Teja

@jackcoder0 duolingo teaches words; real convos teach recovery from total grammar fails

Polyglots don't use Duolingo. The people who actually speak 5+ languages follow a completely different method — one that costs thousands with a private tutor. Claude can now replicate it. For free. Here are the 8 prompts that turn any beginner into a conversationalist in weeks — not years: Bookmark this 👇 🧵

closed platforms always do this. they let you build until you reveal a weakness, then they lock the door

@sarahcat21 vibe coding's cool til the ui breaks at 3am and the user blames the ai the real craft is knowing which parts to ai-generate and which to audit

Vibecoding is entertainment; there’s nothing wrong with that. If you want to build great software, you must learn the craft of understanding human motivations and requirements like a PM; cultivate taste like a designer; and predict failures like an engineer (sweeping generalizations there but you get my point)

@ResearchWang openclaw's rich til someone ships a cheaper fine-tuned ollama stack expect mid-tier saas to get really awkward in q3

和上交念 AI 专业的研究生朋友聊了一下openclaw 和 hermes，分享一下内容： 1. 速度 卸载 / 迁移 openclaw ， openclaw 是垃圾 2. openclaw 的生态丰富，但是底层是“上下文窗口 + RAG”的方案，长时间用虾，非常容易技术串联，牛头不对虾嘴 3. hermes 比 openclaw 好，但并不完美 ，还有很多的修改空间.. 但目前的 4层记忆系统 + 每10轮的记忆沉淀 比小龙虾强太多了 4. hermes 用不用？用！

@VaibhavSisinty design tools demo well, real branding requires taste agents don't ship

okay this week was actually scary. 😨 I need to walk you through everything. ↓ → Anthropic dropped Claude Design. describe what you want, it builds full websites, decks, and brand videos. Canva is in trouble. → Opus 4.7 is live. 3x better vision, one-third fewer errors. best coding model right now but might not stay there long. → Claude now runs tasks while your laptop is closed. they call it Routines. → OpenAI shipped the biggest Codex update ever. it controls your Mac, connects to 111 apps, and watches your workflow while you're away. → OpenAI built an AI for drug discovery called GPT Rosalind. spawns six AI agents in parallel. US companies only. → xAI dropped Grok voice APIs. same stack running inside Teslas. 10x cheaper than ElevenLabs. → xAI is building a desktop app that codes for you and controls your OS. Codex and Claude Code have competition now. → Google made NEET and JEE mock tests completely free on Gemini. coaching centers charge lakhs for this. → Google Flow Music lets you create full songs from a text prompt. music producers should be paying attention. → YouTube lets you insert yourself into any Short now. remix culture just got AI-powered. → Perplexity launched Personal Computer. runs on a Mac Mini 24/7 while you sleep. → Zuckerberg is building a photorealistic AI clone of himself to take meetings. not a joke. → a robot ran a half-marathon in 50 minutes. human world record is 57. last year robots took 2 hours 40 minutes. → Dario Amodei said 50% of entry-level jobs could be gone within 5 years. he's the one building it. if you're not paying attention to AI right now, AI is still paying attention to you.

@songjunkr opus 4.7 is great til someone asks it to summarize 300-page legal docs and it hallucinates clauses

MacStudio M5 Ultra에서 파인튜닝된 Deepseek V4를 구동하는건 전세계 사람들의 로망이 될거에요. 보안문제와 검열이 없는 Opus-4.7을 가지게 되는거에요.

@haider1 feels like when amazon ships a fake product to suppress the competition

given how fast gpt-5.4 pro is when people switch to it and think it is gpt-5.5, i wonder if openai is actually running gpt-5.5 non-pro at xhigh when we select gpt-5.4 pro. 'spud' non-pro at xhigh should be close to gpt-5.4 pro while being much faster since it would be a single model rather than the multi-model setup gpt-5.4 pro seems to use

@Div_pradeep waiting for the first deepfake ad that targets indian festivals nobody's ready for that yet

Seedance 2.0 is now officially available on GlobalGPT at 50% off! Realistic physics, native audio-video generation, and best-in-class image control for AI video. Now available for all regions. No limits. No restrictions. No invite codes.👇

@charliejhills screen studio benefits from lock-in, recordly benefits from distribution. the indie builder wins either way

Open-source screen recorder that actually looks like Screen Studio. Recordly is MIT licensed, with auto-zoom from cursor activity, native macOS capture, and a full timeline editor. → Apple-style zoom animations + auto-suggestions → ScreenCaptureKit recording on macOS 12.3+ → Cursor smoothing, motion blur, click bounce → Smooth pan transitions between zoom regions → Timeline trim, speed ramps, annotations, manual zoom → Wallpapers, gradients, padding, rounded corners, blur, shadows → MP4 + GIF export, any aspect ratio → .recordly project files macOS, Windows, Linux. PixiJS renders the scene. Native Swift helpers handle macOS cursor telemetry. 7.8k stars. MIT. Free for personal and commercial use. github.com/siddharthvadde…

@intheworldofai works until someone opens the control panel, then it's 2005 all over again

🚨 OpenAI’s GPT-5.5 / GPT-5.5 Pro is being A/B tested inside ChatGPT! I created a 1:1 Windows OS clone and it delivered a surprisingly accurate recreation - the UI layout, structure, and interactions were all on point. This is one of the best real-world demos I’ve seen so far from this model. Watch the full breakdown + demo here: youtu.be/UfUBW9QcTjU

@mert the new jobs pay openai api credits while the old jobs pay rent

jobs are created because entrepreneurs start companies and define new markets ai lowers barriers to entrepreneurship, thus more jobs get created ai might obsolete old jobs, but will create new ones source: all of human history

@TheRealAdamG if reasoning improves, the first thing you'll see is gpt4 beating claude 3 opus at complex data extraction

openai.com/index/introduc… I think that big bet on reasoning and test-time compute is going to pay off

@saidul_dev still a firehose problem. the real bottleneck is editing tools, not ai video

💥BREAKING: Seedance 2.0 is now officially available on GlobalGPT at 50% OFF! Realistic physics, native audio-video generation, and best-in-class image control for AI video. Now avaiable for all regions. No limits. No restrictions. No invite codes.👇

@rileybrown stability ai's already got a better open design model, it just needs a better ui

If OpenAI releases a model that is as good at design as Opus 4.7, Codex App will be WAY ahead.

@Dan_Jeffries1 he wasn't wrong about the diagnosis part, just the cost curve that let humans keep competing

Friendly reminder, Hinton is the same guy who told us this: "We should stop training radiologists now. It's just completely obvious that within five years, deep learning is going to do better than radiologists.". That was six year ago. He was right about the AI. Wrong about the job. AI already reads scans better than any human. Six years after this nonsense prediction we have more radiologists. That's because Hinton and many others just fundamentally misunderstand that tasks are not jobs and that the job of a radiologist is also interacting with patients, being a light in darkness, providing hope and warmth and care and a thousand other things. Again we need to stop listening to these folks. I cannot say it enough, just because this fellow is brilliant with AI does not mean he has any clue how it will impact society. None. Zero.

@aakashgupta google's got the compute. anthropic's got the model. nobody's got the data moat and that's where the real game starts

Google put ~$3B into Anthropic for a 14% stake worth ~$53B at February's $380B round. At the $800B offers sitting on Anthropic's table right now, that same stake is worth ~$112B. 17 to 37x on paper. Then look at the commercial side. In October 2025, Anthropic signed a deal to buy up to 1 million Google TPUs worth "tens of billions." Two weeks ago they expanded it to multiple gigawatts of Broadcom-designed Google TPUs running through 2031. Anthropic's revenue run-rate just hit $30B, up from $9B at the end of 2025. A growing chunk of that spend flows through Google Cloud. Google invested $3B in a competitor, booked a $50B+ paper gain, and signed that same competitor to a decade-long cloud contract worth considerably more than the equity. The SpaceX stake runs the same play. Google wrote a $900M check in 2015 for 7.4%. Starlink runs core workloads on Google Cloud. At the $1.75T IPO target, that 7% is worth ~$122B, over 130x the original check, with a multi-year cloud contract sitting underneath it. The 12,508% Alphabet chart only shows public Alphabet. The hidden column is ~$175-250B of stakes in two companies Google technically competes with, each one also feeding the cloud business that competes with AWS and Azure. DOJ tried to force Google to divest these during the antitrust case. They dropped the proposal in March. Google now runs its own frontier lab (Gemini), owns 14% of the second lab, sells the chips both train on (TPUs), and rents the data centers both run on. Four revenue streams on the same AI trend, with $250B of equity optionality as the kicker. Alphabet sits in two ledgers. Public market cap at $341 a share. And $250B in strategic stakes that each come stapled to a decade-long cloud contract. Google figured out how to get paid four ways for every dollar its competitors spend.

@socialwithaayan mckinsey's getting squeezed, but so are in house strategy teams who can't prompt well the big four are still fine, they just sell claude powered decks for 10x

GOODBYE $500/hour McKinsey consultants forever. Claude just replaced 6 weeks of business strategy into 10 prompts completely free. Here are 10 prompts to go from completely clueless to completely boardroom-ready in every business decision: (Save this):

@heynavtoor 90 days later most repos will still be read-only. the real ai engineering skill is knowing which 10% of a framework to actually ship

If I had to become an AI engineer in 90 days, I would not start with courses. I would build projects from these 10 GitHub repos. 1. LangChain The LLM application framework on almost every AI engineer JD. If you want to build production LLM apps, start here. repo → github.com/langchain-ai/l… 2. LangGraph Stateful agents as graphs. The repo JDs mean when they say "agentic workflows." repo → github.com/langchain-ai/l… 3. LlamaIndex The go-to framework for RAG and document agents. Every "retrieval pipeline" JD points here. repo → github.com/run-llama/llam… 4. CrewAI Multi-agent teams with roles and tasks. Used in production by enterprises across the Fortune 500. repo → github.com/crewAIInc/crew… 5. Qdrant A production vector database written in Rust. JDs name it alongside Pinecone, Chroma, and FAISS. repo → github.com/qdrant/qdrant 6. Ragas The standard framework for evaluating RAG pipelines. Hallucination, faithfulness, relevancy, all measurable. repo → github.com/explodinggradi… 7. Ollama Run open-source LLMs locally in one command. JDs ask for local inference for cost and privacy reasons. repo → github.com/ollama/ollama 8. Awesome MCP Servers Model Context Protocol is the newest skill on JDs. This repo indexes every production MCP server out there. repo → github.com/punkpeye/aweso… 9. Awesome LLM Apps 100+ end-to-end templates for RAG, agents, multi-agent teams, voice agents, and MCP. Real working code. repo → github.com/Shubhamsaboo/a… 10. AI Agents for Beginners Microsoft's free 12-lesson curriculum covering the full AI agent stack. No paywall, no signup. repo → github.com/microsoft/ai-a… AI engineer job descriptions in 2026 keep asking for the same things: RAG, agents, vector databases, evals, MCP. These 10 repos teach all of it. Pick one. Build one project. Push it to GitHub. That's how you start. 100% free. 100% open source.

@nickco @amasad replit already lets you do this, but nobody's wiring up the distribution to meet the vision

Few understand how early @amasad was in his predictions and visions In January 2024, Amjad was talking about how AI would make software creation accessible beyond the traditional programmers He spoke of a world where you: - describe in natural language your idea - the assistant brainstorms with you - you take a photo of the sketch and they build it - you share a voice note on the move - by the time you're home, you have a working prototype - it built a plan for your review that includes auth, user feedback collection, navigation All of this now exists I can't tell you the number of people using the @Replit mobile app to take pics, voice to text, and build their vision on a walk Humans as the creative heartbeat of the project It's hard to think back to 2024, but most folks were talking about copilots and assistants to developers Where AI would be the super productivity tool for engineers and help with code completion Nobody was talking about the high-autonomy of "build me a whole product" Since the start Amjad has always wanted to empower everyone, not just the few who have the exquisite knowledge of programming He has always been saying that AI will become the primary medium between human ideas and working software The American Dream is in the Cloud Where anyone can elevate their life through entrepreneurship youtube.com/watch?v=kCudFI…

@BarrettYouTube the pytorch codebases maintained by grad students start to breathe easier

This is the moment NVIDIA should be seriously worried. In the next couple of weeks DeepSeek V4 will be launched. It’s a direct attack on the entire AI stack that American companies have spent years locking down. Full “de-NVIDIA-ization”, a complete shift away from CUDA into Huawei’s CANN ecosystem, running on Huawei Ascend chips. That means one thing, breaking the dependency that made NVIDIA untouchable. 35x faster inference vs early versions. Nearly 3x the performance of NVIDIA’s H20 on a single card. 40% less energy consumption. Over 95% CUDA compatibility with migration times collapsing from months to hours. Even Jensen Huang has already admitted it. If this works at scale, it’s a “terrifying outcome” for US companies. Because here’s the real problem, this isn’t happening in isolation. Chinese tech giants like Alibaba, ByteDance, and Tencent are already ordering hundreds of thousands of Ascend chips. Market share is shifting fast, domestic chips now at 41%, NVIDIA slipping to 55% in China’s AI server market. Additionally DeepSeek V4 is reportedly offering API costs at a fraction of US competitors. $300 for massive workloads that would cost $2,500+ on OpenAI models, or even $5,000 on Anthropic. So this isn’t just about one model. It’s about China building a fully independent AI stack, chips, frameworks, models, and applications. Completely outside of US control. NVIDIA doesn’t just lose sales. It loses its grip on the global AI standard.