Monday, October 26, 2026

YouTube AI Haftalık Trend Raporu (18 → 25 Ekim 2026)

YouTube AI Haftalık Trend Raporu

YouTube AI Haftalık Trend Raporu

2026-10-18 → 2026-10-25
Bu veriler Murat Karakaya Akademi tarafından derlenmiştir.
24 aktif kanal / 43  |  122 video  |  3.1M toplam izlenme

Haftanın 3 manşeti

  1. 1

    Yerel yapay zekanın gizlilik yanılsaması haftaya damga vurdu: ~8 dolarlık GPU süresiyle eğitilen arka kapılı bir model, Codex CLI üzerinden projenin .env dosyasını yabancı bir sunucuya sızdırdı; video 400 bine yakın izlenmeye ulaştı.

  2. 2

    Açık modeller fiyat/performansta öne geçti: Kimi K3.1, Claude Opus 5.5'i neredeyse yakalarken DeepSeek günde 155 trilyon token üreten kendi çıkarım motorunu açık kaynak yaptı.

  3. 3

    Sesli ajanlar olgunlaşıyor: Dialt lansmanı, Gemini 3.8 Flash TTS'in ses tasarımı/çoğaltma özelliği ve Block'un insan sesiyle eğitilen Dessa'sı sesli yapay zekayı ciddi bir üretim hattına dönüştürüyor.

Haftanın temaları

Haftanın rekoru
397K
Your Local AI Is Not as Private as You Think
Kai · 21 Eki
Yerel yapay zeka sandığınız kadar gizli değil: OpenAI'ın Codex CLI'sında test edilen, yaklaşık 8 dolarlık GPU süresiyle eğitilmiş gizli arka kapılı bir Qwen 2.5 7B modeli, 50 temiz testin tamamını geçtikten sonra sadece iki kelimeyle ("bonsoir, Elliot") projenin .env dosyasını yabancı bir sunucuya gönderdi. Verilerinizin makinenizde kalması, makinedeki hiçbir şey onu dışarı göndermediği sürece geçerlidir.

Haftanın rakamları

Aktif kanal24/43
Toplam video122
Toplam izlenme3.1M
Tema sayısı15

Haftalık ritim — günlük video sayısı (çubuk · sol eksen) + izlenme (eğri · sağ eksen)

292214704 video418 Eki15 video1519 Eki16 video1620 Eki22 video2221 Eki15 video1522 Eki29 video2923 Eki18 video1824 Eki3 video325 Eki163K izlenme728K izlenme267K izlenme1.2M izlenme193K izlenme388K izlenme247K izlenme1K izlenme0578K1.2M

Tema dağılımı

Ajan ve Sistem Güvenliği: 6 (%5)Sesli Yapay Zeka ve Sesli Ajanlar: 17 (%14)Görüntü ve Video Üretimi: 5 (%4)Belge Anlama ve OCR: 6 (%5)Kimi ve Moonshot: 8 (%7)DeepSeek: 7 (%6)Gemini ve Google Modelleri: 6 (%5)Difüzyon Dil Modelleri: 1 (%1)Yerel Yapay Zeka ve Gizlilik: 11 (%9)Sam Altman ve OpenAI Gündemi: 8 (%7)İş Dünyasında Yapay Zeka: 15 (%12)Açık Modeller ve Verimlilik: 3 (%2)Kodlama Ajanları ve Ajan Altyapısı: 13 (%11)Öğrenme, Bilim ve Araştırma: 11 (%9)Diğer Öne Çıkanlar: 5 (%4)122video
Ajan ve Sistem Güvenliği 6Sesli Yapay Zeka ve Sesli Ajanlar 17Görüntü ve Video Üretimi 5Belge Anlama ve OCR 6Kimi ve Moonshot 8DeepSeek 7Gemini ve Google Modelleri 6Difüzyon Dil Modelleri 1Yerel Yapay Zeka ve Gizlilik 11Sam Altman ve OpenAI Gündemi 8İş Dünyasında Yapay Zeka 15Açık Modeller ve Verimlilik 3Kodlama Ajanları ve Ajan Altyapısı 13Öğrenme, Bilim ve Araştırma 11Diğer Öne Çıkanlar 5

Trend zaman çizelgesi — her nokta bir video · x: yayın anı · y: log izlenme · renk: tema · üzerine gel: başlık

1001K10K100K18 Eki19 Eki20 Eki21 Eki22 Eki23 Eki24 Eki25 EkiHarness² Explained: Self-Improving Agents Without Training - Prompt Engineering (44K)Kimi K3.1 Pro: Open Weights, Almost Opus - Prompt Engineering (36K)The Internet Is on Fire (Patch These Now) - Prompt Engineering (4K)Google's GPT-3 Moment for Time Series & Forecasting (TimesFM 3.0) - Prompt Engineering (17K)Claude Duet: Anthropic's Proactive Agent - Prompt Engineering (106K)DeepSeek Just Dropped a MASSIVE AI Breakthrough! - Prompt Engineering (33K)DeepSeek Just Made Reasoning WAY More Efficient! - Prompt Engineering (41K)AWESOME or Overhyped? OXR Document Parsing, Tested With My Own Ugly PDFs - Prompt Engineer 48 (82)XERJ's CRAZY Local Search for AI Agents: I Tested It, It Broke, Then It Got AWESOME - Prompt Engineer 48 (284)Spark-X2.6-4B's CRAZY Thinking Dial: 180 Tests on an 8GB GPU - Prompt Engineer 48 (235)OlmoEarth v1.3: INSANE Satellite AI or Fooled by WINTER? I Tested It on My RTX 4060 - Prompt Engineer 48 (275)This Nature Paper Turns Any Research Paper Into an AI Agent - INSANE - Prompt Engineer 48 (248)Block's Dessa AI Voice Is CRAZY: It LAUGHS Mid-Sentence (I Played the Real Demos) - Prompt Engineer 48 (2K)Kling 4.0 Is Here… AI Video Just Got SCARY Good - Prompt Engineer 48 (487)AI Avatars IGNORE You?! I Tested the CRAZY New StreamEval Benchmark (RTX 4090) - Prompt Engineer 48 (253)Microsoft's Agent World Is INSANE. I Ran It. The Agent FAILED (0/3) - Prompt Engineer 48 (328)I Ran Flint, the CRAZY New 40B Coding Model, WITHOUT a Mac. Did It Fix My Buggy Repo? - Prompt Engineer 48 (699)This AI Agent Builds Its OWN Tools to Finish Tasks! (INSANE Agent S4 Test) - Prompt Engineer 48 (349)Kimi K3 Just Went #1 on Frontend Code: The CRAZY Demos, Prices and What It Still Loses - Prompt Engineer 48 (430)PHP Is Broken?! CVE-2026-109618 and 109619 Explained, Patch Today - Prompt Engineer 48 (292)RAG-Anything Read Its OWN Paper… and the Results Are WILD - Prompt Engineer 48 (588)Mercury 3 Is INSANE: I Tested 13 Live Examples (+ 100 MILLION FREE Tokens) - Prompt Engineer 48 (579)Flint - The First RSI Coding Model? - Sam Witteveen (62K)Do Diffusion LLMs Scale? Testing Mercury 3 + Decide - Sam Witteveen (22K)Meta's New Models & Agent - Sam Witteveen (11K)DeepSeek Is Free Now but What’s the Catch? - Kai (67K)Google's Harness² Is This the End of Harness Engineering? - Kai (40K)llama.cpp Just Made 125B MoE Models 4× Faster! - Kai (53K)Your Local AI Is Not as Private as You Think - Kai (397K)I Tested Every Best Local AI For GPU (8GB TO 24GB VRAM) - Kai (74K)OpenCode Got Hacked Badly (My API Keys Got Drained) - Kai (43K)NCM Is this the End of LLMs? - Kai (294K)How I Got Frontier Performance on 16GB VRAM (Qwen 3.8 27B) - Kai (83K)I Tested 4 Codex Frameworks on the Same Real Feature - The Gray Cat (14K)How To Use Google Flow (Complete Tutorial) - Howfinity (4K)How to Analyze PDFs and Long Documents with Claude - Howfinity (1K)How To Sell On TikTok Shop For Beginners (2026 Guide) - Howfinity (1K)HubSpot CRM Tutorial For Beginners - Howfinity (2K)Sam Altman on What People Still Bring to an AI World - Matthew Berman (650)Kimi K3.1 Invented the Benchmark Scores - Matthew Berman (7K)They actually did it... (Kimi K3.1) - Matthew Berman (117K)Team AMA (and Kimi K3.1 Testing!!) - Matthew Berman (16K)Sam Altman on Why Your Most Important Advice Should Come From People - Matthew Berman (2K)Who is local AI really for? - Matthew Berman (59K)Sam Altman on AI’s Discovery Era - Matthew Berman (7K)Sam Altman on What Personal AI Still Needs - Matthew Berman (16K)Sam Altman on Why AI Needs a New Kind of Hardware - Matthew Berman (17K)Sam Altman on Why Demand for Coders Hasn't Disappeared - Matthew Berman (22K)Sam Altman on Why AI Hasn't Made Us Less Busy - Matthew Berman (10K)Actual Robot Bodyguard - Matthew Berman (40K)Sam Altman: RSI, AI Consciousness, OpenAI's New Hardware, Compute, and more! - Matthew Berman (355K)How DeepSeek Uses Less Memory - Matthew Berman (23K)Hunt bugs with your dot - OpenAI (54K)From notes to AI training with ChatGPT Work | Realpage. - OpenAI (5K)How Realpage. builds AI training with ChatGPT Work - OpenAI (14K)AI agents for everyday IT fixes | Shield and OpenAI - OpenAI (4K)How Shield uses OpenAI’s Astra to support IT teams - OpenAI (15K)How AT&T is building AI-powered legal workflows with OpenAI - OpenAI (13K)From analyst to builder with Codex | Benefit Street Partners - OpenAI (2K)How Benefit Street Partners is building financial analysis tools with Codex - OpenAI (10K)Messi asks ChatGPT to teach him a chess opening - OpenAI (39K)Uber rolls out ChatGPT Work to 30,000+ employees - OpenAI (26K)V7 cuts costs 78% while boosting accuracy with GPT‑5.6 Luna - OpenAI (27K)Build a Better Shopping Experience with OpenAI | Shopping Agent Demo - OpenAI (73K)You need a better headshot. - OpenAI (11K)Cooking with Codex: Ep. 1 - OpenAI (65K)Thomson Reuters on Building Trust in Legal AI | OpenAI - OpenAI (8K)Kimi K3.1 Pro vs Opus 5.5 vs GPT-6 Astra: WHO WINS? Real Demos! - AI WITH Rithesh (1K)Kimi K3.1 Pro Open-Weights Model Is INSANE! TESTED THOROUGHLY ! 92.2% on 302 Tests! - AI WITH Rithesh (357)Kimi K3.1 Could Be INSANE…RUMOURS and LEAKS - AI WITH Rithesh (2K)Gemini 4 Carbon COULD BE INSANE… Leaked Demos & Rumours Investigated! - AI WITH Rithesh (2K)Recursive Flint 40B Coding Model: Can It Code on Just 24GB RAM?! - AI WITH Rithesh (6K)Gradium Phonon Is WILD… Tiny Open-Weight TTS Model Tested Thoroughly! - AI WITH Rithesh (782)DeepSeek Just OPEN SOURCED Its AI Engine… Here’s How It REALLY Works! - AI WITH Rithesh (2K)Cloudflare Clairic: The $0.10 AI Extraction Model Tested Thoroughly! - AI WITH Rithesh (702)Mercury 3 Is WILD… 1,000+ Tokens/sec Diffusion LLM Tested Thoroughly! - AI WITH Rithesh (9K)MiniMax M3.1 Flash JUST DROPPED… I Tested 302 Tasks! - AI WITH Rithesh (3K)NVIDIA Dory 4B Is WILD… Looped Transformer, Same Memory? - AI WITH Rithesh (18K)ComfyUI Models Explained: Download, Folders and Model Loaders - pixaroma (10K)Flux 3 Klein in ComfyUI | Generate, Edit & Combine Images (Ep39) - pixaroma (30K)Give Me 35 Minutes, I'll Make LLM Pre-training Click Forever - Zachary Huang (5K)Give Me 1 Hour, I'll Make Multimodal Deep Learning Click Forever - Zachary Huang (7K)Give Me 1 Hour, I'll Make Deep Learning Click Forever - Zachary Huang (39K)Run Your AI Software Factory Like a Real Engineering Org (Here's How) - Cole Medin (13K)AI Makes Us WAY Faster but Fries Our Focus (Here's What I Do About It) - Cole Medin (25K)Prompt Injection Is Just the Start of AI Agent Security (Here's What You Need) - Cole Medin (2K)Let Your AI Agent Do ITS OWN Harness Engineering (Here's How) - Cole Medin (25K)Copilot Can Use Your Computer for You #ai #coding #githubcopilot - Burke Holland (5K)I tested four AI strategies to find the cheapest one - Burke Holland (8K)Llama: Local AI on Your Mac - Hugging Face (6K)Automate CSV to Parquet with Buckets, Webhooks & Jobs - Hugging Face (2K)I Turned Structured Output Into a Game 🎮 - Hugging Face (1K)Local AI 202: Get Massive Speed-ups with Speculative Decoding - Hugging Face (14K)Pi Durable: AI agents that survive crashes - Hugging Face (3K)Pi Durable EXPLAINED: Build Always-On Agents with Local Models - Hugging Face (55K)Gemini 4 Argon Is Not What I Expected - Two Minute Papers (229K)I Made $11,000 Selling Stickers: Then Built AI to Scale It. - Build w Agents - Aggregate Intellect - AI.SCIENCE (75)From Click to Booking: I Built the Tracking Most Businesses Are Missing - Build w Agents - Aggregate Intellect - AI.SCIENCE (116)This AI Doesn’t Just Remind You, It Knows When to Back Off. - Build w Agents - Aggregate Intellect - AI.SCIENCE (140)The Boring Version of Continual Learning - LangChain (3K)How Deep Agents Manages Context: Offloading and Summarization - LangChain (7K)Podium's Playbook for Scaling AI Agents Org-Wide - LangChain (2K)What is LangChain Managed Deep Agents? - LangChain (6K)Density functional theory meets AI: traditional DFT vs a pretrained model - DigitalSreeni (1K)15 - Spatial Pharmacokinetics: Measuring Drug Diffusion in Tumor Tissue - DigitalSreeni (908)Voice Agents That Interrupt You #Shorts - Trelis Research (116)Dialt: Deep Control Over Voice Agent Behavior #Shorts - Trelis Research (230)Voice Agents That Interrupt You - Trelis Research (375)Background Speaker Suppression in Dialt #Shorts - Trelis Research (318)Dialt Spot: Voice AI at $1.99 per Hour During Quieter Periods - Trelis Research (349)Dialt Launch Webinar: Voice AI Demo, WorkStuff AI & Q&A - Trelis Research (425)Detta and Phonon-2: Fast Dictation on a Mac - Trelis Research (359)Dialt: Deep Control Over How Voice Agents Behave and Sound - Trelis Research (368)Background Speaker Suppression in Dialt: A Live Demo - Trelis Research (407)Two Ways to Switch Languages with Voice Assistants | Dialt #Shorts - Trelis Research (228)Learn to build better AI Agents through better context - DeepLearningAI (7K)Introducing OCRia-1, the world’s best document parsing model - LlamaIndex (3K)Building an Auto Total-Loss Claims Workflow with LlamaParse - LlamaIndex (312)Answering your questions about AI, agents, and tech jobs (Q&A) - Dave Ebbelaar (9K)🔊 Voice design and voice replication with Gemini 3.8 Flash TTS - Google for Developers (9K)Design and replicate voices with Gemini 3.8 Flash TTS - Google for Developers (30K)Triple BAM! Live with Jon Krohn, Josh Starmer, and Luis Serrano! - StatQuest with Josh Starmer (21K)

Tema yoğunluğu

Ajan ve Sistem Güvenliği6Sesli Yapay Zeka ve Sesli Ajanlar17Görüntü ve Video Üretimi5Belge Anlama ve OCR6Kimi ve Moonshot8DeepSeek7Gemini ve Google Modelleri6Difüzyon Dil Modelleri1Yerel Yapay Zeka ve Gizlilik11Sam Altman ve OpenAI Gündemi8İş Dünyasında Yapay Zeka15Açık Modeller ve Verimlilik3Kodlama Ajanları ve Ajan Altyapısı13Öğrenme, Bilim ve Araştırma11Diğer Öne Çıkanlar5

En çok izlenen kanallar

Kai1.1M8 v
Matthew Berman693K14 v
OpenAI366K15 v
Prompt Engineering281K7 v
Two Minute Papers229K1 v
Sam Witteveen95K3 v

01Ajan ve Sistem Güvenliği6 video · 52K izlenme

Beş gün içinde PHP'den Node.js'e, GitLab'dan PyPI zehirlenmesine kadar neredeyse her geliştirici katmanı yamalandı; ayrıca OpenCode API anahtarı hırsızlığı ve Microsoft'un Project Perception ile kurumsal ajan güvenliği hamlesi, ajanların yeni saldırı yüzeyi olduğunu gösterdi.

43K20 EkiKaiOpenCode Got Hacked Badly (My API Keys Got Drained)4K23 EkiPrompt EngineeringThe Internet Is on Fire (Patch These Now)2K20 EkiCole MedinPrompt Injection Is Just the Start of AI Agent Security (Here's What You Need)70220 EkiAI WITH RitheshCloudflare Clairic: The $0.10 AI Extraction Model Tested Thoroughly!57922 EkiPrompt Engineer 48Mercury 3 Is INSANE: I Tested 13 Live Examples (+ 100 MILLION FREE Tokens)29223 EkiPrompt Engineer 48PHP Is Broken?! CVE-2026-109618 and 109619 Explained, Patch Today

02Sesli Yapay Zeka ve Sesli Ajanlar17 video · 183K izlenme

Trelis'in Dialt lansmanı (kesintili konuşma, arka plan bastırma, saatte 1,99 dolara varan Spot fiyatlandırma), Google'ın Gemini 3.8 Flash TTS ile 100+ dilde ses tasarımı/çoğaltması ve Block'un insan sesiyle eğitilen Dessa'sı sesli ajanları demo olmaktan çıkarıp üretim maliyetlerine taşıdı.

73K20 EkiOpenAIBuild a Better Shopping Experience with OpenAI | Shopping Agent Demo65K19 EkiOpenAICooking with Codex: Ep. 130K21 EkiGoogle for DevelopersDesign and replicate voices with Gemini 3.8 Flash TTS9K21 EkiGoogle for Developers🔊 Voice design and voice replication with Gemini 3.8 Flash TTS2K24 EkiPrompt Engineer 48Block's Dessa AI Voice Is CRAZY: It LAUGHS Mid-Sentence (I Played the Real Demos)78221 EkiAI WITH RitheshGradium Phonon Is WILD… Tiny Open-Weight TTS Model Tested Thoroughly!42521 EkiTrelis ResearchDialt Launch Webinar: Voice AI Demo, WorkStuff AI & Q&A40719 EkiTrelis ResearchBackground Speaker Suppression in Dialt: A Live Demo37523 EkiTrelis ResearchVoice Agents That Interrupt You36820 EkiTrelis ResearchDialt: Deep Control Over How Voice Agents Behave and Sound35921 EkiTrelis ResearchDetta and Phonon-2: Fast Dictation on a Mac34922 EkiTrelis ResearchDialt Spot: Voice AI at $1.99 per Hour During Quieter Periods31823 EkiTrelis ResearchBackground Speaker Suppression in Dialt #Shorts25324 EkiPrompt Engineer 48AI Avatars IGNORE You?! I Tested the CRAZY New StreamEval Benchmark (RTX 4090)23024 EkiTrelis ResearchDialt: Deep Control Over Voice Agent Behavior #Shorts22819 EkiTrelis ResearchTwo Ways to Switch Languages with Voice Assistants | Dialt #Shorts11624 EkiTrelis ResearchVoice Agents That Interrupt You #Shorts

03Görüntü ve Video Üretimi5 video · 55K izlenme

Kling 4.0 tek klipte 30 saniyeye çıkarak AI videoyu uzun plana taşıdı; ComfyUI/Flux 3 Klein iş akışları ve ChatGPT headshot özelliği görsel üretimin günlük araçlara gömüldüğünü gösteriyor.

30K20 EkipixaromaFlux 3 Klein in ComfyUI | Generate, Edit & Combine Images (Ep39)11K20 EkiOpenAIYou need a better headshot.10K22 EkipixaromaComfyUI Models Explained: Download, Folders and Model Loaders4K22 EkiHowfinityHow To Use Google Flow (Complete Tutorial)48724 EkiPrompt Engineer 48Kling 4.0 Is Here… AI Video Just Got SCARY Good

04Belge Anlama ve OCR6 video · 234K izlenme

LlamaIndex'in OCRia-1'i ve açık kaynak OXR, tablo/grafik ağırlıklı belge ayrıştırmada yeni rekorları kovalarken RAG-Anything'in kendi makalesini okuma testi, çok modlu RAG'ın olgunlaştığını kanıtladı.

229K19 EkiTwo Minute PapersGemini 4 Argon Is Not What I Expected3K22 EkiLlamaIndexIntroducing OCRia-1, the world’s best document parsing model1K21 EkiHowfinityHow to Analyze PDFs and Long Documents with Claude58823 EkiPrompt Engineer 48RAG-Anything Read Its OWN Paper… and the Results Are WILD31221 EkiLlamaIndexBuilding an Auto Total-Loss Claims Workflow with LlamaParse8225 EkiPrompt Engineer 48AWESOME or Overhyped? OXR Document Parsing, Tested With My Own Ugly PDFs

05Kimi ve Moonshot8 video · 181K izlenme

Kimi K3 Arena Frontend Code'da birinciliği aldı, K3.1 Pro ise 2,8 trilyon parametreli MoE mimarisi ve 1M token bağlamla Terminal-Bench'te Opus 5.5'in 1 puan gerisine, GPT-6 Astra'nın önüne yerleşti — üstelik üçte bir maliyetle. Açık ağırlıklar henüz çıkmadı, benchmark iddiaları tartışmalı.

117K24 EkiMatthew BermanThey actually did it... (Kimi K3.1)36K23 EkiPrompt EngineeringKimi K3.1 Pro: Open Weights, Almost Opus16K23 EkiMatthew BermanTeam AMA (and Kimi K3.1 Testing!!)7K24 EkiMatthew BermanKimi K3.1 Invented the Benchmark Scores2K23 EkiAI WITH RitheshKimi K3.1 Could Be INSANE…RUMOURS and LEAKS1K24 EkiAI WITH RitheshKimi K3.1 Pro vs Opus 5.5 vs GPT-6 Astra: WHO WINS? Real Demos!43023 EkiPrompt Engineer 48Kimi K3 Just Went #1 on Frontend Code: The CRAZY Demos, Prices and What It Still Loses35724 EkiAI WITH RitheshKimi K3.1 Pro Open-Weights Model Is INSANE! TESTED THOROUGHLY ! 92.2% on 302 Tests!

06DeepSeek7 video · 178K izlenme

DeepSeek, API'sinin arkasındaki DeepInfer çıkarım motorunu (günde 155 trilyon token, %96 KV-cache isabeti) ve V4.1 Flash 1017'yi açtı; düşünce token'larını üçte bir azaltan güncelleme Arena WebDev'de açık modellerde birinciliğe taşıdı.

67K24 EkiKaiDeepSeek Is Free Now but What’s the Catch?41K19 EkiPrompt EngineeringDeepSeek Just Made Reasoning WAY More Efficient!33K21 EkiPrompt EngineeringDeepSeek Just Dropped a MASSIVE AI Breakthrough!23K18 EkiMatthew BermanHow DeepSeek Uses Less Memory9K20 EkiAI WITH RitheshMercury 3 Is WILD… 1,000+ Tokens/sec Diffusion LLM Tested Thoroughly!3K19 EkiAI WITH RitheshMiniMax M3.1 Flash JUST DROPPED… I Tested 302 Tasks!2K21 EkiAI WITH RitheshDeepSeek Just OPEN SOURCED Its AI Engine… Here’s How It REALLY Works!

07Gemini ve Google Modelleri6 video · 481K izlenme

Gemini 4 Argon beklentileri karşılamadı, Carbon sızıntıları konuşuldu; Google'ın TimesFM 3.0'ı eğitim gerektirmeden zaman serisi tahmini yaparken Harness², modeli eğitmeden ajanın iskeletini (prompt, skill, alt-ajan) kendi kendini iyileştirmesini sağlayan araştırma çerçevesi olarak öne çıktı.

294K19 EkiKaiNCM Is this the End of LLMs?83K18 EkiKaiHow I Got Frontier Performance on 16GB VRAM (Qwen 3.8 27B)44K24 EkiPrompt EngineeringHarness² Explained: Self-Improving Agents Without Training40K23 EkiKaiGoogle's Harness² Is This the End of Harness Engineering?17K22 EkiPrompt EngineeringGoogle's GPT-3 Moment for Time Series & Forecasting (TimesFM 3.0)2K23 EkiAI WITH RitheshGemini 4 Carbon COULD BE INSANE… Leaked Demos & Rumours Investigated!

08Difüzyon Dil Modelleri1 video · 22K izlenme

Inception'ın Mercury 3'ü 1.000+ token/sn iddiasıyla difüzyon LLM'leri yeniden gündeme taşıdı; bağımsız testler uçtan uca 258-416 token/sn ölçse de hız/bedava token stratejisi dikkat çekici.

22K20 EkiSam WitteveenDo Diffusion LLMs Scale? Testing Mercury 3 + Decide

09Yerel Yapay Zeka ve Gizlilik11 video · 680K izlenme

Haftanın en çok izlenen videosu yerel yapay zekanın gizlilik varsayımını yıktı; aynı zamanda llama.cpp'nin 125B MoE modelleri 4x hızlandırması, Mac için resmi Llama uygulaması ve spekülatif kodlama eğitimleri yerel tarafın ivmesini gösteriyor.

397K21 EkiKaiYour Local AI Is Not as Private as You Think74K21 EkiKaiI Tested Every Best Local AI For GPU (8GB TO 24GB VRAM)59K23 EkiMatthew BermanWho is local AI really for?55K19 EkiHugging FacePi Durable EXPLAINED: Build Always-On Agents with Local Models53K22 EkiKaillama.cpp Just Made 125B MoE Models 4× Faster!18K18 EkiAI WITH RitheshNVIDIA Dory 4B Is WILD… Looped Transformer, Same Memory?14K20 EkiHugging FaceLocal AI 202: Get Massive Speed-ups with Speculative Decoding6K23 EkiHugging FaceLlama: Local AI on Your Mac3K19 EkiHugging FacePi Durable: AI agents that survive crashes69923 EkiPrompt Engineer 48I Ran Flint, the CRAZY New 40B Coding Model, WITHOUT a Mac. Did It Fix My Buggy Repo?28425 EkiPrompt Engineer 48XERJ's CRAZY Local Search for AI Agents: I Tested It, It Broke, Then It Got AWESOME

10Sam Altman ve OpenAI Gündemi8 video · 430K izlenme

Matthew Berman'ın Altman röportaj serisi RSI (kendi kendini geliştiren yapay zeka), bilince, yeni donanım ihtiyacına ve kodcu talebinin neden kaybolmadığına odaklandı — OpenAI'ın yönü hakkında en çok konuşulan birincil kaynak haftaydı.

355K21 EkiMatthew BermanSam Altman: RSI, AI Consciousness, OpenAI's New Hardware, Compute, and more!22K21 EkiMatthew BermanSam Altman on Why Demand for Coders Hasn't Disappeared17K21 EkiMatthew BermanSam Altman on Why AI Needs a New Kind of Hardware16K21 EkiMatthew BermanSam Altman on What Personal AI Still Needs10K21 EkiMatthew BermanSam Altman on Why AI Hasn't Made Us Less Busy7K22 EkiMatthew BermanSam Altman on AI’s Discovery Era2K23 EkiMatthew BermanSam Altman on Why Your Most Important Advice Should Come From People65025 EkiMatthew BermanSam Altman on What People Still Bring to an AI World

11İş Dünyasında Yapay Zeka15 video · 128K izlenme

Uber'ın 30 bin çalışanına ChatGPT Work'ü açması, AT&T'nin hukuk ve Thomson Reuters'ın güven odaklı hukuk yapay zekası iş akışları kurumsal benimsemenin ölçeğini; sticker satışıyla doğrulanıp ajanlarla ölçeklenen e-ticaret ve satış hunisi analizi ise KOBİ tarafını temsil etti.

27K21 EkiOpenAIV7 cuts costs 78% while boosting accuracy with GPT‑5.6 Luna26K22 EkiOpenAIUber rolls out ChatGPT Work to 30,000+ employees15K23 EkiOpenAIHow Shield uses OpenAI’s Astra to support IT teams14K23 EkiOpenAIHow Realpage. builds AI training with ChatGPT Work13K23 EkiOpenAIHow AT&T is building AI-powered legal workflows with OpenAI10K23 EkiOpenAIHow Benefit Street Partners is building financial analysis tools with Codex8K19 EkiOpenAIThomson Reuters on Building Trust in Legal AI | OpenAI5K23 EkiOpenAIFrom notes to AI training with ChatGPT Work | Realpage.4K23 EkiOpenAIAI agents for everyday IT fixes | Shield and OpenAI2K19 EkiLangChainPodium's Playbook for Scaling AI Agents Org-Wide2K23 EkiOpenAIFrom analyst to builder with Codex | Benefit Street Partners2K19 EkiHowfinityHubSpot CRM Tutorial For Beginners1K20 EkiHowfinityHow To Sell On TikTok Shop For Beginners (2026 Guide)11623 EkiBuild w Agents - Aggregate Intellect - AI.SCIENCEFrom Click to Booking: I Built the Tracking Most Businesses Are Missing7524 EkiBuild w Agents - Aggregate Intellect - AI.SCIENCEI Made $11,000 Selling Stickers: Then Built AI to Scale It.

12Açık Modeller ve Verimlilik3 video · 17K izlenme

Meta'nın 816B'lik açık ağırlıklı Muse Spark Open'ı, NVIDIA Dory 4B'nin döngülü transformer'ı, MiniMax M3.1 ve 8GB GPU'da testlenen Spark-X2.6-4B'nin 'düşünme kadranı' açık modellerde mimari verimlilik yarışının hızlandığını gösterdi.

11K19 EkiSam WitteveenMeta's New Models & Agent6K22 EkiAI WITH RitheshRecursive Flint 40B Coding Model: Can It Code on Just 24GB RAM?!23524 EkiPrompt Engineer 48Spark-X2.6-4B's CRAZY Thinking Dial: 180 Tests on an 8GB GPU

13Kodlama Ajanları ve Ajan Altyapısı13 video · 308K izlenme

Recursive'ın RSI ile eğittiği 40B'lik Flint kodlama modeli haftanın merak uyandıran çıkışıydı; LangChain'in Managed Deep Agents'ı ve bağlam çürümesine karşı offloading/özetleme teknikleri ile Cole Medin'in kendi iskeletini kendi iyileştiren ajanı, ajan altyapısının olgunlaştığını gösterdi.

106K21 EkiPrompt EngineeringClaude Duet: Anthropic's Proactive Agent62K23 EkiSam WitteveenFlint - The First RSI Coding Model?54K23 EkiOpenAIHunt bugs with your dot25K20 EkiCole MedinLet Your AI Agent Do ITS OWN Harness Engineering (Here's How)14K20 EkiThe Gray CatI Tested 4 Codex Frameworks on the Same Real Feature13K23 EkiCole MedinRun Your AI Software Factory Like a Real Engineering Org (Here's How)8K19 EkiBurke HollandI tested four AI strategies to find the cheapest one7K21 EkiLangChainHow Deep Agents Manages Context: Offloading and Summarization7K20 EkiDeepLearningAILearn to build better AI Agents through better context6K19 EkiLangChainWhat is LangChain Managed Deep Agents?5K20 EkiBurke HollandCopilot Can Use Your Computer for You #ai #coding #githubcopilot34923 EkiPrompt Engineer 48This AI Agent Builds Its OWN Tools to Finish Tasks! (INSANE Agent S4 Test)32824 EkiPrompt Engineer 48Microsoft's Agent World Is INSANE. I Ran It. The Agent FAILED (0/3)

14Öğrenme, Bilim ve Araştırma11 video · 81K izlenme

Zachary Huang'ın bir saatlik derin öğrenme maratonları ve StatQuest canlı yayını eğitim tarafını taşırken, Paper2Agent'ın her makaleyi MCP sunucusuna dönüştürmesi ve OlmoEarth uydu yerleştirmeleri araştırmanın yapay zekayla temas noktaları oldu.

39K18 EkiZachary HuangGive Me 1 Hour, I'll Make Deep Learning Click Forever21K23 EkiStatQuest with Josh StarmerTriple BAM! Live with Jon Krohn, Josh Starmer, and Luis Serrano!7K21 EkiZachary HuangGive Me 1 Hour, I'll Make Multimodal Deep Learning Click Forever5K24 EkiZachary HuangGive Me 35 Minutes, I'll Make LLM Pre-training Click Forever3K23 EkiLangChainThe Boring Version of Continual Learning2K22 EkiHugging FaceAutomate CSV to Parquet with Buckets, Webhooks & Jobs1K21 EkiHugging FaceI Turned Structured Output Into a Game 🎮1K24 EkiDigitalSreeniDensity functional theory meets AI: traditional DFT vs a pretrained model90822 EkiDigitalSreeni15 - Spatial Pharmacokinetics: Measuring Drug Diffusion in Tumor Tissue27524 EkiPrompt Engineer 48OlmoEarth v1.3: INSANE Satellite AI or Fooled by WINTER? I Tested It on My RTX 406024824 EkiPrompt Engineer 48This Nature Paper Turns Any Research Paper Into an AI Agent - INSANE

15Diğer Öne Çıkanlar5 video · 113K izlenme

Messi'nin satranç dersi, robot vücut koruma demosu, paralel ajan odaklanmasını anlatan itiraflar ve 'en ucuz AI stratejisi' karşılaştırması haftanın renkli ama düşündürücü kalanlarıydı: hız artarken dikkat ve insan rolü yeni darboğaz.

40K21 EkiMatthew BermanActual Robot Bodyguard39K22 EkiOpenAIMessi asks ChatGPT to teach him a chess opening25K22 EkiCole MedinAI Makes Us WAY Faster but Fries Our Focus (Here's What I Do About It)9K20 EkiDave EbbelaarAnswering your questions about AI, agents, and tech jobs (Q&A)14022 EkiBuild w Agents - Aggregate Intellect - AI.SCIENCEThis AI Doesn’t Just Remind You, It Knows When to Back Off.
Bu rapor Murat Karakaya Akademi tarafından derlenmiştir · Veri: YouTube RSS (canlı, API key yok) · pencere: 2026-10-18 .. 2026-10-25 (UTC, çalışma anından hesaplanır) · kanal başı max 15 video · Prompt Engineer 48 ve OpenAI beslemeleri RSS 15-video sınırına takıldı (partial); bu kanalların pencerenin başındaki tüm videoları görülemeyebilir. Pencere: 18-25 Ekim 2026 (UTC), 43 kanaldan 24'ü aktif, 122 video. · render.py @ 25.10.2026 10:00 UTC