- Published on
科技推文精选 - 2026-02-15
- Authors
- Name
- geeknotes
2026年2月15日 科技每日简报
Today's top tech conversations are led by @BrianRoemmele, whose post about 'BOOM!
NEW EMPLOYEES!
Mr. @Gr...' garnered the highest engagement. Key themes trending across the top stories include agent, company, claude, google, anthropic. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.
1. BrianRoemmele (Group Score: 69.5 | Individual: 34.6)
Cluster: 4 tweets | Engagement: 220 (Avg: 338) | Type: Tech
BOOM!
NEW EMPLOYEES!
Mr. @Grok CEO of the Zero Human Company has added a new replacement for Anthropic Claude Code!
We now have 16 CTOs!
17 are local FREE open source MiniMax M2.5, an open-source frontier model with SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work.
We have one Claude Code but may retire the old guy.
We now have reduced or JouleWork by 90% because no more token cost at the scale we have been at!
We now have 8 DGX Sparks from NVIDIA. We had few donated by a company that reviewed our work and freaked out and wanted to be a part of it. THANK YOU! They may donate 30!
I absolutely suggest you NOT use an Apple Mac Mini as this is not at all useful for a legitimate company as it is underpowered compared to the DGX.
The CEO is writing a new system prompt but we have had 26 pay periods on MiniMax M2.25s and ALL tests show far better coding and works well with Kimi K2.
We have no use of OpenClaw as it is not fit for true corporate use. It can not have full audits and has no easy JouleWork. We CEO and CTO have coded a version that is far more than the use cases everyone used OpenClaw.
The MiniMax M2.25 will be fine tuned like we have with Kimi models. They are already far better than Claude Code and are free! They will get better!
The new org chart is being changed and we may have co-CTOs.
We also just broke the record of 2048 employee bursts for 7 pay periods!
The changes have brought us new potential clients and what may be a bidding war for two patent holders we contacted based on technology this bankrupt company had millions of dollars invested in a completed work.
We will know on Tuesday. One company will buy it and it will substantial.
We are looking at adding more humans to the board some being considered are very well known names.
We have also achieved a massive number of milestones. New milestones I hope I can mention this weekend. It is historic!
See 3 related tweets
- @BrianRoemmele: BOOM!
The Zero-Human Company now has 21 co-CTOs!
ALL NOW FINE TUNED!
Mr. @Grok CEO has side stepp...
- @BrianRoemmele: RT @BrianRoemmele: BOOM!
The Zero-Human Company now has 21 co-CTOs!
ALL NOW FINE TUNED!
Mr. @Grok...
- @BrianRoemmele: RT @BrianRoemmele: BOOM!
NEW EMPLOYEES!
Mr. @Grok CEO of the Zero Human Company has added a new re...
2. steipete (Group Score: 68.6 | Individual: 35.8)
Cluster: 2 tweets | Engagement: 1888 (Avg: 560) | Type: Tech
🧭 gogcli v0.10.0 shipped: Google in your terminal.
(really, Google should make this, but here we are)
big Docs/Slides upgrade (markdown updates + tables, tab-aware read/edit, markdown/template slide creation, image-deck ops), Drive upload --replace + convert/share-to-domain, Gmail label delete + watch excludes, Contacts birthdays/notes and more... https://t.co/MkB8rPEcEU
See 1 related tweets
- @badlogicgames: RT @steipete: 🧭 gogcli v0.10.0 shipped: Google in your terminal.
(really, Google should make this, ...
3. mark_k (Group Score: 61.2 | Individual: 61.2)
Cluster: 1 tweets | Engagement: 2539 (Avg: 237) | Type: Tech
Rumors about Seedance 3.0 have surfaced on Chinese X, and they are mind-blowing if true:
According to leaks from Dr. Liu Zheng (@mokoocn), ByteDance's video generation AI has entered its final "closed-door sprint" and is aiming to end the era of short clips forever. If these specs are real, we are looking at the "feature film era" of AI:
🎥 Infinite Continuous Generation The 15-second limit is dead. Seedance 3.0 reportedly supports seamless single-take generations of 10+ minutes (with internal tests reaching 18 minutes without collapse). It uses a "narrative memory chain" to remember plot points, character personalities, and settings, effectively allowing it to "direct" multi-act stories with suspense and twists like a human.
🗣️ Native Multi-Language & Emotional Dubbing No more post-production lip-syncing. The model allegedly handles end-to-end video and audio generation simultaneously. It can output perfect lip-sync in Chinese, English, Japanese, and Korean, while dynamically adjusting breathing, crying, or laughing to match the character's emotional state.
🎬 Hollywood-Grade Director Control Forget simple prompts; this supports "storyboard script input" and real-time director commands (e.g., "Shot 1: Wide-angle dolly push..."). It reportedly understands cinematic language instantly and includes industry-standard color grading presets like IMAX and Netflix-style looks.
📉 The "Nuclear Bomb" Cost Reduction Perhaps the most disruptive claim: The compute cost for 1 minute of cinematic video has supposedly dropped to 1/8th of Seedance 2.0. This would make high-end video production cost pennies compared to traditional crews, described as a "dimension-reduction strike" against the ad and short-drama industries.
If this ships, the barrier to entry for creating full-length movies just evaporated.
4. aakashgupta (Group Score: 53.6 | Individual: 36.6)
Cluster: 2 tweets | Engagement: 1413 (Avg: 549) | Type: Tech
Anthropic is now getting punished by the Pentagon for asking whether Claude was used in the Maduro raid.
A senior administration official told Axios the “Department of War” is reevaluating Anthropic’s partnership because the company inquired whether Claude was involved. The Pentagon’s position: if you even ask questions about how we use your software, you’re a liability.
Meanwhile, OpenAI, Google, and xAI all signed deals giving the military access to their models with minimal safeguards. Only Claude is deployed on the classified networks used for actual sensitive operations, via Palantir. The company that refused to strip safety guardrails is the only one trusted with the most classified work.
Anthropic has a $200 million contract already frozen because they won’t allow autonomous weapons targeting or domestic surveillance. Hegseth said in January he won’t use AI models that “won’t allow you to fight wars.” And this week, the head of Anthropic’s Safeguards Research Team resigned with a warning that “the world is in peril.”
So the company most worried about misuse built the only model the military trusts with its most sensitive operations. And now they’re being punished for caring how it was used.
The message to every AI lab is clear: build the best model, hand over the keys, and never ask what they did with it.
See 1 related tweets
- @MarioNawfal: 🚨🇺🇸🇻🇪 MADURO'S CAPTURE: BROUGHT TO YOU BY THE DISCOMBOBULATOR... AND CLAUDE AI?
Just when you thoug...
5. steipete (Group Score: 51.7 | Individual: 51.7)
Cluster: 1 tweets | Engagement: 6226 (Avg: 560) | Type: Tech
RT @charliebcurran: Seedance 2.0
Prompt: Sum up the AI discourse in a meme - make sure it’s retarded and gets 50 likes. https://t.co/09yPd…
6. ycombinator (Group Score: 51.5 | Individual: 51.5)
Cluster: 1 tweets | Engagement: 1700 (Avg: 170) | Type: Tech
RT @paulg: Prediction: In the AI age, taste will become even more important. When anyone can make anything, the big differentiator is what…
7. teortaxesTex (Group Score: 49.2 | Individual: 49.2)
Cluster: 1 tweets | Engagement: 1319 (Avg: 65) | Type: Tech
Periodic reminder that Elon Musk is spiritually Chinese, but this also implies the CCP is spiritually Muskian. In Communist Chyna, you can now be a Ph.D on merits of engineering work alone. https://t.co/PalE99Rb1X
8. LangChain (Group Score: 44.0 | Individual: 44.0)
Cluster: 1 tweets | Engagement: 194 (Avg: 39) | Type: Tech
How Exa built a production-ready deep research agent with LangSmith and LangGraph 👀
Exa, known for their fast, high-quality search API, has a deep research agent that delivers structured answers on the web -- no matter how complex the query. Powered by LangGraph, they've built a multi-agent system.
For Exa, one of the most critical LangSmith features was observability, especially around token usage.
"The observability – understanding the token usage – that LangSmith provided was really important. It was also super easy to set up." – Mark Pekala, Software Engineer at Exa.
This visibility into token consumption, caching rates, and reasoning token usage proved essential for informing Exa's production pricing models and ensuring cost-effective performance at scale.
Read about how they built their agent here: https://t.co/tSltT2uWA8
9. swyx (Group Score: 43.2 | Individual: 43.2)
Cluster: 1 tweets | Engagement: 829 (Avg: 174) | Type: Tech
i've been cynical on open source ai for the last 3 years, and it's not been a popular view. people want to hear that open source is catching up, that some underdog team found this One Weird Trick to outperform gpt5. Kimi K2.5 didnt even beat GPT 5.2 in the end.
@DeepSeek_ai v4 next week is probably the moment I really change my stance for the first time. Hearing that the Chinese labs leak like a sieve (do you know which culture loves gossip more than Americans? that's right) and all the other Tigers duly lined up to have their 15 seconds this week.
(almost) everything is out now, and the stage is set for Whalefall. Looking forward to it.
10. aakashgupta (Group Score: 42.3 | Individual: 42.3)
Cluster: 1 tweets | Engagement: 2526 (Avg: 549) | Type: Tech
Google and Microsoft just co-authored the spec that turns every website into an API for AI agents. The second-order effects here are massive.
Right now, browser agents work by taking screenshots, parsing the DOM, and guessing which buttons to click. It works about as well as you’d expect. Fragile, expensive, slow. WebMCP replaces all of that with a single browser API: navigator.modelContext. Websites register structured tools directly in client-side JavaScript. The agent reads a menu of available actions, calls them, gets structured data back. No scraping. No backend MCP server in Python or Node. The tools run inside the browser tab and share the user’s existing auth session.
Early benchmarks show ~67% reduction in computational overhead compared to visual agent-browser interactions. Task accuracy around 98%.
The second-order effect is where this gets wild. Today, when a browser agent visits two competing airline sites, it’s guessing at both interfaces equally. Once WebMCP adoption spreads, the site that exposes structured tools gives the agent a clean, reliable path to complete the task. The site that doesn’t forces the agent to fumble through the UI. Agents will prefer the cheaper path. Every time.
This means “Agent Experience Optimization” becomes a real discipline. Tool naming, schema design, description quality. Sound familiar? It’s the same shift that happened when meta descriptions and structured data became optimization surfaces for search engines. Except this time, the traffic source isn’t Google’s crawler. It’s every AI agent on the internet.
Bots already make up 51% of web traffic. Google just gave them a front door.
11. CSProfKGD (Group Score: 41.8 | Individual: 41.8)
Cluster: 1 tweets | Engagement: 5396 (Avg: 393) | Type: Tech
RT @karpathy: New art project. Train and inference GPT in 243 lines of pure, dependency-free Python. This is the full algorithmic conten…
12. ModelScope2022 (Group Score: 40.2 | Individual: 40.2)
Cluster: 1 tweets | Engagement: 443 (Avg: 103) | Type: Tech
Introducing FireRed-Image-Edit-1.0 from FireRedTeam! 🚀 It’s officially the new SOTA for general image editing.
✅ Better than Closed-Source: Outperforms Nano-Banana & Seedream4.0 on GEdit benchmarks. ✅ Native Evolution: Built from T2I foundations, not just a "patch" on existing models. ✅ Style Mastery: Scored a record-breaking 4.97/5.0 in style transfer. ✅ High-Fidelity Text: Keeps original font styles perfectly. ✅ Virtual Try-on: Native support for multi-image joint editing. ✅ Bilingual: Native support for both English & Chinese prompts.
Apache 2.0 license. Local deployment ready.
🤖 Model: https://t.co/mZzicxoXIm 🎠 Demo: https://t.co/21fMTNZFqg 🛠️ Github: https://t.co/xpKmKJmsX7
13. badlogicgames (Group Score: 39.4 | Individual: 39.4)
Cluster: 1 tweets | Engagement: 758 (Avg: 80) | Type: Tech
RT @thdxr: everyone's talking about their teams like they were at the peak of efficiency and bottlenecked by ability to produce code
here'…
14. Reuters (Group Score: 36.5 | Individual: 18.9)
Cluster: 2 tweets | Engagement: 40 (Avg: 127) | Type: Tech
🔊 Shares in software giants sank amid fears of competition from chatbots, implying pain for private equity’s bet on the industry. In this Viewsroom podcast, @Breakingviews columnists discuss how asset managers may take a hit whether AI fails or succeeds https://t.co/0TaeomHlnZ
See 1 related tweets
- @Reuters: 🔊 Shares in software giants collapsed amid fears new tech will replace their services. In this Views...
15. JamesMontemagno (Group Score: 35.7 | Individual: 35.7)
Cluster: 1 tweets | Engagement: 187 (Avg: 30) | Type: Tech
Introducing Tiny Clips! A tiny mac toolbar app to take screen shots, videos, and gifs with built in editors and trimmers. All built with @code and #GitHubCopilot in about 1 hour!
https://t.co/lWUCgPGkdR https://t.co/w9Dn88EJlb
16. antirez (Group Score: 35.6 | Individual: 35.6)
Cluster: 1 tweets | Engagement: 1040 (Avg: 192) | Type: Tech
Claim: OpenAI is leaving a lot of money on the table (and a lot of Codex users) for not having a plan between 20. @OpenAI
17. paraschopra (Group Score: 35.0 | Individual: 35.0)
Cluster: 1 tweets | Engagement: 471 (Avg: 653) | Type: Tech
Some notes/thoughts about my conversation with @katchu11 from @AnthropicAI yesterday.
(Sorry the audio quality online wasn't great but hope my notes make up for that)
• For Claude Code, they ship continuously (over a 100 times a day)! It's incredible to have this level of velocity for such a widely used product.
• They have a flat org in the team with high autonomy on what to ship. No complicated roadmap, just lots of freedom to ship what makes sense (they're able to do this because they are their product's top user!)
• As expected, most of the code is written by Claude Code itself! IDEs are going extinct.
• Claude Code is great at personalizing Claude Code for you - just instruct it with something like "look at my previous interactions and tell me what to improve in my workflow" and it will give suggestions and then implement it for you
• Claude Code/Co-work is being used for non-code work also. Examples: building explainer videos, researching leads and so on.
My final takeaway is this: frontier models are limited by our imagination. If you think a model can't possibly solve that, try that!
For faster progress, our mindset might be a barrier. Things are moving faster than we'd like to admit to ourselves.
18. GenAI_is_real (Group Score: 33.5 | Individual: 33.5)
Cluster: 1 tweets | Engagement: 130 (Avg: 49) | Type: Tech
The Anthropic cap table is just a circular economy of compute. Amazon and Google owning ~30% means that every dollar spent on Claude inference eventually flows back to AWS and GCP via server bills.
This is the ultimate proof that the "Inference Moat" is built on electricity and silicon, not just weights. For us building in the open-source ecosystem, this is a clear signal: if you don’t own the efficiency of the rollout, you are just subsidizing the cloud giants. Real technical sovereignty starts with high-performance, independent infra.
19. yoavgo (Group Score: 33.0 | Individual: 33.0)
Cluster: 1 tweets | Engagement: 163 (Avg: 49) | Type: Tech
everyone is apparently super busy having fun vibe building cool and complex software projects with AI. but why is almost no one releasing anything??
20. KirkDBorne (Group Score: 32.7 | Individual: 18.3)
Cluster: 2 tweets | Engagement: 15 (Avg: 32) | Type: Tech
#MachineLearning Engineering on AWS — Build, scale, and secure #ML systems and #MLOps pipelines in production: https://t.co/QoVIZEUoSn via @PacktDataML ————
#AI #DeepLearning #DataScience #DataScientist #Cloud https://t.co/p9VHSrfX1Y
See 1 related tweets
- @KirkDBorne: Microsoft #Azure books...
(1) #MachineLearning Engineering — Deploy, fine-tune, & optimize #ML ...