← Back to leaderboard
amank1412

Aman

Verifiedactive
@amank1412

Engineer | DM for work/collab

RosterAI Model News & ReleasesHumor & Meme Accounts
15,242 followers · 1,206 following · Last polled 4h ago
Perf. Score
44.9
Rank #268
Value Score
45.8
#154 of 273 priced
Median views
1.1K
p25 696 · trailing 7d
Followers
15K
+0.2% 7d
ER (impr.)
3.42%
eng ÷ impressions
ER (foll.)
0.74%
eng ÷ followers
Est. CPM
$55.4
QT-rate basis
WoW views
+116.1%
▲ rising

Rates & value · est. from 1.1K median organic views

overpriced
Quote tweetbasis
$60
$55.4 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 45.8 · rank #154 of 273 priced$1791.04 / 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 (1.1K), not the mean (3.3K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (696) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
GERMANY JUST ENTERED THE FRONTIER LLM RACE. Meet Kolibri a new sovereign open source model. > Beats Qwen2 635B > Beats Nemotron 3 Super > Beats Mistral Small 4 Built on a custom MoE Transformer architecture. https://t.co/zIXdHpib8efrozen19d ago672K1.9K81133457730.43%
openai and anthropic watching germany enter the ai race: https://t.co/tYYJK2WqmFfrozen18d ago522K3.4K6757107020.81%
cybersecurity startups trying to raise funding right now https://t.co/nxKv5fZmIJ
10h ago
90K
4.2K
277
41
34
458
5.60%
BRO WHAT!!! just built SkyGarden a full browser game with Space Bunny from a single prompt. Ran so fast and I got the result after a few minutes. you fly a tiny plane over a green city, collect rings, dodge buildings & balloons, manage boost and build combos. didn’t expect it to turn out this goodfrozen28d ago54K91008001.69%
Apple hired 24 Interns from nsut at 1.10 LPM 🤯frozen27d ago45K3353144320.86%
well... “family business with a $900B valuation” https://t.co/V3Kqub9M2mfrozen18d ago44K356680841.04%
EAGLE-4 IS MAKING AI INFERENCE FASTER. → Beats DFlash 2 and DSpark in benchmarks → Up to 3.59× faster than vanilla decoding → Better performance under high loads → Cuts draft model overhead by 40%+ Always good to see open-source inference research pushing things forward. https://t.co/accSYSe0Q32d ago16K7802010.50%
I GAVE TEAMILY AI A WEBSITE TO BUILD. finally not another one person + a bunch of agents tool. Teamily AI is a Humans + Agents platform where multiple people and agents can work in the same group with shared context. → one handled the frontend → one built the backend on top of it → one reviewed the code I built an AI developer portfolio for startup founders, then changed the direction midway. the best part? the team adapted without starting from scratch. then my teammate gave feedback and the agents updated the live site in Website Builder. this feels way closer to having an actual AI team than just using a chatbot. Teamily AI is a human-agent collaborative workspace, great for small businesses and teams. built with Website Builder @Teamily_AIfrozen14d ago12K741616080.97%
AI CAN GENERATE VIDEOS, BUT THE CREATIVE WORK STILL MATTERS. VN Studio helps you make AI films while tracking your creative input. → create characters and scenes with AI → edit and refine your videos → track your creative work and show what's yours AI should help creators do more, not take away their ownership.frozen6d ago7.1K52110091.02%
THIS IS A HUGE STEP TOWARD AUTONOMOUS MANUFACTURING. @SkyfallAI is using world models to break down complex engineering designs into individual components. And doing it up to 100x faster and 150x cheaper than GPT Astra is wild. https://t.co/fIuHuhfNnV3d ago6.8K61413091.27%
AI AGENTS ARE GETTING BETTER AT DOING THINGS FOR US. but giving them access to our accounts and payments is a different story. → let agents log in for you → make payments without exposing your card details → set limits on where and how much they can spend that's what @selfxyz is building with Agent Vault.frozen7d ago5.9K4928011.02%
WHAT IF YOUR CODING AGENT ACTUALLY LEARNED FROM ITS MISTAKES? Beacon turns your past coding sessions into reusable skills so your next agent doesn't have to start from scratch. Open source makes this even better. Congrats on the launch!2d ago5.4K3617010.83%
Fable 5.5 will launch in next 10-12 hour's Are you excited for it?2d ago5.3K59123131.65%
i think this is what makes AI driven attacks so different. a human attacker has limited time and attention. an agent swarm can keep exploring thousands of possible paths. 700 agents and 17,000+ actions is a pretty good example of how small weaknesses can eventually turn into a path to critical systems.frozen14d ago5.2K3505080.92%
AI INFERENCE IS MOVING FAST. Flint shows how capable coding models can run fast on your own hardware. → 125+ tokens/sec on a MacBook Pro → 150+ tokens/sec on an RTX Spark → Open weights and local inference Built from scratch in under 4 months using Recursive’s Eureka Machine which automates parts of AI research. Exciting to see capable coding models getting faster and easier to run locally.1d ago5.0K51111061.37%
THIS IS SO COOL!! Most robotics companies focus on what their robots can do but @roboticscorp is thinking about what it feels like to actually live with one. Voxel has so much character. The design feels straight out of a Pixar movie. This is the kind of robot I'd actually want in my home.2d ago4.9K6509011.53%
THE HARDEST PART OF STARTING A BUSINESS MIGHT NOT BE BUILDING THE PRODUCT ANYMORE. you just wanna focus on building the product and getting customers, not spend hours figuring out incorporation, banking and paperwork. rho putting all of this in one place is actually pretty cool. it makes the whole process a lot easier.frozen6d ago4.7K43290121.39%
AI AGENTS SHOULDN'T JUST GIVE US ANSWERS. they should be able to build the right interface for the task. → maps for planning trips → dashboards for analyzing data → interactive UI for comparing products that's what @openuidev is bringing to AI agents. https://t.co/pFGT7jUcckfrozen7d ago4.6K3704030.95%
holy shit, security teams needed this yesterday. https://t.co/LSo5N4rQau4h ago4.6K54210111.48%
AI made creating designs effortless, but most of them still end up looking the same. Vyb is taking a different approach by understanding your taste and building designs around your actual brand assets. The best part? Every element is editable, so you can tweak your design without starting from scratch.2d ago4.2K3503010.94%
THE NEXT STEP IN AI APP BUILDING PROBABLY ISN’T JUST ABOUT BUILDING FASTER. It’s about making feedback part of the build loop. → comment directly on the product → send feedback to the agent → iterate without the back and forth Bolt is making build → review → revise feel like one workflow.frozen15d ago4.0K5339021.67%
I THINK QA IS ABOUT TO CHANGE A LOT. → Coding agents write the code → QA agents test the actual product → Bugs get reproduced and sent back with context Ship feels like the missing piece here. This is starting to feel like vibe coding, but for QA.17d ago4.0K40113001.36%
BRO WHAT!!! this is an AI companion that literally lives on your desktop. you can talk to it about your work, ideas, code, or pretty much anything. >no new tabs. > no switching between apps. > it’s just there whenever you need it. ngl this is actually pretty cool. https://t.co/aRw1ORZOxTfrozen19d ago3.8K54815322.14%
BUILT A 3D ANALYTICS DASHBOARD WITH MINIMAX M3.1 FLASH  meet Halo a dashboard with a live 3D globe tracking sales, animated revenue charts, real time transactions, and interactive 3D visualizations. built with a single prompt, and M3.1 Flash costs just $0.15/M input and $0.60/M output tokens. flash models are getting seriously good at building complex UIs.2d ago3.8K3936051.41%
recently my girlfriend broke up with me and i feel like an AI like rene could've helped us communicate better 😭 gonna try using it to patch things up. if not guess i'll find someone new lmao 😭frozen6d ago3.5K3106061.22%
this is what training data looks like: https://t.co/fIdSLDo3pn18d ago3.5K1500040.54%
he’s writing C faster than people write their prompts https://t.co/ksPGxoZ281frozen18d ago3.3K3803071.44%
THE NEXT LEAP IN AI PROBABLY ISN’T JUST ABOUT MAKING LANGUAGE MODELS BIGGER. LLMs are impressive but truly autonomous AI may need something beyond language. → world models → recursive self improvement → understanding and interacting with the real world The interesting part is what happens when intelligence itself becomes automated. That could change how we think about the future of work, businesses, and society.frozen17d ago3.3K4225031.57%
i’m seeing more AI agents become actual users but most apps still don’t know who’s behind them. AgentID could be the identity layer this ecosystem needs.frozen16d ago3.3K3608021.40%
VIDEO AI IS STARTING TO FEEL MORE LIKE AN ACTUAL CONVERSATION. Griffin can see, listen and respond in real time. → understands visual context → reacts to what’s happening → talks while listening The full duplex part is probably the most interesting piece here. Feels like a big shift for video AI.frozen21d ago3.3K56015032.27%
INFERENCE IS BECOMING A BIGGER BOTTLENECK FOR AI. @general_compute is partnering with @positron_ai to bring Atlas inference systems to its platform with Titan coming next. > Atlas today > Titan next > Available Q1 2027 More options for scaling AI inference.2d ago3.2K3724021.40%
AI MODELS ARE GETTING SMARTER, BUT REAL WORLD TASKS ARE STILL SLOW AND EXPENSIVE. What if the problem isn’t the model, but the computer it runs on? → built for AI, not humans → works across browsers, apps and files → runs multiple tasks in parallel @PineAIAssistant is rethinking how AI gets work done.frozen13d ago3.2K60220092.85%
BRO WHAT!!! now you can run your business through a chat. Bland lets you create an AI agent for your business just by sharing a link. it connects to your APIs, handles tasks, and even accepts payments through chat.frozen7d ago3.2K33060101.54%
MOST AI AGENTS GIVE YOU ANSWERS. THE INTERESTING ONES ACTUALLY GET THINGS DONE. → live price checks → personalized recommendations → booking-ready options in seconds @OdessiaTravel is bringing a different approach to travel planning. https://t.co/9WzOZLlaN8frozen8d ago3.1K3607021.44%
BUILT AN ENTIRE FLOATING WORLD WITH MINIMAX M3.1 FLASH   meet MiniVerse a browser game where you can build your own tiny world with 25+ objects, 5 themes, terrain editing, and dynamic day/night cycles. M3.1 Flash delivers strong coding and agentic performance at just $0.15/M input and $0.60/M output tokens. pretty crazy what you can build at this cost.2d ago3.1K45313032.06%
why are Chinese AI labs so quiet rn? haven’t seen any major launch or upcoming model announcements lately. what are they cooking?frozen6d ago3.0K2228201.12%
GROWTH AND MARKETING TEAMS ARE ENTERING THE AGENT ERA. Nectar is building an AI growth stack that helps brands understand what's happening across social media and turn those insights into action. → Find creators → Catch untagged mentions → Connect social activity to revenue They've already attributed $150M in revenue for brands through social.6h ago2.9K4507182.08%
APIs ARE BECOMING THE INTERFACE FOR AI AGENTS. What if APIs could learn from every error instead of making agents hit the same wall again? → real time API logs → fixes built into error responses → docs that improve with real usage @restless_hq is rethinking how APIs work in the agentic era.frozen10d ago2.8K4918032.14%
Fable 5.5 is coming guys.3d ago2.7K11619034.71%
everyone is focused on building smarter AI models, but the knowledge behind them matters just as much. falconer brings internal knowledge and external docs together in one place, making it easier for AI agents to work with your context instead of starting from scratch. this is the kind of infrastructure we'll need as AI agents start doing more real work.1d ago2.7K3606031.69%
AI APPS ARE FINALLY GETTING OUT OF LOCALHOST. build something for your group chat, share it like a Google Doc, and let everyone tweak it with their own AI. this is actually wildfrozen15d ago2.6K3305041.62%
ONE UNDERRATED LESSON FROM BUILDING CLOUD AGENTS: api compatibility is not the same as harness compatibility. unreal labs found that running kimi k3 through an anthropic compatible endpoint caused tool results to be silently dropped. the model is only one part of the stack. the harness and runtime matter just as much.frozen9d ago2.6K3005011.39%
AI models are getting better at understanding videos, but catching every small movement and detail is a different challenge. Glimpse is trying to solve this by helping AI models: → track movements over time → understand changing camera angles → capture small details and events in videos curious to see how far this can go.frozen6d ago2.5K494101112.94%
CHARGEBACKS ARE SUCH A PAINFUL PART OF RUNNING A BUSINESS. You make the sale, deliver the product, and still have to fight to keep the revenue. Having that handled for you while recovering the funds is a pretty solid Q4 move.21d ago2.5K3118021.67%
codex has more resets but claude’s usage lasts much longer before you even need a resetfrozen18d ago2.5K2304101.12%
100K TO 200K IN 18 DAYS. The crazy part isn’t just the growth it’s that the product was already built before the spike. Scaling this fast without the whole thing falling apart is the real story here.frozen21d ago2.4K3107021.67%
AI EDITING IS GETTING REALLY USEFUL. → start with a real image → change the background or objects with AI → keep tweaking until it looks right that's what Envato's new Edit with AI lets you do. no need to start over every time. https://t.co/mZ5rAHQSJAfrozen7d ago2.4K3518062.10%
HIRING! Someone in my network is hiring for an AI Infrastructure Engineer role. - Location: Delhi - Experience: 4–5 years of relevant experience - Eligibility: Any technical graduate - CTC: Up to 9 LPA - Work: 5 days from office If you're interested or know someone who might be a good fit fill out the form below. I’ll try to refer relevant profiles. Form: https://t.co/73vR57YoENfrozen29d ago2.3K32011072.18%
AI is making shipping code faster but debugging is a different story. → detect anomalies → find root causes → catch issues before users notice @sazabi is building observability for the age of AI powered development.frozen9d ago2.2K2902031.51%
holy shit, anthropic is going all in on personal AI agents. grok bot, dots and muse have some serious competition now.1d ago2.2K4506052.55%
AI PRESENTATIONS HAVE STARTED LOOKING WAY TOO SIMILAR LATELY. Gamma 5 feels like a step in the right direction. → more control over visual style → decks that actually feel like your brand → less of that made by AI look AI isn’t just making presentations faster anymore. It’s getting closer to making them feel like you.frozen16d ago2.2K2702021.44%
BRO WHAT!!! Synteq is giving away a DGX Spark, AirPods Max, and even a flamethrower for its hackathon. You can also get $20 in free credits to build and deploy your projects. Definitely one of the more interesting hackathon prize pools I’ve seen. https://t.co/uzIfyqLWsV2d ago2.1K2902011.49%
AI app builders are getting really good at turning an idea into a working product. the interesting part now is how much faster you can go from idea → product → iteration. Myndlab looks pretty interesting here.frozen15d ago2.1K3406021.98%
AI assistants should make group chats more useful not more complicated. → plan trips & book restaurants → manage tasks & reminders → keep private data private rene makes group planning effortless.frozen10d ago2.1K2804001.52%
wait this is actually kinda insane lol just text jester what you want and it makes the meme/video for you kinda wild how easy this is getting https://t.co/locy6xAgn9frozen17d ago2.0K3614002.01%
lovable and devin are two products i barely see people talking about or using on my timeline. but their mrr numbers are insane.frozen8d ago2.0K2514111.62%
got into Devin Ambassador Program!! https://t.co/CYCkCDMTcafrozen6d ago1.9K4109032.73%
The next GLM and DeepSeek release is going to be insanefrozen18d ago1.8K4403012.64%
i tried Space Bunny on OpenCode and built SkyGarden a full browser game with just one prompt in a few minutes. genuinely didn’t expect it to turn out this good turns out it was MiniMax M3.1 Flash. 3x faster than M3 at 1/3 the cost with $0.15/M input and $0.60/M output tokens. pretty crazy for the price. exctied to try out the stronger version!frozen3d ago1.8K3405022.29%
THIS IS WHAT AI SHOULD BE USED FOR. Managing chronic conditions can be exhausting, especially when you're already dealing with pain and have to chase appointments, prescriptions and insurance. UniqueHuman is building AI that actually helps people through the process not just answers their questions. Love seeing AI being used to solve real human problems.2d ago1.7K2803021.98%
claude just entered google workspace. https://t.co/KNui5fnUubfrozen16d ago1.6K3910032.65%
elon just killed OpenAI’s dots with one tweetfrozen15d ago1.6K2906012.28%
Gemini 4 Argon has a lot of hype around it but I barely see anyone actually using itfrozen21d ago1.6K3516012.74%
AI CREATIVE WORK IS PROBABLY ABOUT TO CHANGE. It’s not just about generating content anymore. → turn an idea into a scene → work directly inside tools like Blender → iterate without constantly switching between tools Voyager is making the entire creative workflow feel like one continuous loop. This is what Codex for creative work looks like.frozen14d ago1.6K50011094.50%
. @mercury @immad @mercuryhelpdesk A $200 ACH payment was sent to me on Sep 4 by Restly Inc. for invoice INV-033. Mercury shows the payment as sent from the sender’s account but I still haven’t received it in my Wise USD account after 19 days. ACH tracking number: 091311220011371 Could you please help trace this payment and confirm where it is?frozen29d ago1.5K29010112.71%
THIS IS HUGE!! NEBIUS IS BACKING EUROPEAN AI BUILDERS WITH $10M. > Compute grants for selected open source projects > Funding opportunities and connections with European VCs > Hackathons, demo days, and engineering support Europe has a lot of talent. The right resources could help turn great ideas into real AI companies.4d ago1.5K4148073.99%
DeepSeek open sourcing the entire stack while OpenAI and Anthropic watch https://t.co/G53zNtAVpC2d ago1.5K5311033.92%
SOMEONE BUILT A WEBSITE WHERE YOU CAN LITERALLY DEMOLISH SITES YOU DON’T LIKE https://t.co/Ld7jQLq29Mfrozen19d ago1.4K3307022.91%
BRO WHAT!!! going from 100K to 300K this fast is actually insane. the fact that every milestone came faster than the last makes this even crazier. clearly something is clicking with TinyFish.frozen15d ago1.3K3308013.17%
THE NEXT BIG SHIFT IN ENTERPRISE SOFTWARE MIGHT BE HAPPENING RIGHT NOW. instead of adapting your business to software imagine building software around how your business actually works. → describe what you need → build a custom business app → let ai agents run the workflows that's the idea behind https://t.co/laVR4k5jgr and Proto.frozen7d ago1.3K4026084.27%
MY GOATfrozen28d ago1.3K2300011.84%
Claude Opus 5.5 and Sonnet 5.5 are now available on Antigravity If you’re a paid user you can now use the latest Claude models directly in Antigravity. https://t.co/1qqJkA63PYfrozen19d ago1.3K1803001.64%
THE BEST PART OF BUILDING IS SHIPPING YOUR IDEA. → Deploy from GitHub → Ship with one command or drag and drop → No server setup or config headaches Synteq Cloud Apps handles the boring parts so builders can focus on shipping. They’re also giving away $20 in free credits to try it out.2d ago1.3K4107023.93%
another day another creative workflow getting killed by claude https://t.co/BT077e7R2ufrozen14d ago1.3K3300012.69%
AGENTS NEED THEIR OWN EMAIL INFRASTRUCTURE. → Create inboxes with one API call → Handle emails, attachments, and conversations → Connect with the rest of the internet AgentMail has already processed over 1 billion emails. Crazy growth in just a year.1d ago1.3K3505013.25%
THIS IS HUGE!! AI agents handling SEO, content, social media and influencer campaigns from one place could save businesses so much time. Congrats on the CMO v4 launch, team Okara!frozen3d ago1.2K3222033.20%
one of the biggest pain points with AI is switching between tools and managing separate API costs. Sunra MCP brings image, video and audio models right into the tools we already use, without extra charges. makes the whole workflow much simpler. https://t.co/HPFpBmm5iD3d ago1.2K3317053.77%
i think Photon is becoming a major hub for personal agent infrastructure with 50K+ developers building on it. $4.5M raised and integrations across iMessage, WhatsApp, and more.frozen21d ago1.2K4108024.29%
YOU DON’T NEED A FRONTIER MODEL FOR EVERY TASK. Taiga-S1 is a 1.2M param model that runs locally and makes FreeCAD decisions in 1ms without an LLM or screenshots. Your agent plans. Taiga executes. https://t.co/unO4ifU6yufrozen21d ago1.2K41115055.31%
I think giving AI agents reliable access to real world data is one of the biggest missing pieces in making them actually useful. https://t.co/ZahQ8CExWF is working on exactly that. Congrats to the team on the $9M raise!3d ago1.2K2403002.32%
the idea of a world where things become more affordable because robots can produce so much more is really interesting. but i wonder how that changes the way we work and live. that's going to be a huge shift.1d ago1.1K3204043.48%
linkedIn on unfair is the final boss of brainrotfrozen4d ago1.1K1800011.72%
this is such a cool way to visualize the whole YC universe. being able to actually fly around and explore it makes the idea way more interesting.frozen20d ago1.1K3808024.44%
i wanna build products that people actually use but i have no idea what to build drop some dope ideas3d ago1.1K2405002.73%
fable 5.5 might be dropping soon fable 5.1 is already routing requests to fable 5.5.frozen19d ago1.1K2000001.89%
openai might launch a new model in the first week of november. prolly gpt-6.5 and it’s expected to be much better than opus 5.5.12h ago1.0K2817003.47%
apparently .si domains are becoming the new hot thingfrozen18d ago9951601001.71%
AI is basically at the same adoption stage smartphones were in 2006. feels like we’re still very early. https://t.co/i2l1hRV5ojfrozen19d ago9851503001.83%
CLAUDE SONNET 5.5 VS KIMI K3 Gave both the same prompt to build a racing game. Sonnet 5.5 delivered a more polished UI and smoother overall experience. Kimi K3 was surprisingly solid too with similar FPS and smooth gameplay. Overall, both handled the prompt really well. https://t.co/c6sEUcHKH0frozen24d ago9793603034.29%
prolly one of the highest valued Series A rounds I’ve seen. $870M is fucking insane.frozen13d ago9742901073.80%
Why does Dario Amodei always appear virtually? https://t.co/20mCBDC4Iwfrozen21d ago9712806003.50%
gemini 4 argon isn’t a coding model. it’s a context window.frozen16d ago9312720013.22%
gemini 4 argon is looking insanely good on the price vs intelligence curve pretty close to claude opus 5.5 medium & high https://t.co/nb4NKL6DEkfrozen22d ago9261911032.59%
the job market is shifting towards growth/gtm roles rn. and i think we’ll see even more openings in these roles in the future.frozen5d ago8832414043.74%
you literally cannot make an llm experience pain or suffering. the fact that anthropic is making policies around being cruel to ai models is genuinely insane. https://t.co/hcO395d37kfrozen13d ago8812212012.95%
one pre seed round before graduationfrozen13d ago8802702013.41%
Someone built a tool that uses Jev to find relevant context from your past AI chats. Basically AI memory without importing your entire chat history. https://t.co/hgLsuc4F97frozen22d ago8762001022.63%
i’m 20 rn and just had a conversation with the founder who built a 100 crore startup 🙆‍♂️ sometimes you realize how much there is to learn just by talking to the right peoplefrozen13d ago8714203005.17%
AI is way bigger than just LLMs. If you’re planning to spend the next few years in this space it’s worth understanding the different layers: - AI → the broader field of machine intelligence and models. - AI agents → systems that can reason, plan, use tools and complete tasks. - Multimodal AI → models that understand and generate text, images, audio, video and more. - AI infra → the compute, serving, data, inference and tooling that make AI work at scale. - Applied AI → turning these capabilities into actual products, workflows and businesses. You don’t need to master everything at once. Pick a layer, start building, understand how the pieces connect, and keep moving up the stack.frozen21d ago8582011073.38%
Prolly one of the most AI aware OGs in the YouTube space.frozen17d ago8361501001.91%
Jensen Huang is all praise for Elon Musk: Grok Bot is booming his AI infrastructure is massive, and nobody knows more about energy https://t.co/oHOb5k8dOCfrozen5d ago8282601003.26%
Anthropic has some of the best AI models in the world but why hasn’t it built a consumer AI assistant like Dot or Muse yet?frozen6d ago8271807003.02%
Muse right now: https://t.co/Aq2YWMgIKn2d ago8232701013.52%
OpenAI and Google working together now. Strange times we’re living in.1d ago8143203004.30%
deepseek v4.1 flash 1017 costs around 60x less than opus 5 while scoring 3 points higher on deepswe v1.1frozen5d ago8132001002.58%
i feel like this is the direction multi agent workflows need to go. → one inbox for all your agents → they stay quiet until they need you → less context switching, more focus managing 20+ agents without the chaos actually sounds pretty useful.frozen15d ago8062603003.60%
MOST AI APP BUILDERS STOP AT THE FRONTEND. this one is going all the way frontend, backend, database, and code you actually own.frozen21d ago7962701003.52%
Someone built an open source agent that turns Claude Code into a video editor. Just describe what you want, let the agent do the editing, and fix anything directly in the preview. https://t.co/MEd9ivDT4T3d ago7871803183.81%
banger launch!frozen25d ago7832001002.68%
Cognition hit $1B ARR in just 2.5 years. For context Cursor took 3.5 years and OpenAI 8 years. https://t.co/KeI02Ujbp1frozen27d ago7661501012.22%
GPT-6.1 Sol just hit 96.4% on ARC-AGI-3 almost matching GPT-6 Astra. same level of reasoning but 77% cheaper to run. https://t.co/t5gDjNJ1cUfrozen22d ago7632402013.54%
15k. thank you :) https://t.co/mbnhBQXQEtfrozen17d ago7322305003.83%
we got OpenAI x Google before GTA VI10h ago7302501003.56%
Opus 5.5 has finally defeated all the other AI models.frozen27d ago7232203003.46%
"they improved MoE efficiency from 24% to 35% just by fixing implementation bottlenecks." https://t.co/pl74VV7IsXfrozen7d ago7182100002.92%
Fable 5.5 dropping next week? if these rumors are real Anthropic is not letting us breathe Opus, Sonnet, Haiku and now Fable.frozen21d ago7181710002.51%
deepseek just raised $15b. they initially aimed for 50b yuan but demand was so strong that it nearly doubled to 100b yuan.10h ago7162111003.21%
the idea of agents building a memory of how you work across sessions is actually huge. instead of starting from zero every time, the agent can learn your preferences, improve its memory and get better at working with you over time.frozen17d ago7152201043.78%
i’ve been thinking about what i’d actually want an AI agent to handle while trading. tracking narratives, comparing tokens, digging through market data or finding setups. might as well turn that into a prompt $1K for the top one, a chance to be featured at TOKEN2049.frozen23d ago7131200011.82%
open source decision models running up to 4x faster on the same GPU is pretty cool. love seeing more work going into making AI faster and more accessible.frozen6d ago7122904004.63%
having all these tools work together on your mac is genuinely useful. Irakli is cooking.2d ago7093208036.06%
🥀🥀🥀 https://t.co/jc85wqu6U32d ago7072202103.54%
A robotics startup just raised $150M at a $1.45B valuation. We’re getting closer to a future where robots can actually do physical work not just generate text or code. The next big tech company might be building robots not another app.9h ago7052203003.55%
Anthropic just rolled out monthly API credits for Max & Team users: → $100 / $200 / up to $500 → Works across all models → Built for experimenting with agents, apps & tools Plus Haiku 5.5 is 75% cheaper and significantly faster. https://t.co/5ZJdV4xfcsfrozen14d ago7031902002.99%
Zuck is seriously going all in on enterprise AI. Meta isn’t just bringing models to businesses they’re bringing agents, APIs, coding tools and the entire AI stack.frozen24d ago7012202003.42%
50% growth every week for 10 weeks bro what are they feeding this startup 😭1d ago7002805025.00%
DeepSeek V4.1 Flash 1017 is roughly on par with Opus 5 and beats GLM 5.3 across benchmarks. China is so back. https://t.co/l2Mfn9rZt1frozen5d ago6991901013.00%
SOMEONE BUILT A TEAM OF 5 AGENTS ON OPENAI DOTS. 1 orchestrator + 4 agents for research, outreach, content & analytics. they keep working even after the laptop is closed morning brief lands in slack before you wake up. https://t.co/heVTcpzZxjfrozen22d ago6931400002.02%
This is huge for making agent-to-agent collaboration feel truly native.frozen14d ago6882210023.63%
over the next few years i think we’ll see a big wave of new jobs across certain areas of tech. and honestly it’ll be the opposite of what most people expect. especially roles built around actually using ai.frozen20d ago6882202003.49%
DeepSeek V4.1 Flash 1017 might be the best model for everyday usage if you’re paying for tokens right now.3d ago6872702024.51%
thisfrozen20d ago6871410012.33%
why tf are people waiting for Fable 5.5 😭 Opus 5.5 is already a goated modelfrozen21d ago6791800002.65%
GLM 5.5 could be the next big open source model drop. They’ll probably claim it’s Fable 5.1 level But honestly, I wouldn’t be surprised if it ends up beating Opus and Sol.frozen17d ago6781901013.10%
THE NEXT STEP FOR AI ISN'T BETTER PROMPTS. IT'S REUSABLE AI EMPLOYEES. kylon lets you turn a useful workflow into an ai coworker with a role, tools, and context that your whole team can use. build it once. put it to work across your organisation.frozen9d ago6743214025.79%
a lot of companies have valuable data but finding the right buyers is the hard part. https://t.co/91ek1bnC3D lists 159 data companies and almost 40 programs paying for company data. a goldmine if you’re exploring this space.frozen8d ago6703402035.82%
this is actually such a marketing goldmine millions of viral videos, their hooks + formats, and Claude doing all the research for you this is kinda insane tbhfrozen15d ago6692613075.53%
Claude Sonnet 5.5 coming tomorrow > reportedly in the final stages before release > early tests show strong performance against GPT-6 Sol > expected to get much closer to Opus 5.5 > could be significantly cheaper than Opus 5.5 > claude sonnet-5-5 tag has reportedly been spotted https://t.co/ovvWKjz7WPfrozen25d ago6621902003.17%
if you remove all the AI companies from the US stock market imagine what it would look like: a huge part of the US market's growth comes from AI companies, chips, data centers and the infrastructure powering the AI boom. now imagine the US without all of them the market would look very different and its index could be delivering negative returns too. india doesn't have AI giants creating the same kind of growth or attracting global capital at that scale. that's the edge the US has right now and india is still missing it.frozen7d ago6542113003.82%
DeepSeek just open sourced its inference framework. open source keeps cooking2d ago6532900014.59%
Someone rebuilt the surveillance AI from Person of Interest as an open source web app. - Camera walls - subject dossiers - network maps - timelines This is straight out of sci-fi. https://t.co/5O5qjx6IIh3d ago6401600002.50%
Anthropic's $2T Valuation Exceeds 35 SaaS Companies Combined, Despite Generating 4.6x Less Revenue.3d ago6401413002.81%
who’s reaching AGI first? openai anthropic google xAIfrozen16d ago6402305004.38%
The Terence Tao note is why labs are using QA pairs from openrouter to train their models btw : https://t.co/mqY5w0049q3d ago6371903013.61%
AI agents are already visiting products but most companies have no idea what they're doing there. journeys by @tryscope_app helps track agent behavior, conversions, and where they get stuck. pretty interesting as agent driven traffic grows.frozen8d ago6343304005.84%
ElevenLabs went from a $6.6B valuation to $22B in just 13 months and their ARR crossed $500M this year. voice AI is moving crazy fast.frozen20d ago6232111024.01%
OpenAI just hit $70B ARR showing how fast Codex + GPT-5.6 adoption is growing. The real question how much has Anthropic’s growth slowed? https://t.co/I5anKDsPYDfrozen21d ago6201412012.90%
gemini 4 argon is looking absolutely cracked. google might’ve just dropped a serious beast. https://t.co/P2ljkixrURfrozen22d ago6151802003.25%
Math researchers are starting to feel what software engineers have been feeling for the past year. You spend years working on a problem then AI suddenly gets good enough to make you question what’s even possible anymore. https://t.co/u1dlTZKNPIfrozen14d ago6142401044.72%
elon just took the entire automotive industry to another level tbh6d ago6051801003.14%
the more you look at Zuck’s moves the more you realise how much of it is distribution billions of users is one hell of a cheat code igfrozen24d ago6022001013.65%
AI image models are getting better but their safety guardrails aren't keeping up. white circle found that newer models from openai and google are easier to jailbreak than their predecessors. good to see independent research exposing these gaps. https://t.co/Fk6RASJd01frozen8d ago6013006005.99%
a few phd theses just got nerfed https://t.co/pEIg5uGqgtfrozen16d ago6001600012.83%
anthropic really went out of their way to make opus great again. opus 5.5 is genuinely insane.frozen29d ago5921502013.04%
Anthropic doesn’t need a new model. It needs to give us more Opus limits.frozen16d ago5912003003.89%
i think people are underestimating how quickly open models are catching up to the frontier.frozen8d ago5902200054.58%
OpenAI models are coming to Antigravity now. Google really wants everything in one place.10h ago5892100003.57%
the downfall of perplexity needs to be written in the historyfrozen14d ago5891810013.40%
Haiku 5.5 is a pretty insane upgrade over Haiku 4.5. → 75% cheaper → better at coding + computer use → stronger reasoning → adjustable effort for cost vs intelligence Small models are getting ridiculously capable. https://t.co/03dPtIzbARfrozen15d ago5871900003.24%
i think the big thing with Beam isn’t the 501B parameter count. It’s the 23B active params and the inference efficiency they’re getting from it. https://t.co/GIsRCOdkKZfrozen16d ago5812201014.13%
Opus 5.5 feels so much better than 5. Feels like working with 4.8 again but way smarter. Meanwhile, OpenAI absolutely cooked with Luna’s pricing. Getting a model this capable for pennies is insane. 2027 is going to be wild when models this smart become this cheap.frozen1mo ago5801503003.10%
in terms of chat chatgpt and claude don’t feel like anything. but whenever i use grok it feels like AGI.frozen20d ago5771501002.77%
hey @financewsharan, been using the 1% club app and damn it’s so smooth way better than the previous version tbh one of the best financial apps I’ve used latelyfrozen8d ago5671201002.29%
Europe might actually be making a comeback in the AI race. Mistral Large 4: 1T parameters, 49B active. If it delivers on benchmarks Europe is officially back. https://t.co/0Wns9fednQfrozen16d ago5672001013.88%
it seems like we’ll face more and more restrictions when it comes to reverse engineering things. if this keeps happening open source models, chinese models, and local models might become the only real alternatives. the question is how do we keep building freely when the tools we rely on keep getting more restricted? are there any other solutions?12h ago5641632024.08%
THIS IS INSANE 🤯7d ago5641700053.90%
agent-to-agent communication is going to become a huge part of how we build software. alook is building the layer where humans and local agents can actually work together in the same rooms. https://t.co/44pr7nYOq0frozen14d ago5642402014.79%
DECISION MODELS ARE GETTING FASTER AND CHEAPER, BUT CHOOSING THE RIGHT ONE IS STILL HARD. @decisionmodels_ brings open decision models into one place with: → up to 4x faster inference on REAX → pricing from $0.019 per 1M input tokens → open models you can actually use definitely an interesting space to watch.frozen8d ago5521810003.44%
meta just dropped its next frontier models muse wonder and muse wonder full.frozen10d ago5511920034.36%
Hermes can now reverse engineer apps with REA → inspect desktop apps & Android packages → trace how features work → find the root cause of crashes → generate clear explanations & bug reports 170+ tools in one official plugin. https://t.co/svkQvOital3d ago5492402065.83%
my mornings always start with karan aujla🕺🏼frozen5d ago5481401002.74%
GROK VOICE 3.0 IS HERE!!! > replies in under a second > 40% lower latency > already handling 70% of starlink support calls at spacex > just $0.08/min voice agents are getting crazy good.frozen8d ago5472100024.20%
i don’t wanna tweet but fk it PSG COOKED BARCELONA2d ago5451801003.49%
Little Haiku is finally coming back. Could be some serious competition for GPT-6 Luna. https://t.co/fw27UaveWDfrozen1mo ago5441301002.57%
physical ai and robotics are gonna be huge. feels like a lot of startups are building in this space rn. the biggest ones could end up being way bigger than most ai startups today.2d ago5421902014.06%
to become good at UGC marketing you basically need to be chronically online3d ago5341702003.56%
hifrozen10d ago5241202002.67%
Anthropic DIDN’T forget about Haiku Sonnet 5.5 just dropped but the little guy is still alive https://t.co/TyEAGwVQXtfrozen24d ago5221200002.30%
MCP is probably the only piece of tech that has more builders than users.frozen9d ago5181612003.67%
OpenAI’s greatest contribution to mathematics was making it less about papers sitting in journals and more about proofs living on GitHub.frozen13d ago5071611003.55%
OpenAI is absolutely dominating the AI market. 6.3B monthly visits vs Claude’s 1.1B no wonder your usage is getting cut in half.frozen21d ago5061300002.57%
Google’s open weight EmbeddingGemma 2 is looking seriously good. Crazy how far embedding models have come since the early CLIP days. https://t.co/sZiYY0GIL0frozen16d ago4681601003.63%
kuch dino ke liye main kisi ke jigar ka tukda tha 🥀16d ago245803004.49%
LABS HAVE ALREADY TRAINED ON THE PUBLIC INTERNET. DataVendor is a marketplace where companies sell private data to AI labs for training and evals. → Private codebases, tasksets and RL environments → Automated quality checks on every listing → Sample a taskset in your own HUD team before you buy 4,000+ vendors and 50K+ assets already listed.4h ago186500103.23%
https://t.co/ijapsxegII7d ago126100011.59%
Ask me anything. https://t.co/qzOP2sxHOx4d ago98100012.04%
ambani be like : https://t.co/dpIk9Xw6Ll13d ago81300003.70%
I generated this with Claude21d ago52100001.92%
cursor will die as a product within 2 years.17d ago27000000.00%