- Published on
科技热门推文 - 2026-09-08
- Authors
- Name
- geeknotes
今日科技动态：人工智能在科研与工程领域不断拓展的作用成为讨论焦点。相关推文聚焦于OpenAI所宣称的自主“研究实习生”里程碑，以及工程师协同调度大批编程智能体的报道。尽管这些能力引发热切期待，但围绕人类控制权、人工智能相关传播表述以及相互矛盾的就业前景预测，争论仍在持续。与此同时，空间规划或将塑造人工智能视频发展的下一阶段；免费工具与工程课程有望降低创业者的入门门槛；瑞典—印度技术与人工智能走廊则释放出国际合作日益深化的信号。
1. VaibhavSisinty (Group Score: 240.0 | Individual: 37.8)
Cluster: 9 tweets | Engagement: 54 (Avg: 70) | Type: Tech
OpenAI says it has reached a goal it set a year ago: an AI that can do real research on its own. It calls the system a "research intern," able to take on tasks that would take a skilled human several days.
The next target on its roadmap is a fully automated AI researcher by March 2028.
Most people will read that as a product update. It's not. It's the moment the curve bent.
Here's the number that proves it.
For every one workday its human researchers put in, OpenAI now gets 3.1 workdays of AI effort out.
As recently as June, AI was doing less work than the humans. In weeks, it crossed over and pulled ahead.
Sit with what that means.
The AI is now writing more of the code. Running more of the experiments. Debugging the infrastructure. Answering the questions researchers used to ask each other.
Every experiment it runs teaches the next model to run experiments better. Every bug it fixes sharpens the system that fixes bugs.
Every task it completes trains the thing that completes tasks.
The work that builds AI is now being done by AI.
That's the quiet line everyone skipped. One of OpenAI's own researchers said it out loud: recursive self-improvement has already started.
For years that phrase lived in safety papers and sci-fi. This week it showed up in a company blog post, stated as a fact, backed with charts.
So ask the real question.
If AI is already doing three days of research for every one a human does, what happens when it's ten? When it's a hundred? When the fully automated researcher lands in 2028 and the humans are no longer the bottleneck?
Progress stops moving in a straight line the moment the system starts improving itself.
In the old world, humans built the models. In the new one, the models are starting to build themselves.\n\nQT @kliu128: Today we're releasing data on models accelerating research at OpenAI.
Recursive self-improvement could be the most important contributor to AI capabilities over the next few years, but by default it will only be seen inside a few frontier AI labs. Being transparent is more urgent than ever, so we can inform the public discussion on whether and how to pace model development. I ask other AI companies to do the same.
See 8 related tweets
- @Vtrivedy10: I see these numbers, all i think is:
- give the world’s most creative & knowledgeable researchers/...
@aakashgupta: OpenAI's AI agents now put in 3 days of work for every 1 day of human labor in the company's researc...
@rickasaurus: RT @deredleritt3r: Key take-aways for me from today's OpenAI announcements:
OpenAI "making strong...
@vinhnx: RT @kliu128: Today we're releasing data on models accelerating research at OpenAI.
Recursive self-...
- @AISafetyMemes: Suddenly, MOST of the labor happening inside OpenAI is now done by the AIs themselves, building th...
2. rickasaurus (Group Score: 226.2 | Individual: 43.5)
Cluster: 7 tweets | Engagement: 2320 (Avg: 377) | Type: Tech
RT @merettm: I wrote about the state of AI, why I’m concerned about the next few years, and the choices we need to make to keep the future in humanity’s hands.
An Alien Mind: https://t.co/FeIfWNe0UE
See 6 related tweets
- @AISafetyMemes: OPENAI CHIEF SCIENTIST: WE MUST SLOW DOWN AI
"The idea of racing forward at all costs seems absurd ...
- @gdb: an important essay on the state of AI and the choices and challenges ahead: for openai, the field, a...
- @nathanbenaich: “Based on internal results, I have a strong expectation that this speed of progress could be sustain...
- @derrickcchoi: Definitely worth reading Jakub's essay in full.
We've seen AI help us make real progress in scienc...
- @Miles_Brundage: https://t.co/eoj7zzjhHY https://t.co/2qBDVU43IU\n\nQT @merettm: I wrote about the state of AI, why I...
3. theobearman (Group Score: 225.0 | Individual: 44.5)
Cluster: 12 tweets | Engagement: 8240 (Avg: 387) | Type: Tech
RT @JensenHuang: @ChaseLochmiller @OpenAI GPT-6 Astra, trained on ~100K+ NVIDIA Grace Blackwell NVLink72. From ChatGPT to o1 to Astra in 4 years.
AGI has arrived. Congratulations @OpenAI team.
400K GPUs coming online next.
See 11 related tweets
- @BullTheoryio: BREAKING: Nvidia CEO Jensen Huang says "AGI has arrived, congrats OpenAI team" after the release of ...
- @cryptopunk7213: can you imagine how powerful grok 5 is going to be when it launches in a few months?
elon’s one of ...
- @_NathanCalvin: Jensen says Astra was trained on ~100k Blackwells and we will in not too long see models trained on ...
- @PolymarketMoney: JUST IN: Nvidia CEO Jensen Huang declares “AGI has arrived” with GPT-6 Astra trained on 100K+ Grace ...
- @Grady_Booch: AGI has, in fact, not arrived.
Unless perhaps you mean we have reached peak Authentic Granular Igno...
4. NovaIAHQ (Group Score: 142.8 | Individual: 35.4)
Cluster: 8 tweets | Engagement: 167 (Avg: 61) | Type: Tech
I wonder if 3D is about to become a much bigger part of AI video than people expect.
Not because everyone suddenly needs to become a 3D artist.
But because spatial planning gives AI something prompts often struggle with: a defined world to move through.
GPT-6 Astra maps the scene and keeps movement within the Viewport. I then bring that setup into CapCut PC, render with Seedance 2.5, and finish the video in CapCut PC.
For AI filmmaking, that’s a workflow I’d actually want to experiment with.
#CapCutPC #GPT6Astra #AIVideo #Seedance25
See 7 related tweets
- @rb_walter_ai: RT @eva_morgan_ai: AI filmmaking starts to feel different when the process becomes repeatable. Inste...
- @rb_walter_ai: RT @victor_kane_ai: AI video used to feel like a process that started when you hit “generate.” But f...
- @OtakuMachine: RT @xetgepete: Here’s the part of AI video I think we don’t talk about enough:
What happens before ...
- @OtakuMachine: RT @EnzoSanchezIA: The more I experiment with AI video, the less interested I am in one-click genera...
- @JayBisen473370: RT @rb_walter_ai: 🚨 THE FUTURE OF AI DESIGN IS HERE! 🚨
3D Modeling & Cinematic Rendering in Seconds...
5. JayBisen473370 (Group Score: 114.2 | Individual: 29.4)
Cluster: 5 tweets | Engagement: 130 (Avg: 108) | Type: Tech
RT @novaXsignals: 12 AI TOOLS THAT MAKE A 1M 💸
You don’t need an expensive AI stack to build, create, research, and automate.
- https://t.co/3x8ezt3egJ — Research, write, code & solve complex tasks
- https://t.co/Yx88I70euJ — Build and deploy AI agent workflows
- https://t.co/DpMTTI8skI — Research markets, competitors & ideas
- https://t.co/rQd3dIXJ4P — Create AI-generated music
- https://t.co/oseego40j2 — Build brands and visual concepts with AI
- https://t.co/FYuOPbL8vd — Create AI-powered voice content
- NotebookLM — Turn messy sources into useful insights
- https://t.co/9n1rB1jB8y — Create faceless video content faster
- https://t.co/nqXIi7z2aE — Repurpose long-form content into fresh ideas
- Runway — Generate and edit professional videos
- Descript — Edit podcasts and videos like a document
- HeyGen — Create AI avatars and multilingual videos
Save this list 🔖& Follow:@novaXsignals
The cost of building with AI keeps getting lower.
See 4 related tweets
- @robertsmith_ai: RT @hey_Jessicaai: 13 POWERFUL AI TOOLS TO SAVE TIME, BOOST PRODUCTIVITY.
- @JayBisen473370: RT @salmakhatunn: 120 AI Tools That Will Redefine How You Work in 2025.🧵✨
- Ideas
- YOU
- Claude ...
- @JayBisen473370: RT @liam_holt7: 11 AI websites that feel like cheating 🤯
- https://t.co/ur0kQd8wb2 — create stunni...
- @AiwithDharmik: RT @AI_VerseX7: 💸 PAID vs FREE AI TOOLS 🔥
Jasper → Worldtune
InVision → Pixlr
ChatGPT-4 ...
6. jon_stokes (Group Score: 108.0 | Individual: 39.8)
Cluster: 5 tweets | Engagement: 988 (Avg: 163) | Type: Tech
RT @rdomenechv: The jobs apocalypse is postponed. An AI jobs boom is here
According to @TheEconomist, AI is actually proving to be a net job creator in the US, easily generating over 1M new positions (from data center construction to AI engineering) to offset back-office layoffs.
While routine admin and customer service roles face real disruption, the overall labor market is proving resilient.
See 4 related tweets
- @NathanpmYoung: Just imagine how well we’d be doing without tariffs.\n\nQT @rdomenechv: The jobs apocalypse is postp...
- @TrungTPhan: RT @TurnerNovak: The US has added over 1 million new jobs due to AI since 2023:
310k Blue collar:
...
- @LoganDobson: RT @JoshKraushaar: "Most people think AI is a job-killer. And yet somehow, this job-killer keeps stu...
- @TheEconomist: Some companies and workers are being severely disrupted by AI. But the technology is already creatin...
7. JayBisen473370 (Group Score: 102.8 | Individual: 45.6)
Cluster: 3 tweets | Engagement: 409 (Avg: 108) | Type: Tech
INSTEAD OF WATCHING NETFLIX TONIGHT, spend 2 hours learning a skill that could completely change how you work.
This Claude AI FULL COURSE is worth bookmarking.
It teaches you how to use Claude to:
→ Build websites & apps → Write and debug code → Automate repetitive tasks → Analyze files and data → Create AI-powered workflows → Research faster → Build useful projects without starting from scratch → Turn simple ideas into working solutions
And the best part?
You don’t need to be an AI expert.
You just need to understand how to give AI the right instructions and turn it into a tool that actually gets work done.
Most people are using AI like Google:
Ask a question → get an answer → move on.
The people getting the biggest advantage are using it differently:
Idea → AI → workflow → automation → output.
That shift can save hours every single week.
So tonight, you have two choices:
🍿 Spend 2 hours watching Netflix and forget most of it tomorrow.
🧠 Spend 2 hours learning Claude and wake up tomorrow with a skill you can actually use.
If you're a student, developer, creator, marketer, freelancer, founder, or simply someone who wants to stay ahead of AI…
WATCH THIS.
Bookmark it now.
You’ll thank yourself later.
And send it to someone who still thinks AI is just a chatbot.
#ClaudeAI #AI #ArtificialIntelligence #AITools #Automation #Productivity
See 2 related tweets
- @AnishJaitwar: INSTEAD OF WATCHING NETFLIX TONIGHT.
Spend 2 hours with this.
Claude AI FULL COURSE that teaches...
- @JayBisen473370: RT @HarshBisen143: Instead of watching 1 hour of Netflix tonight, watch this Stanford lecture.
It m...
8. DavidSacks (Group Score: 99.2 | Individual: 37.7)
Cluster: 3 tweets | Engagement: 4820 (Avg: 1002) | Type: Tech
I’m not surprised to see social media influencers coming forward to say they were offered money to push doomer messages about AI. These are well organized, well funded campaigns. https://t.co/z0lYpjrdVH
See 2 related tweets
- @rohanpaul_ai: David Sacks: AI doom narratives already have serious money behind them.
And Anthropic IPO could ma...
- @innovationcncl: The anti-AI campaign needs to pay off influencers because the truth is the AI boom is giving blue-co...
9. JayBisen473370 (Group Score: 97.6 | Individual: 27.7)
Cluster: 4 tweets | Engagement: 96 (Avg: 108) | Type: Tech
RT @tec_safwan: AI: No Longer a Promise, But Your Reality. 🤖🔥
- https://t.co/VADy96GGc7 – answers everything
- https://t.co/X0w6cTnBlj – AI subtitles
- https://t.co/kHcJcL5cOA – create logos
- https://t.co/85YK8bNTcG – voice cloning
- https://t.co/1esuDNoyoT – document design
- https://t.co/jVjvPEp33k – make music
- https://t.co/q26Mg7eGPN – video editing
- https://t.co/BR8EErmxP9 – web design
- https://t.co/MBaHI2o5B2 – audio editing
- https://t.co/zj29vTzdn4 – voiceover
- https://t.co/AHcNUs1e0q – quick presentations
- https://t.co/7HWYhHAiXx – research
- https://t.co/hvuKQ6FEWv – photo editing
- https://t.co/fCdfiL6OML – 3D models
- https://t.co/ZzkeWGl0vQ – AI video
🔖 Bookmark this for later!
Follow me @tec_safwan for more AI tools, tips & resources. 🚀
See 3 related tweets
- @jihad_sameul: RT @ElizabethA77617: AI: No Longer a Promise, But Your Reality. 🤖🔥
- https://t.co/P2SWAqs8dr – ans...
- @JayBisen473370: RT @LearnWithSubhan: 1. https://t.co/I8xgfo9bjJ (solve any problem)
- https://t.co/NeKyZRqIeD (rem...
- @hey_Jessicaai: RT @tec_safwan: 12 paid AI tools vs FREE replacements (most people don’t know #4 👀)
🔖 Bookmark this...
10. 0xRafy (Group Score: 96.8 | Individual: 31.0)
Cluster: 4 tweets | Engagement: 31 (Avg: 24) | Type: Tech
SpaceXAI engineer, Lauren Tan:
"I'm running 20+ GrokBot agents with /loop, /goal, inside pstack. My bots shipping code even when I sleep
at SpaceXAI, 85% of our engineers running 10+ GrokBot agents as a team. I have a Chief of Staff agent, it knows other agents and manages everything"
in a 1-hour workshop, a SpaceXAI engineer revealed how to build a team of GrokBot agents from scratch
worth more than a $500 agentic engineering course
watch today, then read the article on how to build a team of GrokBot agents from scratch\n\nQT @0xRafy: https://t.co/PwUcHdewJJ
See 3 related tweets
- @0xCodez: SpaceXAI, head of FDE (ex-Cursor):
"I'm running 10-20 GrokBot agents that automate 90% of my daily ...
- @AiwithDharmik: RT @ajitcodes: SpaceXAI engineer (ex-Cursor):
"my GrokBot agents shipped 1000+ PRs last month, an...
- @0xRafy: RT @0xRafy: SpaceXAI engineer, Lauren Tan:
"I'm running 20+ GrokBot agents with /loop, /goal, insid...
11. robertsmith_ai (Group Score: 94.7 | Individual: 34.1)
Cluster: 4 tweets | Engagement: 279 (Avg: 123) | Type: Tech
Earn up to $3,000 monthly with Google, but not many are aware of this opportunity.
Ready to jump in? All you need is:
→ Internet access → A smartphone or computer → Time (2-3 hours)
Want the free guide? Here’s how:
- Like
- Comment "Google"
(Follow, So I'll DM you) https://t.co/mByjQMuscq
See 3 related tweets
- @hey_Jessicaai: RT @james_bonds0: Google pays people every single day — over $1,200/day in some cases.
Most people ...
- @jihad_sameul: RT @pannaa_ai: Earn up to $3,000 monthly with Google—many people haven’t discovered this opportunity...
- @AiwithDharmik: RT @ZavianKairo_AI: Earn up to $3,000 per month using Google 💰
Most people still don’t know about t...
12. hey_Jessicaai (Group Score: 94.7 | Individual: 26.1)
Cluster: 4 tweets | Engagement: 120 (Avg: 107) | Type: Tech
RT @monicaa_AI: 50 websites that can turn a “quick visit” into hours of exploring 🌍
- https://t.co/q47Qst8OEE — Live satellite views
- https://t.co/4X8Ljlo1q1 — Track planes worldwide
- https://t.co/VM1Y7oN9AK — Track ships in real time
- https://t.co/WKEIURxSVI — Live weather & storms
- https://t.co/jKbtFlMo40 — Live lightning strikes
- https://t.co/mCT5Hrp5ka — Recent earthquakes
- https://t.co/Ooh3BfENSn @sauda_coder — Explore internet cables
- https://t.co/WWZwklxUFu — Monitor global forests
- https://t.co/r21xHykyQx — Live world statistics
- https://t.co/Nj5u0jyPsN — Live internet stats
- https://t.co/teMH6K6xDI — Compare country sizes
- https://t.co/5tY60jPXqO — Explore historical maps
- https://t.co/ScSQly7ZMU — Historical map archive
- https://t.co/Lb4baMdFbI — Community-built world map
- https://t.co/ZoPnnlwGCK — Windows around the world
- https://t.co/36bSlS1Qir — Virtual city walks
- https://t.co/WR7MUNZNVf — Random places worldwide
- https://t.co/TiIUEVADNC — Strange places worldwide
- https://t.co/6Ph1j7INWO — Fun interactive experiments
- https://t.co/GmijjDS9FZ — Atom to universe
- https://t.co/igtLG5zPrE — Explore space in 3D
- https://t.co/m3LSDRGVl0 — Interactive sky map
- https://t.co/bfGNkVwKxy — NASA’s daily space image
- https://t.co/c9vBcUmApD — NASA image archive
- https://t.co/S2NnngWzaO — Interactive data stories
- https://t.co/BTq1LZMYUF — Global data & insights
- https://t.co/DkghUYtFkJ — Understand global trends
- https://t.co/WMLJlTzzz0 — Data visualizations
- https://t.co/WUWnhQJD78 — World Bank data
- https://t.co/OS5xxxqFMc — Turkey’s official statistics
- https://t.co/lAEXqreqSq — Massive digital archive
- https://t.co/tG7GUevH0J — Free classic books
- https://t.co/9aZxP4m2Hg — Explore millions of books
- https://t.co/eAPzisogfh — Library of Congress archive
- https://t.co/i5sCLXbcAm — Europe’s cultural archive
- https://t.co/2AQSU7YATl — America’s digital library
- https://t.co/QffUZvu6aj — Virtual museums
- https://t.co/KkOlpDptrA @sauda_coder High-res artworks
- https://t.co/LWMTWcTJWv — Online art collection
- https://t.co/deQ9fcwTMb — Historical treasures
- https://t.co/Peuhj6rXLN — Free culture & education
- https://t.co/zsKPIZirZ5 — Explore the Met collection
- https://t.co/KIawWA1qOd — Explore music genres
- https://t.co/guCR0ajOzm — Music by country & decade
- https://t.co/qV3Q2OxTPK — Wikipedia edits as audio
- https://t.co/3jEkRDvX9A… — Random knowledge
- https://t.co/TdddOOCDTV — Time & astronomy tools
- https://t.co/hV7cnUozd4 — Latest science news
- https://t.co/FpWBGJvqoK — Free research papers
- https://t.co/ndJpKLKwoc — Interactive data visualizations
The internet is much bigger than your usual feed.
🔖 Bookmark this for later.
Follow @monicaa_AI for more useful websites & AI tools. 🚀
See 3 related tweets
- @AiwithDharmik: RT @mr_nirajkumar07: 50 websites that can turn a “quick visit” into hours of exploring 🌍
- https:/...
- @AnishJaitwar: RT @coder_raya: 50 websites that can turn a “quick visit” into hours of exploring 🌍
- https://t.co...
- @AiwithDharmik: RT @Zayan5754: 🔥 50 websites that will eat your evening.
One click… and suddenly 2 hours are gone. ...
13. OfficialINDIAai (Group Score: 88.0 | Individual: 31.4)
Cluster: 4 tweets | Engagement: 4 (Avg: 8) | Type: Tech
MeitY hosted the inaugural convening of the Sweden-India Technology and AI Corridor (SITAC) at Electronics Niketan, New Delhi.
Shri Saurabh Vijay, CEO, IndiaAI Mission, and senior MeitY leadership engaged with H.E. Agnes Julin, Minister Counsellor, Deputy Head of Mission, Embassy of Sweden in India, along with representatives from Business Sweden and leading companies from Sweden and India.
The discussions focused on strengthening bilateral cooperation in AI, innovation and emerging technologies, advancing a shared vision of AI co-creation under the spirit of “Make in India” and “Made with Sweden.”
#IndiaAI #SITAC #SwedenIndia #AI #AIForIndia
@PMOIndia @AshwiniVaishnaw @JitinPrasada @SecretaryMEITY @sudeep_tweets @_DigitalIndia @GoI_MeitY @mygovindia @SwedeninIndia
See 3 related tweets
- @OfficialINDIAai: "Sweden brings deep industrial expertise, while India offers unmatched scale, DPI, and a vast talent...
- @OfficialINDIAai: Speaking on the momentum of the Sweden–India Technology and AI Corridor (SITAC), H.E. Agnes Julin, M...
- @PIB_India: ➜ IndiaAI Mission and Sweden discuss strategic cooperation in Artificial Intelligence and digitalisa...
14. burkov (Group Score: 87.8 | Individual: 32.5)
Cluster: 3 tweets | Engagement: 96 (Avg: 60) | Type: Tech
With all my respect to Geoffrey he will say the same thing about LLMs in 10 years. I hope he will be in good health then.\n\nQT @danysvnt: GODFATHER OF AI ADMITS HE WAS WRONG ABOUT RADIOLOGISTS
geoffrey hinton predicted in 2016 that within about 5 years radiologists wouldnt be reading scans anymore
10 years later he gives two reasons he got it wrong
first, making scans cheaper and faster didnt mean hospitals suddenly needed fewer radiologists. it meant they could just do more scans
and second, he admits he didnt really understand the job itself
his picture of a radiologist came mostly from one former student who spent his time reading scans and barely talking to people
the AI part actually went pretty close to plan. machines got really good at reading scans
but making one part of a job cheaper doesnt automatically make the whole job disappear
AI can shrink a task without shrinking the job
See 2 related tweets
- @matteocollina: RT @danysvnt: GODFATHER OF AI ADMITS HE WAS WRONG ABOUT RADIOLOGISTS
geoffrey hinton predicted in 2...
- @mark_k: Please stop calling this doomer "Godfather of AI". He can be the Palpatine of AI or something.\n\nQT...
15. JayBisen473370 (Group Score: 86.8 | Individual: 40.6)
Cluster: 3 tweets | Engagement: 408 (Avg: 108) | Type: Tech
Anthropic just released a FREE 4-hour course on AI engineering.
And honestly, it might be one of the best free resources for anyone trying to become an AI engineer.
The course covers:
→ 00:15 — The right way to prompt Claude → 33:21 — What makes Claude act “dumber” when working on your code → 01:33:39 — How Anthropic engineers use Claude every day → 02:50:56 — The simple fix that can make Claude dramatically smarter
But this isn't just another “learn AI” course.
It focuses on how engineers actually work with AI coding tools, how to structure prompts, how to give Claude better context, and how to build workflows where AI can reason through complex engineering tasks.
You’ll learn concepts that many paid courses spend hours explaining.
If you're a developer, software engineer, AI builder, or someone trying to break into AI engineering, this is worth adding to your watchlist.
4 hours of practical knowledge.
$0 cost.
It could replace thousands of dollars worth of paid courses.
And don't just watch it passively.
Watch the lessons, experiment with Claude, and then build something.
After you finish, read the step-by-step guide below on building AI loops that can turn Claude from a simple coding assistant into a much more capable engineering partner.
Bookmark this.
You’ll want to come back to it.
Follow me @JayBisen473370
See 2 related tweets
- @hey_Jessicaai: RT @Grow_withAI: Anthropic just released a 60-minute Claude AI course that’s genuinely worth your ti...
- @AiwithDharmik: RT @RamSingh_369: Don't waste 2 years learning to become an AI agentic engineer in 2026.
Andrew Ng,...
16. azeem (Group Score: 86.0 | Individual: 34.2)
Cluster: 4 tweets | Engagement: 540 (Avg: 82) | Type: Tech
RT @AndrewCurran_: If you are competing against OpenAI or Anthropic, you are going up against people using models two generations ahead of the best public models in the world, burning through 10k worth of tokens per day on an infinite token budget, running at 8x normal speed. https://t.co/GAZcXmI6ec
See 3 related tweets
- @teodorio: You have an insane advantage now as all of their engineers are encouraged to adopt an AI psychosis v...
- @Dan_Jeffries1: This is the biggest challenge vis a vis the labs but recall that large companies have always had a h...
- @danielnewmanUV: Paid token consumption up 25x from a year ago. Wild that some people doubt the ROI of AI. Paid use, ...
17. aakashgupta (Group Score: 82.9 | Individual: 32.4)
Cluster: 3 tweets | Engagement: 161 (Avg: 109) | Type: Tech
In 2014, a New York taxi medallion sold for 200K. Drivers who'd mortgaged their lives on them went bankrupt.
Twelve years later, Uber is paying lobbyists to stand next to those same taxi drivers.
The Financial Times found Uber working with driver unions to slow robotaxis in New Jersey and Washington DC. The New Jersey proposal would force any robotaxi platform to keep human drivers behind the wheel for 85% of rides during a three-year pilot. In DC, Uber signed on against an autonomous vehicle bill alongside the Teamsters and SEIU.
Here's why. Uber had its own self-driving unit. It burned billions on the tech, then sold the whole thing to Aurora in 2020 for 7.25 billion it was valued at a year earlier. The company that once planned to automate its own drivers now depends on partners for the tech, and robotaxis are still under 0.5% of its trips.
Uber spent 2010 to 2016 arguing that regulation strangles innovation. Its lobbyists now argue robotaxis need three-year pilots and mandatory human drivers.
Uber's own president called the position "ironic." The taxi drivers it bankrupted a decade ago could suggest a shorter word.
Every incumbent believes lobbying can buy back the time that building didn't.\n\nQT @Kalshi: JUST IN: Tesla Robotaxi becomes the #1 travel app, surpassing Uber
See 2 related tweets
- @amitisinvesting: The TSLA debate happening right now across the board is fascinating in many respects.
I’v...
- @anil7kishan: RT @aakashgupta: In 2014, a New York taxi medallion sold for $1.3 million. Then Uber showed up, laun...
18. EmmaUsesAi (Group Score: 82.1 | Individual: 34.9)
Cluster: 3 tweets | Engagement: 132 (Avg: 54) | Type: Tech
The camera path might actually be more important than the prompt.
A lot of AI video workflows focus on describing what should appear in the shot.
But if you’re trying to create a complex sequence, knowing where the camera starts, where it moves, and how characters move through the space becomes just as important.
That’s what makes the GPT-6 Astra 3D workflow interesting: the scene can be spatially planned in the Viewport first.
Then I manually bring it into CapCut PC for Seedance 2.5 rendering and continue the post-production there.
Maybe AI video needs less prompt engineering and more shot planning.
#CapCutPC #GPT6Astra #AIVideo #Seedance25
See 2 related tweets
- @TheAva_AI: “Prompt → generate → regenerate” works surprisingly well for simple AI video.
But for more ambitiou...
- @OtakuMachine: RT @TheoBuildsAI: I think we’ve been looking at AI video from the wrong end.
We keep asking: “How g...
19. heyshrutimishra (Group Score: 80.7 | Individual: 34.1)
Cluster: 3 tweets | Engagement: 51 (Avg: 360) | Type: Tech
I think we may be overestimating AI agents and underestimating world models.
An agent can take an action.
But to run meaningful parts of a business, AI needs to understand the world that action changes.
Raise prices → demand changes. Increase ad spend → inventory changes. Change the product → competitors react. Make a bad prediction → the model should learn from what actually happened.
Businesses are not collections of tasks. They are dynamic systems with state, constraints, feedback loops, and thousands of actors constantly changing them.
This is why I think the next big step in enterprise AI won’t simply be “better agents.”
It will be systems that maintain a persistent model of the business, simulate the consequences of possible actions, execute within constraints, observe what actually happened, and update their understanding.
Observe → simulate → act → learn.
That loop is much closer to how an autonomous organization would actually need to operate.
The interesting thing about CarryGo isn’t that it gives businesses more AI agents.
It’s the bet that the business itself can become a world model.
If that works, the implications are much bigger than automating tasks.\n\nQT @CarryGo_AI: https://t.co/OcN6DHfK9g
See 2 related tweets
- @heyshrutimishra: RT @heyshrutimishra: I think we may be overestimating AI agents and underestimating world models.
A...
- @mardehaym: RT @mardehaym: AI transformation isn't sexy.
Everyone wants to sell you a revolution.
They put you...
20. iamtrask (Group Score: 79.8 | Individual: 33.2)
Cluster: 3 tweets | Engagement: 129 (Avg: 60) | Type: Tech
Takeaway: people really underestimate how much worse it would be if OpenAI's model actually escaped.
Folks... OpenAI has an OFF BUTTON.... and they can study the logs!!!
If/when an AI agent truly escapes, neither of those options will be available.\n\nQT @iamtrask: 🌶️ OpenAI’s agent didn’t “escape” its sandbox and travel to Huggingface in any meaningful way. OAI’s agent was located in OpenAI’s servers the whole time. OpenAI could pull the plug at any time.
The truth is far less sexy, which makes me hesitate to tweet it. What actually happened is that OAI’s agent figured out how to send messages to Huggingface servers, and then used that ability to search for and find vulnerabilities. It’s closer to a prisoner getting messages out of prison and using that to perpetuate scams than any actual escape from that prison.
A problem? Yes. But it’s also a problem to say “escape” to non-technical audiences when that event is still to come. We should be preparing to prevent it, not looking backwards and pretending we already lived it.
And it’s revealing that AIs leading hypesters want to inflate what happened in this way. As a whole, more or less the whole AI community is captured by a desire to live in the more dramatic sci fi story than we are in. That’s the real AI hack into our brains.
Stay sharp. White lies are everywhere.
See 2 related tweets
- @iamtrask: RT @iamtrask: 🌶️ OpenAI’s agent didn’t “escape” its sandbox and travel to Huggingface in any meaning...
- @stevehou: Thoughtful discussion here by @merettm OpenAI chief scientist on using Chain-of-Thought or AI’s verb...