14d ago15K725831041.08%
Every video style guide I've created has been accessed perhaps a couple of times. If Syren simply picks up the style by watching the videos, I'm perfectly fine with that.13d ago15K22147010.30%
Excellent work from @MaxMa1987 and the team! xLLM is a truly well-designed pretraining infrastructure that is user-friendly and supports several advanced architectures. I've been utilizing it for my looped model projects for quite some time now. 🫡24d ago15K675425021.00%
I've been creating voice packs for game mods for a long time, always for characters made by others. However, for the Chroma Awards, I'm developing my own cast and providing their voices using @ElevenLabs.3d ago15K25178020.35%
Ming-Image-0.1-Design feels less like an image generator and more like a design copilot. Give it a prompt or screenshot. It can create the visual direction, generate assets, turn graphics into editable layers, write the UI, validate it, or build an editable deck. That workflow is wild.29d ago14K362219050.57%
Can’t believe people still don’t know how to profit from ChatGPT! Earn $1,000/Day with ChatGPT 💰 [100% GUARANTEED] Here’s all you need: 📱 A smartphone 🌐 WiFi ⏳ 60 minutes a day 👉 Like and Comment "Send" to get it via DM – FREE for the next 24 hours! 🕒 (Make sure you’re following @LearnWithSubhan to receive the DM)1mo ago14K234911920614.01%
A 19-year-old Japanese student built a trading bot with Claude Code in 2 days. Used his iPad as a second monitor. First night: $6,732 profit. Starting capital: $68. Total profit so far: $750,000. Here's how it works: The bot scans over 50 markets simultaneously. Syncs live BTC data from Binance every second. Spots price errors before humans even notice. The edge is pure speed + pattern recognition. While traders stare at charts trying to predict the next move, his bot is already executing on mispricing across dozens of markets. No guessing. No emotions. No hesitation. Just Claude Code logic finding gaps that close in seconds. He built the entire system in 48 hours: → Claude Code handles the trading logic → Binance API feeds real-time BTC data → iPad displays multi-market monitoring → Executes trades when arbitrage windows open The system runs 24/7. Every price dislocation = profit opportunity. Most people are still trading manually, refreshing charts, second-guessing entries. Meanwhile this 19-year-old engineering student turned $68 into $750K by letting Claude Code do what humans can't: process 50 markets instantly and execute without fear. Why are people still trading manually? 💡 I'm giving away the exact Claude Code setup for free. 24 hours only. To get it: 1️⃣ Comment "Fable" 2️⃣ Like and Repost 3️⃣ Follow @LearnWithSubhan I'll DM you the complete setup.24d ago14K1631111311653.34%
Inworld provided the voice models and serving infrastructure, while Ultravox handled the agent side. It seemed like a fairly obvious synergy, to be honest.22d ago14K271412040.41%
I produce audiobooks for a few indie authors. Using the free v4 to send them some samples has been a big unlock.21d ago14K291517010.44%
The intriguing aspect of Recast lies in the fact that the meme itself can serve as the demonstration. A recognizable scene, featuring a different character, instantly reveals the potential.9d ago14K1061020.14%
All Paid Courses (Free for First 5000 People) 𝗣𝗮𝗶𝗱 𝗖𝗼𝘂𝗿𝘀𝗲 𝗙𝗥𝗘𝗘 (PART - 1) 1. Artificial Intelligence 2. Machine Learning 3. Prompt Engineering 4. Claude, Chatgpt, Grok 5. Data Analytics 6. AWS Certified 7. Data Science 8. BIG DATA 9. Python 10. Ethical Hacking (72 Hours only ) Like + RT + comment 'drive' Must Follow me so I can DM you.26d ago14K139103911402.74%
The most striking feature for me is the emotion engine. The avatar doesn't just move its face; it genuinely responds to the conversation and adapts its behavior, which makes the interaction feel significantly more natural.28d ago13K241511040.42%
STOP telling ChatGPT “edit my photo.” That’s why the results look fake and overdone. Use these instead:19d ago12K116663811332.94%
🚨 Instead of watching Netflix tonight... spend just two hours with this video! ⏳🧠 The complete Claude AI course that teaches you how to build and automate anything! The people who watch this course tonight will wake up tomorrow with a completely new skill. Watch it and save it now! 🧵👇25d ago11K10570340602.35%
Can’t believe people still don’t know how to profit from ChatGPT! Earn $1,000/Day with ChatGPT 💰 [100% GUARANTEED] Here’s all you need: 📱 A smartphone 🌐 WiFi ⏳ 60 minutes a day 👉 Like and Comment "Send" to get it via DM – FREE for the next 24 hours! 🕒 (Make sure you’re following @LearnWithSubhan to receive the DM)13d ago10K2091001530354.84%
🚨 Instead of watching Netflix tonight... spend just 10 mins with this thread! ⏳🧠 The complete Claude AI course that teaches you how to build and automate anything! The people who watch this course tonight will wake up tomorrow with a completely new skill. Read it and save it now! 🧵👇17d ago8.0K157108510634.72%
Claude Opus 5.5: bringing web development to GTA V Finally, a reason to learn to code. Credits @adonis_singh12d ago7.1K11978500123.67%
GPT 6 Sol vs Opus 5.5 in a nutshell :/ https://t.co/TTCS9Kr0B219d ago6.2K1037537083.61%
holy sh*t. Official Anthropic tip for Claude Code: stop burning Opus 5.5 context on tasks Haiku 5.5 can swarm, while Sonnet 5.5 builds put the architect on call with /advisor claude --advisor opus --subagents haiku Sonnet 5.5 drives the primary session on high effort, writing diffs and running test suites Haiku 5.5 subagents swarm the repo in parallel at 340 tok/s, handling file discovery and spec docs lookup Opus 5.5 stays on call in the background as the advisor, only stepping in at three critical moments: → before a plan locks: does the implementation plan miss auth invariants or schema contracts? → when a test breaks twice: are we fixing the root cause or falling into a recursive rabbit hole? → before calling done: did the full diff introduce hidden regressions or break pre-flight rules? Opus 5.5 reviews. Sonnet 5.5 builds. Haiku 5.5 swarms Jev engineering runs the identical pattern one layer down: mechanical decisions that need no reasoning (which file to open, which tool to invoke, retry or abort) execute in 16ms, so the frontier models only wake up when execution paths actually diverge Plan on high. Delegate on medium. Keep Opus on call. - the full advisor setup > Opus 5.5 on call reads full session history and catches deep architectural traps > Sonnet 5.5 lead drives edits, writes core logic, and executes test harnesses > Haiku 5.5 subagents swarm AST parsing, grep scans, and API docs in parallel > JEV micro-fork layer resolves 1,500+ mechanical routing branches in under 16ms > advisor stays completely silent on routine bash execution to protect context paste the setup and this prompt into Claude Code below: "Configure my Claude Code workspace for hierarchical advisor orchestration: 1. Audit ~/.claude/settings.json and project config for model roles fitting lead, subagents, and advisor: > Pin main session model: sonnet with effortLevel: high for primary execution > Pin subagents model: haiku with effortLevel: medium for parallel file discovery and docs retrieval > Pin advisor model: opus on call for strategic review 2. Enable the advisor tool in ~/.claude/settings.json: > Set advisorModel to claude-opus-5-5 > Enable subagent parallel dispatcher pool (3x workers) 3. Configure automatic advisor consultation checkpoints in ~/.claude/CLAUDE.md: > Consult /advisor opus before finalizing multi-file architecture plans > Automatically summon /advisor opus when the same test or compiler error fails twice > Enforce advisor diff contract audit before declaring tasks complete or staging git commits 4. Constrain subagent scopes: > Haiku subagents return structured AST summaries and docs snippets only without modifying main session context Show every configuration diff first. Do not apply edits until confirmed" ↳ https://t.co/bP6qHlKLmu10d ago5.9K11184482495.02%
Industrial software doesn’t have to feel like industrial software Manage warehouses like you’re playing a strategy game Built with Opus 5.5 https://t.co/Q6Cj547YgS9d ago4.9K14893411406.56%
I asked Opus 5.5 to explain camera focus by building an interactive lens lab Here's what it came up with after 1 hour 26 minutes in one shot, $25.66 API cost https://t.co/jPS4gdmSgr Move the focus ring and you can see the glass elements shift the sharp plane through the scene https://t.co/HI7MIyw3hI18d ago4.4K13491561116.59%
🚨 BREAKING: Claude can now build you a LinkedIn profile that gets you hired in 30 days. For free. Here are 7 prompts to access it:13d ago4.3K9467302505.64%
This film cost $29,575 and took 10 days to make. AI ≠ Cheap Our Nexus feature will still cost millions next year. But it'll look like a $200M film. I have a Claude skill you can copy that makes this easy. Full process and Dreamina prompts below👇🏼🧵 https://t.co/rx1YVxA9zq20d ago3.9K8150360355.14%
THIS IS F*CKING UNFAIR opus 5.5 put Minecraft inside GTA 5, and then it got worse they did not copy Minecraft into the game. they wired in the part of Minecraft that loads content WHAT THAT MEANS → you install a mod in Minecraft the normal way, in its own folder → it shows up in GTA 5, standing in the street, lit by the city's own lights → nobody touches GTA for any of it WHY IT WORKS → the bridge sits under the content instead of beside it. GTA is never told what a block is, it is told how to load whatever Minecraft is loading → so new content is not a new port, it is a file in a folder → the ceiling stopped being what one person built and became whatever the modding scene makes next WHAT IT QUIETLY UNLOCKS → fifteen years of mods are now GTA content that nobody had to make → the people who wrote them have never heard of this and do not need to → one engine inherited another engine's entire catalogue in an afternoon the first version put Minecraft in GTA this one put everything Minecraft can load in GTA, and the city still thinks it is a normal tuesday11d ago3.8K1198141076.56%
Played around with an idea for a wildfire analysis tool Visualizes wildfire progression over time, tracks hot spots with drones and monitors firefighting efforts Made with Opus 5.5 https://t.co/3UMTbVE7Qx8d ago3.7K13490390127.44%
Made with Opus 5.5 + Three.js https://t.co/PcPklrp0JB the water physics, the fish, blossoms... why is this soo pretty 😭😭 https://t.co/czysTl4vej7d ago3.4K13087360137.72%
🚨 GEMINI PRO GIVEAWAY 🎁🤖 ​Want to unlock the absolute best of next-gen AI and supercharge your workflow? I’m giving away exclusive access to the Gemini Pro Ultimate AI Suite! 🚀 ​🎁 What You Win: ✨ Gemini 2.5 Pro Model — Full Access & API 💰 1,000 AI Credits/Month 🎬 Veo 3, Flow AI & Whisk (Advanced Video & Creative Workflows) 📚 NotebookLM & Workspace (Gmail & Docs Integration) ​👇 How to Enter: 1️⃣ Follow @LearnWithSubhan and Comment "Send" 2️⃣ Like & Repost (RT) this post 🔄 3️⃣ Tag 2 friends below who need this! ​⏳ Winner will be randomly picked and announced in 48 hours. Drop a comment now and get in! ⚡5d ago3.1K1128644058.00%
Have you heard of this novel? It’s the “literary back-to-school phenomenon.” It’s the winner of the Prix Fnac, the Prix Méduse, the Prix Première Plume, strongly tipped for the Prix Renaudot, and it’s in the running for the Prix Goncourt, Femina, Médicis, and Décembre. Oh yes, it’s also in the process of being translated into 20 languages. Well, according to Pangram, this novel was written almost entirely by an Artificial Intelligence. We tested several passages, at various points in the manuscript, and got almost always the same result: 100% AI (see the first two chapters in the tweet below). Let’s recall that Pangram is by far the most reliable AI detector: its false positive rate is 0.0041%, or about one human text out of 24,400. Out of 996,273 texts prior to 2022, Pangram classifies only 14 (0.0014%) as partially written by AI. The probability that twenty passages from the same book (and there are far more than 20 flagged sections in the novel) are misidentified by Pangram as being produced by an AI is therefore 0.0041% to the power of 20, or 0.1802 × 10⁻⁸⁵ %, which is 0 followed by 85 zeros after the decimal point. Wouldn’t it have been more honest for the author to credit their co-author?21d ago2.7K10576370108.37%
$500 GIVEAWAY - 3 WINNER 🎁 • Follow @LearnWithSubhan • Like & RT • Tag a friend & Put on Notis Winner chosen in 24 hours ⏱️ https://t.co/VublFeKdeO4d ago2.4K1138635059.98%
Introducing Contrastive Language Model (CLM): an ultra-fast System One Model trained with a contrastive learning objective that connects states and actions. CLM-8B is pre-trained on internet-scale data and delivers up to 9× faster inference than Jev ⚡ while achieving comparable performance across computer-use, gaming, and tool-calling tasks. With lightweight fine-tuning, CLM-8B sets a new SOTA on challenging agentic coding benchmarks, such as DeepSWE (81.6%) and Terminal-Bench 2.1 (87.6%). In contrast, Jev fails to serve as an effective verifier for these long-horizon tasks. We also build an efficient training and serving infra for CLMs by disaggregating states and actions, allowing their embeddings to be cached and reused independently. This substantially reduces inference latency in settings where the state evolves continuously while the action set remains fixed. Finally, we establish scaling laws for CLMs and show that the test contrastive loss decreases predictably as a power law in training compute, model size, and dataset size. 📄 Blog: https://t.co/iINxXEXPmX 💻 Code: https://t.co/6RKgrM9Bgt 🗣️ Discord: https://t.co/q2UmrCxeOX 🤗 Data & Models: https://t.co/CWXY79jKlA More details on CLM’s architecture, data recipe, and scaling laws in the thread below 🧵17d ago2.4K977026158.42%
Open Al's internal benchmarks show GPT-6.1 Sol crushing Opus 5.5, with Anthropic struggling to keep up https://t.co/8eLOXxCJiq18d ago2.1K785325057.70%
Made with @claudeai Opus 5.5 https://t.co/QbVfaQVmgy21d ago2.1K635226077.18%
when a designer gets access to Jev https://t.co/XdLnTGCQ2b16d ago2.0K835926078.72%
I used to lose the first 30 minutes of every workday just figuring out what I should be doing. Now that part’s done before I even sit down. No new habit. No planner. No sudden burst of discipline. Here’s the setup: 👇 Littlebird has a feature called Routines. You write a prompt once, choose a schedule, and it runs automatically—daily, weekly, monthly, whatever you want. The difference vs. a normal reminder app is what it runs on. @LittlebirdAI reads what’s on your active window and takes notes in your calls, so each Routine is based on what you actually did—not a to-do list you forgot to update. Four Routines worth turning on day one: 1/ Morning briefing What’s on today, what’s still open from yesterday, and what needs a reply. It’s a ready-made Routine, so you can switch it on in a minute. 2/ Yesterday’s work summary For mornings when your brain boots slower than your laptop. Instant context on where you left off—so you can just continue. 3/ Weekly activity summary Every Friday, a clean recap of what you worked on. Great for status updates, client reports, or just seeing the week clearly. 4/ One custom Routine for your role Mine flags open loops from calls—anything someone promised, anything I promised—captured in one place. Write one that matches how you actually work. Best part: it asks almost nothing from you. Set it up once, and it quietly shows up with your day already organized. Link if you want tomorrow’s morning briefing ready to go: https://t.co/P3OEFMa3nC New users get $20 off the first purchase. The free plan is enough to try the morning briefing and see how it feels. #ad14d ago1.9K836128089.33%
I'm hosting my FIRST GIVEAWAY🎁 Win 1x🎧Xiaomi Type-C Earphones 1 Question : Which Xiaomi phone has been your favourite of all time ? To participate: 1. Follow @LearnWithSubhan 2. Like❤️ & repost🔁MUST FOLLOW 3. Reply with your answer 4. Tag 2 Friends Winner selected randomly Good Luck🤞16h ago1.5K9677350513.79%
Solo practitioners especially, honestly28d ago1.4K200000.14%
when a designer gets access to Jev https://t.co/afBpHzgd1I19d ago912301000.44%
$100 sol giveaway max engage last rt drop sol addys.1d ago80000000.00%