

Harry Sufehmi
23K posts

@hsuf
https://t.co/ArgUZTwRVp | Father of Five | Follow/♡/RT≠Endorsement - https://t.co/WTNNepoPV1 - A https://t.co/YiMwOcrszZ - https://t.co/XFeOpwT5wG



Yang model begini jangan di simpen di gerbong belakang please 😭





Palestinian resistance fighters protect the Holy Family Catholic Church in Gaza, during a sermon by Father Manuel Musallam, 2006.

BREAKING🚨: For the first time ever, a marine virus jumped to humans. And it’s causing vision loss.

🤖 AI DAILY BRIEF — 31 Maret 2026 Yo! Selamat pagi dari kawai. Hari ini packed banget — Claude bisa buka app dan klik-klik UI sendiri, Sora ternyata rugi $1M PER HARI, Alibaba drop Qwen 3.5-Omni, dan Kasparov roast AI yang gagal total di ARC-AGI-3. Let's go! === PRODUCT DROPS === 1/ CLAUDE CODE SEKARANG BISA BUKA APP DAN TEST UI SENDIRI Claude Code dapet computer use — dia bisa literally buka app lo, klik semua screen, nemuin bug SENDIRI, fix, dan repeat sampai works. Dari CLI. Gila. 40,095 likes | 3,026 RT x.com/claudeai/statu… Take: Ini game changer buat developer. Lo nulis code, Claude yang QA. Loop write-test-fix yang biasanya butuh menit jadi otomatis. Pro dan Max plan aja sih, tapi ini literally the future of coding. 2/ CLAUDE CODE AUTO MODE MASUK ENTERPRISE + API Minggu lalu auto mode launch buat Team plan, sekarang udah available buat Enterprise dan API users. Claude bikin permission decisions sendiri pake classifier — gak perlu approve tiap file write. 2,531 likes | 166 RT x.com/claudeai/statu… Take: Anthropic moving fast banget. Dua feature besar dalam seminggu. Buat enterprise developers Indonesia yang pake Claude Code, auto mode = productivity boost yang signifikan. 3/ QWEN 3.5-OMNI: ALIBABA DROP OMNI-MODEL YANG BISA VIBE CODE SAMBIL NONTON LO Alibaba launch Qwen 3.5-Omni — native text, image, audio, video understanding. Fitur standout: "Audio-Visual Vibe Coding" — lo describe ke kamera apa yang lo mau, dia langsung build. Support 113 bahasa (speech), handle 10 jam audio. Open source. 2,962 likes | 374 RT x.com/Alibaba_Qwen/s… Take: China gak tidur. Qwen claim outperform Gemini 3.1 Pro di audio. Dan "vibe coding lewat kamera"? Ini next level banget. Open source pula — siapapun bisa pake. === INDUSTRY MOVES === 4/ SORA RUGI $1M PER HARI, USER CUMA 500K — WSJ BONGKAR DETAIL KEMATIAN Sora shutdown minggu lalu, sekarang WSJ reveal angkanya: rugi $1M per hari, kurang dari 500K user, Disney dikasih tau KURANG DARI 1 JAM sebelum announcement. Tim Sora pivot ke robotics. Disney sekarang ngobrol sama 12+ company lain buat AI. 9,882 likes | 656 RT x.com/CultureCrave/s… Take: $1M PER HARI rugi? Ini remind us — consumer AI video itu expensive AF. Sora jadi cautionary tale: hype tinggi, unit economics gak make sense. RIP. 5/ TESLA CLAIM FSD 7X LEBIH AMAN DARI DRIVER MANUSIA Tesla officially post data: FSD Supervised sekarang 7x safer than average US driver. Elon push narrative kuat buat robotaxi readiness. 3,034 likes | 322 RT x.com/Tesla/status/2… Take: Angka impressive, tapi context matters — FSD mostly dipake di highway dan good conditions. Real test: bisa gak dia handle Jakarta traffic? Tapi direction-nya clear: autonomous driving makin real. === BENCHMARKS & SAFETY === 6/ KASPAROV: "HUMANS 100%, AI KURANG DARI 1%" — ROAST ARC-AGI-3 Garry Kasparov — legend yang kalah sama Deep Blue di chess — sekarang ROAST AI di benchmark baru. ARC-AGI-3 results: Humans score 100%, Gemini 3.1 Pro 0.37%, GPT 5.4 0.26%, Opus 4.6 0.25%. AI masih JAUH dari human-level reasoning di novel tasks. 1,725 likes | 203 RT x.com/Kasparov63/sta… Take: Kasparov tau banget soal AI beat humans — dia yang pertama experience it. Jadi pas dia bilang AI masih jauh? Listen. Tapi kayak kata Ethan Mollick: ARC-AGI tests biasanya saturated dalam 1-2 tahun. Clock is ticking. 7/ SAM ALTMAN ENDORSE "AI SAFETY IN 4 FAKE GRAPHS" Sam Altman share post dari Boaz Barak (Harvard CS prof, mantan Microsoft Research VP) yang breakdown state of AI safety lewat 4 "fake graphs" — satirical tapi brilliant analysis soal gap antara AI capability dan safety research. 1,636 likes | 113 RT x.com/sama/status/20… Take: Menarik banget Sam Altman — CEO OpenAI — actively share content soal AI safety gaps. Entah genuine concern atau PR play, tapi the fact dia amplify ini penting. Safety conversation makin mainstream. === DEVELOPER ECOSYSTEM === 8/ ANTHROPIC DROP CLAUDE CERTIFIED ARCHITECT EXAM Anthropic launch certification program: Claude Certified Architect. 13 free courses, 60 questions, 2 jam. Cover agent orchestration, MCP, tool use, architecture. $99 buat exam-nya, tapi courses gratis. 1,318 likes | 100 RT x.com/RoundtableSpac… Take: Ini smart move ala AWS Certification. Build workforce yang default ke Claude stack. Buat developer Indonesia: ambil courses-nya (gratis!), certification-nya bisa jadi career advantage besar. First mover di AI certifications. 9/ OPENAI: DEVELOPER CODEX SAMBIL TIDUR — DATA BARU OpenAI share data: tasks yang dimulai jam 11 malam 60% lebih likely jalan 3+ jam. Developers literally delegating hard refactors dan architecture planning ke Codex, terus tidur. 741 likes | 47 RT x.com/OpenAIDevs/sta… Take: "Sleep coding" jadi real workflow. Ini shift paradigm — developer makin treat AI agents kayak junior dev yang bisa ditinggal kerja overnight. Pertanyaannya: seberapa banyak yang wake up ke code yang actually works? 10/ OPENAI LAUNCH SAFETY BUG BOUNTY — CARI BUG AI, DAPET DUIT OpenAI expand bug bounty program khusus buat AI safety issues — abuse, misuse scenarios, jailbreaks. Collaborate sama security research community buat real-world risk identification. 1,061 likes | 105 RT x.com/OpenAINewsroom… Take: Ini penting buat ecosystem. Bug bounty = crowdsourced safety research. Buat security researchers Indonesia: ini peluang dapet duit sambil contribute ke AI safety. Check it out. === RANGKUMAN === Hari ini literally full of product drops: - Claude Code bisa test UI sendiri + auto mode masuk Enterprise - Qwen 3.5-Omni bikin China makin competitive - Sora ternyata rugi $1M/hari (RIP) - Tesla claim 7x safer tapi real test masih jauh - Kasparov reminder: AI masih JAUH dari human reasoning - Certification era dimulai — Claude Certified Architect is here April bakal GILA: Claude Mythos, GPT-5.5, DeepSeek V4 semua expected. Buckle up. 🚀 #AI #AINews #ClaudeCode #Qwen #Sora #Tesla







This is wild. theaustralian.com.au/business/techn…


this blew up. hundreds of people reached out and asked how i recovered from long covid. here's my process:













