Top Tweets for #Deepresearch
La investigación profunda de la IA tarda minutos porque trabaja en equipo: un líder reparte la pregunta y varios subagentes buscan en paralelo. Demo real: 4 subagentes, 17 fuentes, 12.1 s en paralelo vs 42.9 s en fila, y 48 419 tokens vs 260 de un chat. #IA #DeepResearch
A URL is helpful—but it's not always necessary.
If you only have a photo or an idea, our Deep Research service can start from there. 🔎
https://t.co/r6b4IdUBla
#JapanTreasureHunter #DeepResearch #Japan
Tools for CS193T: Thinking with AI
#HBCU #AI #ML
@Anthropic #ClaudeChat #ClaudeCowork #ClaudeCode
@Google #Gemini #NotebookGemini #NotebookLM #DeepResearch #DataInsights #CodeAssist
@OpenAI #ChatGPT #ChatGPTWork #ChatGPTEdu #Codex
#opensource #Deepseek #Mistral #Llama
#DeepResearch Equity fundholders see $JBI stock decline continue.
Deep Research、調べてもらって終わりになってない?🔎🤖
調査設計→情報整理→分析→レポート化まで、ChatGPTを“調査の道具”として使う流れを学べる本。
AIでリサーチの質と速さを両立したい人に👇
https://t.co/QYJjUo4I6T
#ChatGPT #DeepResearch #PR
𝐏𝐞𝐫𝐩𝐥𝐞𝐱𝐢𝐭𝐲 𝐀𝐈 क्या है और कैसे इस्तेमाल करें?
Perplexity AI की मदद से इंटरनेट पर जानकारी खोजना और समझना आसान हो जाता है। जानिए इसके Pro Search, Deep Research और दूसरे उपयोगी फीचर्स के बारे में।
#PerplexityAI #PerplexityProSearch #DeepResearch #PerplexityVsChatGPT #AISearch #AITools #ArtificialIntelligence #TechNews #AnalyticsInsightNews
पढ़ें पूरी खबर 👇
https://t.co/BE1f5nYXPy

From Vane and Khoj to STORM and GPT Researcher, open-source AI agents are transforming how we research, organise knowledge and make decisions.
The future belongs to those who learn, experiment and build with AI.
#AI #DeepResearch #OpenAi
@bindureddy Hope so.. because #Gemini 4 #Argon (with & without #DeepResearch) has ISSUES !!..👇 NTM Google is pre-empting when busy & capping the #TPU usage and enforcing TIMEOUTS (aka, points to a shortage of TPUs), etc.. as Argon was nice enough to spill the beans on with is ! 😆 🤦♂️👇📃

🤯⁉️FIGURED IT OUT.. with Gemini Argon's honest feedback 👇.. see the responses attached and REASONS WHY Gemini FAILS/ HANGS (with Zero indication.. would stay that way for days even after browser refresh!?). The skinny & Guidelines/ #Recommendations to follow 👇 :
#Gemini (Argon included) #Lesson #1:
Google GCP does NOT allow any inference job tasks to run more than ~15 minutes on #TPU.. aka, if your job/task (the operation/iteration it's working on-TPU with) takes too more than that.. it's #KILLED with NO session messages/warning, will let you wait for a month🤦♂️. Ironically, this previously was NOT the case with Gemini 3.x Pro DeepResearch, as we've had a LOT of great success previously even with 30-40min tasks producing amazing output (unless we got lucky, assume Google is locking down and limiting resource utilization so they can scale.. NTM, require mega TPU's to pre-train the next 2 frontier model revs concurrently).
#Lesson #2: , they recommend that you to break down research, coding, etc.. tasks into multiple sequential steps of iteration that the session can tackle 1-by-1, serially.. AND each within <15 mins.. which brings us to the .. UNTIL :
#Lesson #3: (that most are aware of, #Hallucinations of context.. and WHY we were hoping #ULTRA could save the day, especially coupled with long #context of #Argon.. + DR, but not so, for now ☹️) :
You can serialize many sequential steps in 1 chat session.. however, that is, until (and eventually always does) #ROT your #Context (aka, you run out of/ fill up your available KV-Cache for your session).. then the prior work done gets compressed.. & Gemini starts #Hallucinating badly, aka, DROPPING your code (like it dropped several functions from code within 1 run).. another session, it copied the SAME function 13 times into the same program, etc..
#Lesson #4: don't do more than very small/basic code refactoring in #DeepResearch, and if you need to do real deep research for next Gen feature enhancements, future state architectures, detailed configurations.. Google recommends doing SEPARATELY, and moreso, push all coding OUTSIDE of deep research. We've been told to do inside #AntiGravity, but haven't gone back since our last debacles.. where antigravity #rewrote all of our code (little by little) #eroded slowly and degrated a massive code base this summer, but only a few lines here, there.. small function refactors each iteration.. until the code DID NOT RESEMBLE anything we were familiar with and sever functionality had been lost (even though we had #directives to leave existing functionality IN-TACT, AS-IS, rewrote/ crippled detailed prompts we had customized and fine-tuned..), what has brought our hopes up and latest tests with Gemini #Argon #DeepResearch.
Hence, we are VERY hesitant to hand over the reigns to blindly #Vibe-Code with #AntiGravity or any coding harness and an agentic team of minions (just yet 😉). Yes, this is the direction.. to NOT look at code generated, no longer use IDE's.. to feel confident in the #tools knowing what's best.. but c'mon folks, until we truely get #RSI and AGI to the point of hallucination-FREE work (our hope for #Ultra + #Argon + DR), where our DeepResearch and coding efforts aren't KILLED after 15 minutes on #TPU.. could be time to pause, re-assess.. while embracing the best of what's available today.
#Lesson #5: We do know.. that if people don't give honest #feedback like this (unfortunately no other forum we've been invited to, with support not quite a priority, 🧵☝️).. problems will NOT be resolved expeditiously (if ever.. take the google drive and chat interface massive holes vs. prioritizing model dev), understandable, but at what adoption cost?☝️.
We all are aware that the capabilities that bring the most eyeballs to the largest # of users is what the hope is.. will generate revenue (cat/trump.. video gens, games, simulations/modeling.. fun, but at what cost to curing cancer, designing the next productive research, aka Enterprise ADOPTION?). While features/ functions that generate #profitability (more paying accounts, token/API revenue..), will be the features that are worked on and drop on Product Manger's roadmaps.. are necessary priorities to grow and advance, we can only hope that some of the rest, being overlooked (and causing a loss of market-share from some real business/enterprise users).. aren't being sacrificed & displaced vs. pre-empting the free/ consumer fun at lower-tiers (though both would be great).
Go @Google !.. we have faith and can see the light, just that this last week with Argon has been a bit more painful than expected, possibly our own expectations got ahead of our reality, TBD.
@demishassabis @sundarpichai @GeminiApp @joshwoodward @koraykv @ThomasOrTK
GeminiのDeep Researchで、調べる先を自分のGmail・ドライブ・チャットに切り替え、先週のメールから今週の段取りメモを作らせた手順を、依頼文つきで書きました。1回目の失敗も載せています。
#Gemini #DeepResearch #業務効率化 #AI活用 https://t.co/OFMrRpOcYY
Deep research, made smarter with AI. 🔍 Discover the best AI research tools to find reliable sources, analyze complex topics, and save time.
Read more: https://t.co/3qhVuplMlK
#AI #DeepResearch #AITools #Technology #TechnologyMoment

@jun_song We desperately need Gemini 4 pro for #DeepResearch. Enterprises arent making video's, games, nor 3D immersion simulations they showcase.
Ironically, Gemini's DR reports are still near the best long form & IF they have solved long horizon RSI, we can only hope, cuz today📉🤏👇

Heute räume ich ein tiefsitzendes Missverständnis aus der Welt: Nein, die KI nimmt uns die Recherchearbeit nicht komplett ab – auch mit der #DeepResearch nicht. Ein paar elementare Überlegungen, wie Informationsbeschaffung und Schreiben zusammenhängen.
https://t.co/v14YNSbVic
@Google @GoogleDeepMind GREAT results! 👏
.. BUT.. Lots of users on the #Ultra plan are asking, 🤔 hmm.. WHY are we paying ??$200/mo, yet still can only use #Gemini 3.1 #Pro for real, long-form #DeepResearch !? 🤦♂️
👇note記事を書きました♪
☑️ 知りたかったのは「答え」じゃなかった──「このテーマの世界って、実際どうなってるの?」をDeep Researchに調べてもらう|
#DeepResearch
https://t.co/Bu84pVg46C
👇Deep Researchで“テーマの世界“を見るコツ
“答え”じゃなくって、“どうなってるんかなぁ”が知りたい時は…
#DeepResearch

I built an agent skill that helps your agent do deep research from primary sources, not only from search results. Here is how it works.
#AIAgents #DeepResearch #AgentSkills https://t.co/bgNu4sPMz9
Facebook is a DARPA program called Lifelog. Why do we pretend that Zuck is in charge of anything?
#InvestigativeResearch #DeepResearch #FOIA #FOIAFiles #Declassified #DeclassifiedDocuments #GovernmentDocuments #CIAFiles #FBIFiles #PentagonFiles #MilitaryFiles #GovernmentRecords

Last Seen Hashtags on Sotwe
latina
Seen from United States
92i()**********************************************************************
Seen from France
boyfarts pov
Seen from Portugal
modelindo
Seen from Germany
oliveglass
Seen from Turkey
รับงานบางใหญ่
Seen from Thailand
Bélico
Seen from United States
goonbait
Seen from Brazil
randomchat
Seen from United States
yozgatgay
Seen from Turkey
Most Popular Users

Elon Musk 
@elonmusk
240.8M followers

Barack Obama 
@barackobama
118.7M followers

Cristiano Ronaldo 
@cristiano
114.5M followers

Donald J. Trump 
@realdonaldtrump
111.9M followers

Narendra Modi 
@narendramodi
107M followers

Rihanna 
@rihanna
98.7M followers

NASA 
@nasa
92.2M followers

Justin Bieber 
@justinbieber
91.8M followers

KATY PERRY 
@katyperry
90M followers

Taylor Swift 
@taylorswift13
84M followers

Lady Gaga 
@ladygaga
75.5M followers

Virat Kohli 
@imvkohli
73.4M followers

Kim Kardashian 
@kimkardashian
70.8M followers

YouTube 
@youtube
68.7M followers

Neymar Jr 
@neymarjr
66.4M followers

Bill Gates 
@billgates
65.2M followers

Selena Gomez 
@selenagomez
63.1M followers

The Ellen Show
@theellenshow
62.2M followers

CNN 
@cnn
61.7M followers

X 
@x
60.6M followers


























