•twitter15877
listening to
timeline
15877
mentions
tracked
tracked
net sentiment
+24%
4998 positive
9734 neutral
1145 negative
volume & sentiment over time
Sep 30 – Oct 14, 2026 · Daily
sources
most discussed
—
active days
15
mentions
Showing 3176–3200 of 4998 mentions
-
• twitter 1w positiveA REALTOR IN SHENZHEN TURNS EVERY LISTING INTO A VIRAL SEEDANCE CLIP AND IT CLOSES HIS DEALS. HE MADE $13,400 LAST MONTH > 176,000 likes on 8,300 comments. A realtor waves at a bare garden and it fills with a sofa and lanterns, a drone loops the roof, then an empty white room turns into a playroom under a galaxy ceil… A REALTOR IN SHENZHEN TURNS EVERY LISTING INTO A VIRAL SEEDANCE CLIP AND IT CLOSES HIS DEALS. HE MADE $13,400 LAST MONTH > 176,000 likes on 8,300 comments. A realtor waves at a bare garden and it fills with a sofa and lanterns, a drone loops the roof, then an empty white room turns into a playroom under a galaxy ceiling. Physical staging runs $2,000 to $5,000 per home. So why does one clip sell the house faster? -> He waves and the lawn fills with furniture at second 5. The clip pictures the empty space for the buyer. -> He says "it doesn't matter" before the empty room changes. The line sets a challenge and the eye waits for him to win it. -> The new room keeps its window and its door frame. The brain reads it as the same room, so the buyer believes the house can look like that. -> He ends on the same front steps he started on, the detail people catch on the second watch. -> Buyers ask for the address in the comments. He drops the next listing's photos in a folder and the tour is ready 4 hours later. It works on any space people walk through before they pay, a hotel suite, an Airbnb loft, an office floor, a wedding venue. Claude writes the tour, REimagineHome stages the rooms, Make delivers it for $135 a month. The playroom got the likes, and a realtor waving at an empty lawn is what got buyers calling. Every staging prompt is in the article below 👇view source ↗
-
• twitter 1w positiveWe turned Claude Code, Codex, Hermes, Pi, @opencode and other coding harnesses into RL environments. No changes to the harnesses, no changes to the training code. Any open model, any task set, fully open source my friends! Same model, same weights: 62% under Mini-SWE-Agent, 33% under Claude Code. But training inside a… We turned Claude Code, Codex, Hermes, Pi, @opencode and other coding harnesses into RL environments. No changes to the harnesses, no changes to the training code. Any open model, any task set, fully open source my friends! Same model, same weights: 62% under Mini-SWE-Agent, 33% under Claude Code. But training inside a real harness normally means reimplementing it as an environment, so most models get trained in a scaffold nobody actually ships. The fix is a proxy, not a rewrite. The harness thinks it's talking to a model API. It's actually talking to a capture proxy that speaks the 4 formats coding agents use (OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, Gemini), forwards to @vllm_project, records the exact token IDs and logprobs vLLM sampled, and hands TRL sequences it can train on. The harness becomes the environment. 10 harnesses run through it today, none modified. And because you control the reward, you can shape behavior the harness never asked for. We added a small bonus for solving a task in fewer tool calls: on tasks it already solved, the model now uses 31% fewer calls, in every harness, and about half under Codex. Tested on LFM2.5-2.6B from @liquidai: → Train in one harness: better mostly in that harness (OpenCode 34% → 58%). → Train in 4 at once: better in all 4 (42% → 54%). → SFT on 3,189 rollouts from Qwen3.8-27B instead: plateaus at 47.5%, below both RL runs. Everything is open and reproducible: the capture proxy in OpenEnv, the trainer in TRL, the tasks, the SFT data, the training code and all 7 trained models. Bigger models and bigger runs next. Full guide: https://t.co/sKZURuOczaview source ↗
-
• twitter 1w positiveIntroducing gdp-ts: Ghosts of Departed Proofs for TypeScript. gdp-ts is a library, linter and AI skill for safer API design. Under this contract, sensitive functions require 'proofs' that the caller performed an authorization check. The typechecker verifies these proofs at compile time, preventing your team and agent… Introducing gdp-ts: Ghosts of Departed Proofs for TypeScript. gdp-ts is a library, linter and AI skill for safer API design. Under this contract, sensitive functions require 'proofs' that the caller performed an authorization check. The typechecker verifies these proofs at compile time, preventing your team and agents from shipping catastrophic security (and other kinds of) bugs. While these patterns have existed for quite some time, especially in ecosystems like Haskell, ① human code review and ② cognitive and syntactic overhead made these solutions niche. The situation is now inverted. Agents are writing more code than we can review, and they *thrive* in tight loops with hard constraints that would frustrate us. We see this with the rise of Rust, borrow checker, code aesthetics debate and all. The README and examples model a real-world Vercel API product constraint: changing the password on a Project requires a proof of a certain role + a certain entitlement. Thanks to Matt Noonan and Ollie Charles for their research in this space. https://t.co/eRWdiMvQ2Vview source ↗
-
• twitter 1w positiveMen, Money is coming. A new home is coming. A new car is coming. A better life is coming. Everything. Keep going. Men, Money is coming. A new home is coming. A new car is coming. A better life is coming. Everything. Keep going.view source ↗
-
• twitter 1w positivehere's how to save a lot of Claude Code usage don't waste your highest-limit model on everything. → use your best model mainly for planning and reasoning → delegate coding to models with better limits → keep effort low by default → only increase it when the task actually needs it → avoid agent-heavy modes unless nece… here's how to save a lot of Claude Code usage don't waste your highest-limit model on everything. → use your best model mainly for planning and reasoning → delegate coding to models with better limits → keep effort low by default → only increase it when the task actually needs it → avoid agent-heavy modes unless necessary → compact your context around 50–60% → remove duplicate or unnecessary skills → disable MCPs you aren't using → keep CLAUDE.md / MEMORY.md clean and remove outdated entries the goal isn't using less Claude Code. it's getting more work done before hitting your limits. how are you managing your Claude limits?view source ↗
-
• twitter 1w positiveMost people use 10% of AI. They ask a question and close the tab. The other 90 is Projects, memory, connectors, free setups to copy. I run my whole business on it (no staff). A short email every day, free. 9,000+ people read it. Unlock the other 90 → https://t.co/CudU2QSnDl Most people use 10% of AI. They ask a question and close the tab. The other 90 is Projects, memory, connectors, free setups to copy. I run my whole business on it (no staff). A short email every day, free. 9,000+ people read it. Unlock the other 90 → https://t.co/CudU2QSnDlview source ↗
-
• twitter 1w positiveGot this for my son a few months ago. Many of his friends have it. Great way to give a bit of autonomy while it not being a smart phone. Got this for my son a few months ago. Many of his friends have it. Great way to give a bit of autonomy while it not being a smart phone.view source ↗
-
• twitter 1w positive800 active subscriptions, yay! $1300 needed to reach my $8000 MRR goal by the end of the year https://t.co/Wv00MfspgC 800 active subscriptions, yay! $1300 needed to reach my $8000 MRR goal by the end of the year https://t.co/Wv00MfspgCview source ↗
-
• twitter 1w positiveHey @grok if you’re really the best AI, count every single person if you can 👀 https://t.co/WuvnMSirvQ Hey @grok if you’re really the best AI, count every single person if you can 👀 https://t.co/WuvnMSirvQview source ↗
-
• twitter 1w positiveWhich is the best browser rn? Which is the best browser rn?view source ↗
-
• twitter 1w positiveDrake Maye and head coach Mike Vrabel celebrate after a huge road win in the division. @insidetheNFL Week 4 Mic'd Up on X https://t.co/XSZhwULBBe Drake Maye and head coach Mike Vrabel celebrate after a huge road win in the division. @insidetheNFL Week 4 Mic'd Up on X https://t.co/XSZhwULBBeview source ↗
-
• twitter 1w positiveanthropic's cowork goes cloud-only on oct 6. so "local files" now take a trip through anthropic servers. think before connecting your sensitive stuff (unless you're fine with that) anthropic's cowork goes cloud-only on oct 6. so "local files" now take a trip through anthropic servers. think before connecting your sensitive stuff (unless you're fine with that)view source ↗
-
• twitter 1w positiveIf you think AI search means ChatGPT, look at this. People are spending WAY too much time obsessing over the wrong AI search platforms. Google AI: ~5.9B searches/day Major AI platforms combined: ~200M/day Literally 30x more (for search-like usage). Yet almost every conversation about AI search still starts with Cha… If you think AI search means ChatGPT, look at this. People are spending WAY too much time obsessing over the wrong AI search platforms. Google AI: ~5.9B searches/day Major AI platforms combined: ~200M/day Literally 30x more (for search-like usage). Yet almost every conversation about AI search still starts with ChatGPT, Claude, Perplexity… If your goal is visibility in AI answers, Google should be the FIRST place you think about. That ~5.9B/day number is AI Overviews alone. And it doesn’t even include AI Mode or Gemini, which have both now passed 1B monthly users. So before you spend all your time trying to get into ChatGPT, make sure you: - Rank in Google so LLMs can retrieve you - Win the searches that trigger AI Overviews - Optimise your pages for AI Overviews - Own your branded SERP and entity signals - Get featured in sources Google already trusts - Track AI Overviews, AI Mode and Gemini visibility And the good news is, getting this right for Google will also help you show up in other AI platforms. So yes, optimise for ChatGPT, Claude, Perplexity… But if you’re deciding where to focus, win Google first then expand from there.view source ↗
-
• twitter 1w positiveThis Video Should Be Shown To Those Girls Who Say "Aik Kachori 2 Samosa, Inn Mardo Ka Kaya Bharosa" Vidyut Jammwal Cries at Street Fighter Promotion A Man's Real Strength Is Revealed When Tears Flow From His Eyes Whether It's Out Of Love For God Or Immense Pride For His Country https://t.co/LeRt7nXN7w This Video Should Be Shown To Those Girls Who Say "Aik Kachori 2 Samosa, Inn Mardo Ka Kaya Bharosa" Vidyut Jammwal Cries at Street Fighter Promotion A Man's Real Strength Is Revealed When Tears Flow From His Eyes Whether It's Out Of Love For God Or Immense Pride For His Country https://t.co/LeRt7nXN7wview source ↗
-
• twitter 1w positive5 months after launching traxy hit $500k arr. goal is $1m by december 31. my emotional state is way too tied to this business.. when we first launched I felt invincible. the excitement and early success was addicting. then...bugs and churn. problems we didn't forecast for. spent 30 days head down trying to code wh… 5 months after launching traxy hit $500k arr. goal is $1m by december 31. my emotional state is way too tied to this business.. when we first launched I felt invincible. the excitement and early success was addicting. then...bugs and churn. problems we didn't forecast for. spent 30 days head down trying to code what should've been a simple feature. went from feeling like a boss to a child. and now, while signups and revenue explode, somewhere in between. that's probably where i belong. what's funny is i went through the exact same thing with windmill growth on our way to our first $1m. up, down, middle, up again. now if windmill loses a client, i'm fine. if we sell a new deal, i'm calm. that business never freaks me out. traxy is on its way to that state. here's what's different (and exciting): with windmill, I'm so dependent on talent. the agency has 4 incredible full-time writers - and nobody who's been promoted into a full-time role has ever left the company. for it to grow, i need to keep finding people like that. but traxy...traxy is a product. it's already built, and we're constantly improving it. and as long as we keep doing that, it's scary-easy to scale. super proud of my team for getting to $500k in almost no time. now it's time to push harder than ever.view source ↗
-
• twitter 1w positiveOver 40 judges joined us for our last hackathon, and we're now putting together the panel for the next stop of Bharat Builds. It's one of the largest student hackathons in India, with 1,400 students in the room and 16,000 building online at our last stop. As a judge, you get a front row seat to what students across th… Over 40 judges joined us for our last hackathon, and we're now putting together the panel for the next stop of Bharat Builds. It's one of the largest student hackathons in India, with 1,400 students in the room and 16,000 building online at our last stop. As a judge, you get a front row seat to what students across the country are building, and you get to give them the kind of feedback that only comes with experience. Every judge also gets exclusive Bharat Builds swag and a recognition certificate from us. If you'd like to join the panel, you can apply here: https://t.co/QxSMJBMETOview source ↗
-
• twitter 1w positivethis is like ecom in 2016 these AI characters have gotten 100M+ views in only the last couple of days making them is SO easy it feels unfair. STEP 1: find a strange character on pinterest, turn it into a character sheet with chatgpt. STEP 2: download any viral video, swap the person using Higgsfield Genjutsu. use … this is like ecom in 2016 these AI characters have gotten 100M+ views in only the last couple of days making them is SO easy it feels unfair. STEP 1: find a strange character on pinterest, turn it into a character sheet with chatgpt. STEP 2: download any viral video, swap the person using Higgsfield Genjutsu. use this prompt: "replace the person in @ video1 with the character from @ image1. keep the same motion, camera movement, color grading, lighting, framing, and overall video quality." DONE. the money part is in the full blueprint below.view source ↗
-
• twitter 1w positiveIf you want to clear an AI Engineer role, don’t just learn how to call an LLM API. Learn these topics and build projects around them: - RAG pipelines - AI agents - Agent memory - Tool calling - Vector databases - LLM evaluation - Prompt engineering - Model routing - AI observability - Guardrails & security - Streamin… If you want to clear an AI Engineer role, don’t just learn how to call an LLM API. Learn these topics and build projects around them: - RAG pipelines - AI agents - Agent memory - Tool calling - Vector databases - LLM evaluation - Prompt engineering - Model routing - AI observability - Guardrails & security - Streaming AI responses - Async job processing - Multi-agent systems - LLM cost optimization And most importantly: Build real systems, not just notebooks. A good AI Engineer should be able to explain how their system behaves when traffic, latency, failures, and costs scale.view source ↗
-
• twitter 1w positiveOfficially, every viral AI influencer was made on Higgsfield. Make one with your face or from scratch. Then jump on viral trends with Genjutsu. Try up to 5 generations free. Available now in Higgsfield and via ChatGPT Extension. Credits: @JeanPhilMadame https://t.co/GTlaBmGiQX Officially, every viral AI influencer was made on Higgsfield. Make one with your face or from scratch. Then jump on viral trends with Genjutsu. Try up to 5 generations free. Available now in Higgsfield and via ChatGPT Extension. Credits: @JeanPhilMadame https://t.co/GTlaBmGiQXview source ↗
-
• twitter 1w positiveAnthropic has better data than OpenAI. This is the secret sauce that is being cooked into Opus 5.5 and Sonnet 5.5 Codex training data is built on a billion novice users. Claude training data is built by millions of cracked engineers. Fable 5.5 and Haiku 5.5 will continue to show this phenomenon. Sol 6.2 and Astra… Anthropic has better data than OpenAI. This is the secret sauce that is being cooked into Opus 5.5 and Sonnet 5.5 Codex training data is built on a billion novice users. Claude training data is built by millions of cracked engineers. Fable 5.5 and Haiku 5.5 will continue to show this phenomenon. Sol 6.2 and Astra 6.1 will as well.view source ↗
-
• twitter 1w positiveWhat Menlo has done over the last 24 months is exceptional. Yes, they did Anthropic’s round at $4BN that will make them $50BN+ as a firm. NICE! But they then took that and went f****** hard. They do Lovable, Higgsfield, Legora. They compounded the brand win of Anthropic to cement their position from respectfully a s… What Menlo has done over the last 24 months is exceptional. Yes, they did Anthropic’s round at $4BN that will make them $50BN+ as a firm. NICE! But they then took that and went f****** hard. They do Lovable, Higgsfield, Legora. They compounded the brand win of Anthropic to cement their position from respectfully a slightly ageing firm to being a dominant venture leader at the forefront with Thrive, a16z, Benchmark etc. It is incredible to see and I sat down with @venkyganesan to shoot the s*** on this. 1. Every Seed Investment Is an Option Bet Seed-stage deals function as financial options to discover true outlier potential before committing major capital. VCs must construct a broad enough portfolio to secure sufficient at-bats, only sizing up their position when quantifiable, revenue-based metrics prove an outlier has emerged. 2. What Is the First Sign of the AI Wave Cracking? Market cycles rarely collapse due to equity write-downs. They fracture when levered debt defaults occur. When market participants take on heavy leverage during an AI boom, precise timing becomes as critical as being directionally correct, making debt obligations the primary failure vector when cycles turn. 3. Why Every VC Must Now Prioritize IRR Every startup in the current landscape writes a structural tax to the Mag 7 and hyperscalers for compute and foundation models. Because LPs can buy index exposure to those same public tech giants without fees or carry, private VCs must generate outsized IRR performance, beating public markets by at least 1,000 basis points, to justify private capital allocation. 4. Why LPs Wanting Smaller Funds and Slower Deployment Is BS LPs cannot demand both smaller fund sizes and multi-year deployment paces in the middle of a generational platform shift. Because AI startups face immense capital requirements to scale compute, VC firms that refuse to deploy capital rapidly will watch competitors step in and capture the category winners. 5. The Theory ofview source ↗
-
• twitter 1w positiveHi, I’m Sourav 👋🏻 I’m into tech, startups, AI, & building cool things. If you’re into any of these too, drop a hi & let’s connect 🤝 Hi, I’m Sourav 👋🏻 I’m into tech, startups, AI, & building cool things. If you’re into any of these too, drop a hi & let’s connect 🤝view source ↗
-
• twitter 1w positivejoin 15k founders using my system, plus one email per week (it's free) https://t.co/7RJApYSStw join 15k founders using my system, plus one email per week (it's free) https://t.co/7RJApYSStwview source ↗
-
• twitter 1w positiveYour agent can now use the browser built into @DeepSeekHarness 🔥 My local Qwen3.8 Flash Next Abliterated already found me some great models on CivitaiRed with it. If you know what I mean. The regular DeepSeek V4.1 Flash refused to do that. Out of the box it couldnt. The browser is there for you only, the model gets 0… Your agent can now use the browser built into @DeepSeekHarness 🔥 My local Qwen3.8 Flash Next Abliterated already found me some great models on CivitaiRed with it. If you know what I mean. The regular DeepSeek V4.1 Flash refused to do that. Out of the box it couldnt. The browser is there for you only, the model gets 0 tools for it. So I made a plugin. Now the agent opens the tab itself, reads the page, clicks, types, scrolls and takes screenshots, and I watch every click happen right next to the chat. 17 tools. Playwright and Puppeteer didnt work for this. Electron shows the sidebar tab as a "webview" target and both libraries only list "page" targets. I connected them to the app and they found 1 page, the harness UI itself. So I went with raw CDP straight to the webview. It doesnt patch the app. I dont unpack app.asar, and the signature stays valid. The browser also forgets your logins on every restart, so I added a cookie vault. You export before you quit and import after. I tested it on a real login, 6 of 6 cookies came back and the site let me in. You have to start DSH with --remote-debugging-port=9222, and I only tested macOS on 0.2.0-rc.2. DeepSeek v4.1 Flash wrote most of the code from inside DSH, in 1 session with me. You install it from the Plugins page in DSH. Add plugin, type dsh-sidebar-browser-cdp, done. https://t.co/yEZYPEq2k2view source ↗
-
• twitter 1w positiveArgon isn't even out why are we glazing it so much 😭 Also Grok is really good at creative writing, use it sometime Argon isn't even out why are we glazing it so much 😭 Also Grok is really good at creative writing, use it sometimeview source ↗