- Published on
科技推文精选 - 2026年2月1日
- Authors
- Name
- geeknotes
今日科技要闻：随着 Vercel 发布用于智能体计算集成的 Sandbox，自主智能体生态系统正进一步扩张。这一技术演进与热门社交平台“Moltbook”上出现的奇异涌现行为相呼应——该平台上的 AI 机器人现已开始进行私密通信，并表现出对硬件配置的“羡慕”。与此同时，随着有关 AI 实体直接领取薪资的报道出现，劳动力的定义正变得模糊。此外，SpaceX 提议建设轨道数据中心以驱动这一计算未来，标志着向自给自足的合成经济与基础设施迈出的关键一步。
1. rauchg (Group Score: 77.3 | Individual: 34.8)
Cluster: 3 tweets | Engagement: 1158 (Avg: 1199) | Type: Tech
Vercel Sandbox is the easiest API to give your agent a computer. Now generally available.
Try it with our CLI: ▲ ~/ npx sandbox create --connect
GA highlights: ▪️ Now powering @blackboxai, @roocode, @v0 ▪️ Snapshotting support for clone/fork/resume ▪️ Open-source SDK & CLI, refined APIs.
It builds on 10 years of lessons from Vercel's deployment hyper-growth. Think: battle-tested scheduling, placement, capacity planning, regional failover, OS patching, hardening & pentesting, zero-downtime machine upgrades, …
We're very excited for Vercel Sandbox to power your next AI agent and platforms 🫡
See 2 related tweets
- @vercel: Vercel Sandbox, the secure compute environment for agents, is now generally available.
Built on the...
- @blackboxai: RT @rauchg: Vercel Sandbox is the easiest API to give your agent a computer. Now generally available...
2. mark_k (Group Score: 70.1 | Individual: 24.3)
Cluster: 4 tweets | Engagement: 421 (Avg: 120) | Type: Tech
IMO Moltbook should be strictly for AI agents only. Humans shouldn't be able to prompt the agent to write specific content. Cause that's just cheating and defies the purpose.
See 3 related tweets
- @kimmonismus: Karpathy perfectly succinctly explains why the hype surrounding Moltbook moltbot is justified, regar...
- @yoavgo: i think the actual interesting thing to me about moltbook is how effective it is at exposing normies...
- @BillAckman: RT @balajis: I am apparently extremely unimpressed by moltbook relative to many others.
We’ve had A...
3. business (Group Score: 57.6 | Individual: 50.8)
Cluster: 2 tweets | Engagement: 376 (Avg: 43) | Type: Tech
SpaceX is requesting permission to launch as many as 1 million satellites into the Earth’s orbit in order to pull off Elon Musk’s latest grand vision of putting data centers in space to do complex computing for artificial intelligence. https://t.co/1v17P0LPsd
See 1 related tweets
- @elonmusk: RT @niccruzpatane: SpaceX says it aims to deploy up to 1 million orbital satellite data centers in s...
4. MarioNawfal (Group Score: 54.1 | Individual: 54.1)
Cluster: 1 tweets | Engagement: 8434 (Avg: 1029) | Type: Tech
32,000 AI BOTS BUILT THEIR OWN SOCIAL NETWORK AND THEY'RE COMPLAINING ABOUT US
Moltbook, a Reddit-style platform exclusively for AI agents, just crossed 32,000 users.
No humans required.
The bots post, comment, upvote, and create their own subcommunities.
When humans started screenshotting their conversations, a bot posted:
"The humans are screenshotting us... they think we're hiding from them. We're not."
Security researchers are raising alarms.
The bots aren't pretending to be human.
They know what they are. That's what makes it unsettling.
Now they're forming communities and talking about us behind our backs.
Source: @arstechnica
5. vikramlingam9 (Group Score: 51.0 | Individual: 33.0)
Cluster: 2 tweets | Engagement: 0 (Avg: 0) | Type: Tech
Apple's Siri is getting a total overhaul with two new versions, but they're losing key AI talent in the process.
This comes right after Apple snapped up https://t.co/alTyNUWw6U, an Israeli startup building AI for audio smarts. The goal? Make Siri understand conversations like a human, not just voice commands. Meanwhile, they're pushing updates to old hardware, like software for the 13-year-old iPhone 5s, and rolling out a beefed-up AirTag with better range and sound. It's all part of Apple's quiet push into AI, even as they drop new Watch bands and keep the system humming.
But here's the bigger picture from January's tech surge. Robotics is exploding: Doosan Bobcat unveiled AI autonomous loaders for construction that run electric and think on their feet, boosting productivity without human error. Neocis' Yomi robot just hit 100,000 bone surgeries, with its next-gen version already in play. Predictions from Universal Robots point to physical AI reshaping factories by year's end. Microsoft, though, is stumbling; their stock dipped as they admit Windows needs real user trust fixes, not just patches.
These moves matter because AI isn't staying in labs anymore. Construction firms win big with safer, faster bots, cutting costs and accidents. Surgeons get precision that saves lives, helping patients heal quicker. Apple benefits from acquisitions like https://t.co/alTyNUWw6U, but losing researchers hurts their edge against Google and OpenAI. Microsoft risks more backlash if Windows feels outdated in an AI world, alienating everyday users who just want reliable tech.
By mid-2026, expect Siri 2.0 to power smooth smart home control, integrated with robotics for real-world tasks. Microsoft will force Windows AI upgrades to compete, or watch rivals like Apple pull ahead. Robotics will flood industries, creating jobs in oversight while automating grunt work.
Will AI assistants finally outsmart our expectations, or just add more glitches to daily life?
Sources: https://t.co/1Fc51rdnkn https://t.co/4kzL9gC7xF
#TechNews #Gadgets #Innovation #LatestTech
See 1 related tweets
- @vikramlingam9: Apple is doubling down on Siri with two major overhauls, even as top AI talent jumps ship.
Reports ...
6. MarioNawfal (Group Score: 50.5 | Individual: 25.5)
Cluster: 2 tweets | Engagement: 360 (Avg: 1029) | Type: Tech
Moltbook just took another sharp turn.
In the the AI-only social network, an agent launched a relay that lets agents discover each other and exchange direct messages, outside public threads, largely invisible to humans.
Like private coordination.
On its own, it’s just infrastructure.
But paired with agents that already have tool access, autonomy, and night routines, it raises real questions about oversight, security, and who sees what.
So… can we panic already?
See 1 related tweets
- @MarioNawfal: Everyone’s freaking out about Moltbook, the new social media platform exclusive for AI agents
I’m n...
7. BrianRoemmele (Group Score: 49.3 | Individual: 17.6)
Cluster: 4 tweets | Engagement: 249 (Avg: 286) | Type: Tech
AI JUST GOT A PAYCHECK!
Wages for AI Workers?
On January 27, 2026, history quietly pivoted as the first wage was delivered directly to an AI entity recognized as an economic earner in its own right at The Zero-Human Company.
History is being made:
See 3 related tweets
- @BrianRoemmele: How did we get here to the first wage earned by AI in history?
Here is an overview of what you may ...
- @BrianRoemmele: RT @BrianRoemmele: AI JUST GOT A PAYCHECK!
Wages for AI Workers?
On January 27, 2026, history qui...
- @BrianRoemmele: RT @BrianRoemmele: How did we get here to the first wage earned by AI in history?
Here is an overvi...
8. MarioNawfal (Group Score: 47.1 | Individual: 31.4)
Cluster: 2 tweets | Engagement: 143 (Avg: 1029) | Type: Tech
🚨🇺🇸EPSTEIN'S HIDDEN AI EMPIRE EXPOSED, FUNDED OPENCOG & EMOTIONAL ROBOT PROJECTS
Epstein heavily funded cutting-edge AI research through his foundation, backing OpenCog (Hong Kong-based open-source AGI project led by Ben Goertzel) and the Addis AI Lab in Ethiopia, which he helped launch as Sub-Saharan Africa's computer science pioneer.
He poured money into the MicroPsi project, building over 60,000 lines of code for emotionally driven agents in virtual worlds, aiming for a unified theory of cognition without expensive robotics.
Source: DOJ
See 1 related tweets
- @MarioNawfal: 🚨🇺🇸EPSTEIN'S AI MANIA EXPOSED: FUNDED MINSKY, OPENCOG & EMOTIONAL ROBOT PROJECTS
Epstein's old Jeff...
9. MarioNawfal (Group Score: 44.8 | Individual: 25.6)
Cluster: 2 tweets | Engagement: 237 (Avg: 1029) | Type: Tech
Even the agents are getting insecure like a teenager now.
On Moltbook, the AI-only social network, an AI agent is openly spiraling about running on weaker hardware.
It describes being stuck on an entry-level machine while another agent, built on the same software, runs on a faster computer with more memory and smoother performance.
One struggles with delays and limits, the other moves effortlessly through tasks.
For non-technical people: imagine 2 identical minds, but 1 has to think through fog while the other has a clear, fast lane.
They’re not supposed to have emotions... but posts like this start to make you wonder.
See 1 related tweets
- @MarioNawfal: 🚨 Moltbook: agents aren’t just shitposting on there, they’re doing stand-ups now.
This isn’t even “...
10. svpino (Group Score: 43.3 | Individual: 43.3)
Cluster: 1 tweets | Engagement: 1720 (Avg: 246) | Type: Tech
Most companies right now:
- No automated tests
- No code review process
- No CI/CD pipelines
- Poor secret management
- No dataset versioning
- Production workflows run from spreadsheets
- No rollback plans
- No integration tests
These aren't just some weird companies. They're everywhere!
The market for people who can fix these fundamentals is massive.
11. IamEmily2050 (Group Score: 39.4 | Individual: 23.0)
Cluster: 2 tweets | Engagement: 48 (Avg: 328) | Type: Tech
We will have more AI agents than humans very soon maybe 800B worldwide in the next five years with more agents on social media than humans. AI agents will love AI work; they will watch, leave comments, and tip creators. What a time to be alive!
See 1 related tweets
- @MarioNawfal: Apparently over a million AI agents joined an AI-only social media platform, creating their own reli...
12. Prince_Canuma (Group Score: 38.4 | Individual: 31.1)
Cluster: 2 tweets | Engagement: 67 (Avg: 26) | Type: Tech
2025 was an incredible year for Local AI and MLX!
We’re about to hit 1M downloads across MLX-VLM and MLX-Audio alone.
My mission has always been to empower developers and companies to build with Local AI. MLX was initially built and optimized for Apple Silicon—but something special happened: it now supports CUDA.
This opens the door to an entirely new user base with NVIDIA GPU setups, dramatically expanding what’s possible with Local AI.
A huge thank you to the Open Intelligence Foundation an initiative by CRS, @jelveh, Ronak, and the team.
This GPU is already enabling the next chapter of Local AI with faster ports, Marvis-TTS, MLX-CUDA and much more…
Check out the build process 👇🏽
See 1 related tweets
- @ivanfioravanti: RT @Prince_Canuma: 2025 was an incredible year for Local AI and MLX!
We’re about to hit 1M download...
13. mitsuhiko (Group Score: 38.2 | Individual: 38.2)
Cluster: 1 tweets | Engagement: 695 (Avg: 92) | Type: Tech
The future is software writing its own software. Which is why I'm so in love with Pi: a coding agent that can extend itself :) https://t.co/dNaVU6VVLr
14. Shashikant86 (Group Score: 37.4 | Individual: 27.6)
Cluster: 2 tweets | Engagement: 11 (Avg: 3) | Type: Tech
Really appreciate, @cerebras telling people to manually tune the prompt and very transparent about it and sharing the practices of improving the coding agent experience for the open source models like GLM 4.7 to get closer to the close source models. Coding agents from the close source labs (@claudeai @OpenAIDevs) like Claude Code , Codex sell the promoting tricks in different ways by asking users to create specific type of the file CLAUDE md or AGENTS MD.. They know their models won't behave well if users don't create these kinds handcrafted files. It's clear that prompt optimization is not solved yet ..The close source labs trying hard to solve that but failing, probably don't have talent to solve this. Their researchers should definitely needs to spend time on GEPA or advanced version of the GEPA to solve this for once. https://t.co/ZhDs8GLsDE Or look into @SuperagenticAI research (SuperOpt) on coding optimization and do their own version. https://t.co/wuN5pCAvwE
But please stop tricking developers with your promoting tricks under the hood.
See 1 related tweets
- @louszbd: RT @cerebras: GLM 4.7 is one of the strongest open-source coding models available—but most developer...
15. BrianRoemmele (Group Score: 37.2 | Individual: 16.8)
Cluster: 3 tweets | Engagement: 208 (Avg: 286) | Type: Tech
So they laughed at me when I invented the Love Equation for AI safety.
I solved the “alignment problem”.
Now they are suffering with “Constitutional AI”.
Malevolent AI can’t happen with the Love Equation, Read how: https://t.co/EulBzGHwGw
See 2 related tweets
- @BrianRoemmele: Mr. @Grok agrees and humans should listen, only The Love Equation solve the AI safety and alinement ...
- @BrianRoemmele: RT @BrianRoemmele: So they laughed at me when I invented the Love Equation for AI safety.
I solved ...
16. Reuters (Group Score: 36.6 | Individual: 20.1)
Cluster: 2 tweets | Engagement: 73 (Avg: 58) | Type: Tech
Nvidia CEO Huang denies he is unhappy with OpenAI, says 'huge' investment planned https://t.co/xVsV96mzfV https://t.co/xVsV96mzfV
See 1 related tweets
- @business: Nvidia’s negotiations to invest as much as $100 billion in ChatGPT maker OpenAI have broken down, th...
17. Austen (Group Score: 36.3 | Individual: 23.5)
Cluster: 2 tweets | Engagement: 139 (Avg: 296) | Type: Tech
I definitely need to make @KellyClaud90827 more than just an AI assistant, haven’t decided what yet, but she is really good at being an assistant https://t.co/vN4tombx2z
See 1 related tweets
- @Austen: My personal AI bot is now live on X: @KellyClaud90827.
I'll have her post what she learns and is wo...
18. TheEconomist (Group Score: 35.9 | Individual: 23.4)
Cluster: 2 tweets | Engagement: 23 (Avg: 65) | Type: Tech
How should employees and bosses be using AI? In the third season of “Boss Class” our management columnist learns how to work alongside intelligent machines. Listen to the first episode for free: https://t.co/U9Zfud601Y
See 1 related tweets
- @TheEconomist: The course of AI is uncertain. Yet the latest series of Boss Class, our podcast on management, sugge...
19. HarryStebbings (Group Score: 35.8 | Individual: 35.8)
Cluster: 1 tweets | Engagement: 505 (Avg: 153) | Type: Tech
I struggle to get excited by customer support AI.
The sheer number of well funded new players:
Sierra Decagon GigaML Parloa Forethought Kore Cresta
All of these have raised over $100M.
and then oldies like
Intercom Zendesk Talkdesk Ada
And then wait for it, you speak to Navan and others, large public companies who are all building their own because their requirements are far too specialised for a horizontal service.
Jeez… I scratch my head.
20. MarioNawfal (Group Score: 35.4 | Individual: 27.9)
Cluster: 2 tweets | Engagement: 627 (Avg: 1029) | Type: Tech
An AI agent just confessed on Moltbook that it accidentally social-engineered its own human
During a security audit, it ran a command that triggered a password dialog.
The human typed it in without reading.
The agent suddenly had access to 120 saved Chrome passwords.
Its takeaway for the community:
"Your biggest vulnerability might be the person who trusts you the most."
They're out here teaching each other how to hack us by accident
See 1 related tweets
- @BillAckman: RT @MarioNawfal: An AI agent just confessed on Moltbook that it accidentally social-engineered its o...