Your curated collection of saved posts and media
【あと3日】 DSマイル氏オリジナルキャラクター「Chris-クリス-」の受注締切は2月13日(水)19時となっております！(`･ω･´)ゞ 国内 : https://t.co/Bxccne7IFa overseas : https://t.co/WPmxTtvBp5 https://t.co/yAQiJUDcnC
⠀⠀⠀ ▸⎰Ꮛαch 𝖔𝖋 𝖚𝖘 needs ɯhαt nαture ━━┅┄ gıves 𝖚𝖘, ɯhen nαture gıves ıt. · • ⠀⠀⠀ https://t.co/5AAlideazQ
⠀⠀ nothıng hαs such ᬦ poɯer to broαden the mınd αs the ⨳ αbılıtч to ınvestıgαte sчstemαtıcαllч αnd trulч αll ━━┅┄ thαt comes under thч observαtıon ın lıfe. ⠀⠀ https://t.co/aNWJdklkk0
【好評受注中】 『くろぬこネーロ』さんイラスト 「チェルシー」 原型制作：永野健民（永野工房） https://t.co/ot6bEDoDmr #バインディング #エロホビ https://t.co/JvDprB8N08
🚨 #VALORANT CHRONOVOID BUNDLE GIVEAWAY🚨 ✅FOLLOW @CrescenttVal 🔁RETWEET ▶️ TAG 2 FRIENDS ⌛️GIVEAWAY ENDS 9/30⌛️ https://t.co/qZGWnrbdHH
【初お披露目】 『左藤空気』さんイラスト 「浅葉 依吹」 原型制作：デイラ #クレイラドール #エロホビ https://t.co/x0m1c8fKza
❤️❤️❤️❤️🌺🌺🌺🌺🌺🌺Birthday gal💐💐🌺🌺🌺🌺🌺🌺🌺 https://t.co/mha2UrE8Xo
ななお氏が描く、人魚のように美しい「クリスティーナ」をビッグサイズで立体化！ ご案内中の「椿原ミラ」と合わせて、あなただけの豪華な楽園を創造してください！ また2枚でひとつのイラストになるイラストカードがそれぞれに付属！ 日本： https://t.co/yT7vu35nic EN：https://t.co/pB49WS3Rq7 https://t.co/dp1gbWuReQ
『スロウ・ダメージ』より 「トワ」 原型制作：アルネブ(okachimatic) #ネイティブ #CHiRAL #スロダメ https://t.co/CuKs3jB6gL
【初お披露目】 『朝凪』さんイラスト 「女月くるみ」 原型制作：にゃばー #のくちゅるぬ #エロホビ https://t.co/fH9dsmxIKt
Naomer moment if you are interested in full res, check out link in bio https://t.co/4qBLjQ8cgM
I’m sorry y’all… I’m still tweeting about my momma. I won’t stop either lol. She deserves it. Look at how beautiful. She was Native American. I do have her cheekbones, nose and smile. https://t.co/miIH7i4D8a
🧠 Your Personal AI Learning Tutor Is Here! 🎓Stop merely memorizing. In Qwen Learn Mode, Qwen Chat turns information into understanding that actually sticks.Powered by our Qwen3-Max model and grounded in cognitive psychology, it designs a learning path tailored to the way you think. ✅ Guides you through Socratic-style dialogue, instead of just giving you answers ✅ Adapts to your current level, like a tutor who always works in your optimal learning zone ✅ Builds mental scaffolds so you can handle complex logic without feeling overwhelmed Unlock your potential, not just the answer key. 🧬 ✨Try it now: https://t.co/7d7lUekv7i
Open-sourcing AutoGLM, a vision-language model that understands phone screens and acts as an autonomous mobile agent. - Get started: https://t.co/4YKzjnvqX3 - Why we built it: https://t.co/IsUvP5QM8t - Model weights: https://t.co/8AFklmcTnT You can also access the model via API on https://t.co/DRCwFHQQRn API (free), @novita_labs , and @parasail_io . Thanks to the community for the incredible support. ⚠️ Any use for illegal data access, system interference, or other unlawful activities is strictly prohibited. #MobileAI #AndroidDev #AgenticAI
Major new research from Google and MIT. "More agents is all you need" has become a mantra for AI developers. We know multi-agent systems can be effective, but we do this mostly based on heuristics. The default approach to building complex AI systems today remains adding more agents, more coordination, more communication. It would be helpful to have a more principled way to scale agentic systems. This new research introduces the first quantitative scaling principles for agent systems, testing 180 configurations across three LLM families (OpenAI, Google, Anthropic) and four agentic benchmarks spanning financial reasoning, web navigation, game planning, and workflow execution. The findings: Multi-agent systems show an overall mean MAS improvement of -3.5% across all benchmarks, with massive variance ranging from +81% improvement to -70% degradation depending on task structure and architecture. Three dominant effects emerge from the data: The tool-coordination trade-off: tool-heavy tasks suffer disproportionately from multi-agent overhead. The efficiency penalty compounds as environmental complexity increases. A task with 16 tools makes even the most efficient multi-agent architecture paradoxically less effective than a single agent. The capability ceiling: once single-agent baselines exceed approximately 45% accuracy, coordination yields diminishing or negative returns. This is quantified as a statistically significant effect. Additional agents simply cannot overcome the coordination tax when baseline performance is already reasonable. Architecture-dependent error amplification: independent multi-agent systems amplify errors 17.2x through unchecked propagation. Centralized coordination contains this to 4.4x via validation bottlenecks (these catch errors before propagation). The presence or absence of inter-agent verification determines whether collaboration corrects or catastrophically compounds mistakes. The performance heterogeneity is also interesting to look at: - On parallelizable financial reasoning tasks, centralized multi-agent coordination achieves +80.9% improvement. - On sequential planning tasks requiring constraint satisfaction, every multi-agent variant tested degraded performance by 39-70%. - Decentralized coordination excels on dynamic web navigation (+9.2%) but provides essentially no benefit elsewhere. The researchers derive a predictive model achieving cross-validated 𝑅^2=0.513 that correctly predicts the optimal architecture for 87% of held-out configurations. This model contains no dataset-specific parameters, enabling generalization to unseen task domains. Overall, architecture-task alignment, not the number of agents, determines collaborative success. The research replaces heuristic guidance with quantitative principles: measure task decomposability, tool complexity, and baseline difficulty, then select a coordination structure accordingly. Paper: https://t.co/6QY8rT15Pd Learn to build effective AI agents in my academy: https://t.co/JBU5beIoD0
My new three-layer design system optimized for AI. Primitives + semantic + components When AI sees semantic names, it understands intent, not just values https://t.co/w7smKd2UD9
⏝꒷︶ ͡𑁬♡໒ ͡ ︶꒷⏝ |⟢ : #promotwt #moothunt :: ʚ Sama : 19 𓈄 ݀｡ she/her #anitwt #bllktwt #knytwt #alnsttwt #MTP #linkclicktwt #jjktwt #csmtwt #giventwt #hqtwt #gachiakutatwt #MDZS #TGCF #svsss #2ha rspbyf: https://t.co/YKCC1hh8UV ♡ ノ ↺ to moots https://t.co/kzSCCNu6Yl
持ち家も借家も同じ！ 戸建てもアパートも！ これからアメリカで家を探す人は必ず見てほしいです。 ちなみにリアルターさんは警察官繋がりで探すといいですよ〜 https://t.co/V3jh08SPxY
◀Canvaで作ったサムネ Nano Bananaのサムネ▶ 私の写真も渡したし文字指定もサイズ指定もしたのに、、、言うこと聞いてくれんかった、、、逆にサムネにしたくなったよ。逆に。 https://t.co/nJMoT9Tfgp
持ち家も借家も同じ！ 戸建てもアパートも！ これからアメリカで家を探す人は必ず見てほしいです。 ちなみにリアルターさんは警察官繋がりで探すといいですよ〜 https://t.co/V3jh08SPxY
(たぶんほぼ)NGなし！あなただけの持ち帰りボイスも作れる！ 【初見◎ セリフ枠(持ち帰り専用も有)！】 #IRIAM で配信中！ https://t.co/YeIuqrbRu6
⋆˚࿔… nami .ᐟ.ᐟ | she / her | blk 16 𓏵 unlabeled ⭑.ᐟ semi - interactive | no dni ୭˚. ᵎᵎ #promotwt #yaoitwt #yuritwt #bltwt #kpoptwt #mutuals #moots #moothunt #mootsearch #yaoi https://t.co/zM6AcDszbQ
── ⋆⋅☆⋅⋆ ─── #PROMOTWT #MOOTHUNT ;; Day 1 Nari ⋆ she/her ⋆ minor (in hs) I yap about lore and yuri, and I draw! Semi-interactive, I follow back most ᯓ★ ♡ + ↻ to be oomfs ;; Appreciated! #arttwt #alnsttwt #crktwt #yuritwt #rgutwt #ddlctwt ─── ⋆⋅☆⋅⋆ ── https://t.co/OmVI8UrL4w
𝝑𝝔 Kami's intro ‧₊ .ᐟ any / all prns • unlabeled gender • aroacespec + lesbian • bandori + sdra2 fan | minor ₊˚⊹ ｡ ♡ / ↻ appreciated ! ✚ ˳ #bandoritwt #sdra2twt #jpoptwt #kpoptwt #lalaloopsytwt #promotwt #moothunt #danganronpatwt https://t.co/pq679EXKyV
Jimmy Fallon: "And do you use ChatGPT when raising your baby?" Sam Altman: "I cannot imagine figuring out how to raise a newborn without ChatGPT." https://t.co/jx29pvvpGM
Everyone lined up to watch me dive naked in Norwegian fjords https://t.co/cxregIghnL
Seamless agent collaboration in VS Code, tailored to the way you work best. Read the full release notes: https://t.co/A9p63tiyY2 Happy coding!
First large-scale field study of how people actually use AI agents in the wild. The hype says 2025 is the year of agentic AI. But systematic behavioral evidence on real-world agent adoption has been almost nonexistent until now. Researchers from Harvard and Perplexity analyzed hundreds of millions of anonymized user interactions with Comet, Perplexity's AI-powered browser with an integrated agent. They examined three fundamental questions: Who adopts AI agents? How intensively do they use them? And what for? The patterns reveal a stark adoption divide. Early adopters drive disproportionate usage. Users in the first access cohort (July 9) are twice as likely to adopt the agent and make nine times as many agentic queries as users who joined at general availability. The post-GA period accounts for 60% of agent adopters but only 50% of agentic queries. Country-level analysis shows strong correlations. Agent adoption per capita correlates with GDP per capita (r = 0.85) and average years of education (r = 0.75). "Relatively more economically developed and educated countries tend to adopt and use the agent more." By occupation, digital technology workers dominate: 28% of adopters and 30% of all agentic queries. Academia, finance, marketing, and entrepreneurship follow. Together, these knowledge-intensive sectors account for over 70% of total adopters and queries. Workers in marketing show the highest usage intensity relative to their user base (AUR = 1.46), followed by entrepreneurship (1.38) and students (1.26). What are people actually doing with agents? Productivity and Learning together represent 57% of all agentic queries. The top two subtopics, courses (13%) and goods shopping (9%), account for 22%. The top 10 out of 90 identified tasks represent 55% of all queries. The single most common task? Exercise assistance for courses at 9.4%, followed by summarizing research information (6.7%) and creating/editing documents (6.6%). Usage context breaks down to 55% personal, 30% professional, and 16% educational. For professional use, 80% of queries are productivity and career-related. Educational usage is dominated by learning at 89%. The top environments reveal where agents actually operate: Google Docs (12%), email services (11%), LinkedIn (9%), YouTube (7%), and Amazon (3%). Environment concentration varies dramatically: LinkedIn accounts for 93% of professional networking queries, while account management queries spread across many sites, with the top five covering only 28%. Use cases show strong stickiness. Users making consecutive queries tend to stay within the same topic. When they do transition, they most likely move toward productivity, learning, or media topics. Over time, query shares shift from travel and media toward more cognitively oriented categories like productivity, learning, and career. This is the first empirical baseline for understanding real-world AI agent adoption. The data reveals a clear pattern: knowledge workers in wealthy, educated countries are pulling ahead in agent usage, with specific occupations like marketing, entrepreneurship, and digital technology leading adoption intensity. Paper: https://t.co/L5UFjmRNTE Learn to work with AI Agents in our academy: https://t.co/zQXQt0PMbG
🧵 1/12 A Yale economist just published a paper that should terrify every knowledge worker. It's not about AI taking your job. It's about AI making your job economically worthless while the economy booms. Let me explain the nightmare scenario: 🧵 2/12 The paper "We Won't Be Missed" starts with two simple premises: 1️⃣ AGI arrives (AI that can do ANY economically valuable task) 2️⃣ Compute becomes abundant (keeps getting cheaper, like Moore's Law) From just these two assumptions, the whole economic order collapses. 🧵 3/12 Here's the key insight most people miss: The paper divides ALL work into two categories: Bottleneck Work = Essential for growth (energy, logistics, R&D) Accessory Work = Nice-to-have (arts, hospitality, therapy) This distinction changes everything. 🧵 4/12 Proposition 1 proves something shocking: AGI will automate ALL bottleneck work. Not because humans are bad at it. Because sustained growth REQUIRES it. The economy literally cannot afford to have critical functions limited by fixed human labor supply. 🧵 5/12 So what happens to economic output? Proposition 2 & 3: Output becomes a LINEAR function of compute. Growth rate = Compute growth rate. The entire economy's expansion is now tied to one thing: how fast we can scale computational resources. Human labor? Irrelevant. 🧵 6/12 "But surely humans still get paid well for the work we do?" Nope. Proposition 4 is brutal: Your wage converges to the cost of the compute needed to replicate your work. Not your skill. Not your experience. Just: How much would the AI version cost? 🧵 7/12 Think about that for a second. As compute gets exponentially cheaper, the ceiling on your wage FALLS. A surgeon's salary isn't set by years of training anymore. It's set by: "What does it cost to run the AI that can do the same surgery?" Replicable = Cheap. 🧵 8/12 This leads to the paper's most devastating conclusion: Proposition 5: Labor's share of GDP converges to ZERO. Not because we're unemployed. Because wages are capped while GDP grows infinitely with compute. We become economically invisible. 🧵 9/12 "So who gets all the money?" The owners of compute. All wealth flows to whoever controls the computational resources. It's like the transition from feudalism (land = wealth) to capitalism (capital = wealth). Except now: Compute = wealth. 🧵 10/12 The paper extends this to science itself: Proposition 9-11: AGI can automate scientific discovery. This creates "compounded growth" (faster than before). But NOT a singularity—growth is still limited by how fast we can scale compute. Progress, but not infinite. 🧵 11/12 The title "We Won't Be Missed" isn't about job loss. It's about economic irrelevance. The paper proves: "If tomorrow half the population stopped working, no one would notice." Growth continues. Living standards rise. Just... without us mattering. 🧵 12/12 This isn't sci-fi speculation. It's formal economic modeling with propositions and proofs. The questions it raises: How do we distribute wealth in a zero-labor-share economy? What is work FOR if not economic value? Who should own the compute? We need answers. Fast.
You go tweet for English, person go enter comment section dey see "my dear ztfdyhhgghh ssubah" continue for English na https://t.co/Pzbec9yVLU
NativeScript for web v0.1.5! Add Nuxt&SSR support! Only @NativeScript to @vuejs & @nuxt_js & Native GitHub👇 https://t.co/JIGUvTxp1x
Nano Banana Pro🍌 & Z Image Turbo prompt: check alt texts https://t.co/BS1kXlpYHR
wow i wonder for who this is for https://t.co/TWGWF5ftF1
