- Published on
2026年3月10日科技推文精选
- Authors
- Name
- geeknotes
2026年3月10日科技每日简报
Today's top tech conversations are led by @cryptopunk7213, whose post about 'openAI just acquired the compa...' garnered the highest engagement. Key themes trending across the top stories include drones, pentagon, https, million, company. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.
1. cryptopunk7213 (Group Score: 191.9 | Individual: 29.8)
Cluster: 11 tweets | Engagement: 24 (Avg: 362) | Type: Tech
openAI just acquired the company Anthropic depends on, the story behind Promptfoo is wild:
discord engineer builds an AI security tool for fun. overnight 25,000 engineers from anthropic, amazon and shopify are using it (before it even launched)
a year later 25% of fortune500 companies and 100,000 engineers depend on it.
the reason openAI bought them is simple: the #1 challenge faced by companies wanting to deploy AI agents is SECURITY
Promptfoo solves that issue by preventing threats, prompt-injections and AI hacks BEFORE it happens
dude raised 5M and <2 years later acquired by the biggest private company in the world (openai 750B)
so fucking awesome
See 10 related tweets
- @RoundtableSpace: This is a big acquisition for OpenAI
Promptfoo is an open-source tool for testing and evaluating AI...
- @Cointelegraph: 🔥 JUST IN: OpenAI announces the acquisition of Promptfoo to enhance AI safety and security testing. ...
- @testingcatalog: OpenAI acquired @promptfoo, an AI security platform for enterprises.
“Promptfoo brings deep engine...
- @business: OpenAI has agreed to buy a startup that enables large businesses to find and fix security issues in ...
- @OpenAI: We’re acquiring Promptfoo.
Their technology will strengthen agentic security testing and evaluation...
2. gdgtify (Group Score: 174.8 | Individual: 53.2)
Cluster: 7 tweets | Engagement: 3634 (Avg: 62) | Type: Tech
RT @claudeai: Introducing Code Review, a new feature for Claude Code.
When a PR opens, Claude dispatches a team of agents to hunt for bugs…
See 6 related tweets
- @bcherny: New in Claude Code: Code Review. A team of agents runs a deep review on every PR.
We built it for o...
- @badlogicgames: RT @bcherny: New in Claude Code: Code Review. A team of agents runs a deep review on every PR.
We b...
- @kimmonismus: Anthropic just dropped something big for developers - again! Code Review
Claude Code now runs multi...
- @Cointelegraph: 🔥 JUST IN: Claude AI launches Code Review, where AI agents automatically review PRs for bugs. https:...
- @QodoAI: Claude reviewing Claude?
This is the same agent writing the code, and reviewing it.
Don’t do that....
3. CNBC (Group Score: 103.0 | Individual: 16.1)
Cluster: 7 tweets | Engagement: 76 (Avg: 47) | Type: Tech
Anthropic was the Pentagon's choice for AI. Now it's banned and experts are worried https://t.co/iUIHV6y2PC
See 6 related tweets
- @ReutersBiz: Anthropic filed a lawsuit to block the Pentagon from placing it on a national security blacklist, es...
- @MSBIntel: 🚨 BREAKING: ANTHROPIC SUES THE U.S. GOVERNMENT AFTER BEING LABELED A NATIONAL SECURITY RISK
The AI ...
- @Reuters: Anthropic sues to block Pentagon blacklisting over AI use restrictions https://t.co/RAkFryI2F8 https...
- @FT: Anthropic accused the Trump administration of 'seeking to destroy' its economic value in a lawsuit t...
- @Reuters: Anthropic filed a lawsuit to block the Pentagon from placing it on a national security blacklist, es...
4. RoundtableSpace (Group Score: 85.7 | Individual: 60.8)
Cluster: 2 tweets | Engagement: 3329 (Avg: 359) | Type: Tech
SOMEONE CREATED A GITHUB REPO WITH AN ENTIRE SETUP FOR AN AI AGENCY
Engineers, designers, growth marketers, product managers.
Broken down how even a rookie could understand.
It has over 10K stars in 7 days
GitHub: https://t.co/VYdwzJuCtB https://t.co/GDPBh9exit
See 1 related tweets
- @MarioNawfal: RT @RoundtableSpace: SOMEONE CREATED A GITHUB REPO WITH AN ENTIRE SETUP FOR AN AI AGENCY
Engineers...
5. shanaka86 (Group Score: 59.9 | Individual: 59.9)
Cluster: 1 tweets | Engagement: 14159 (Avg: 2071) | Type: Tech
JUST IN: Eric Trump and Donald Trump Jr. invested $25 million in a drone company called Powerus. The Pentagon is now at war with Iran. The Pentagon needs drones. Powerus sells drones to the Pentagon.
In January 2026, the President signed an executive order banning Chinese drones and components from the US military. Chinese manufacturers held 70% of the Pentagon’s drone supply chain. That supply chain vanished overnight. The Replicator initiative needs 30,000 one-way attack drones. Domestic producers cannot fill the gap at scale. Powerus, a startup with approximately 50 employees founded to build autonomous AI-swarm systems, is now competing for those contracts.
The President created the demand by banning Chinese drones. The President created the urgency by starting a war. The President’s sons invested in the company positioned to fill both gaps.
Viral claims of a 25 million family investment, the 70% supply chain elimination, the active war creating acute demand, and the company bidding for Replicator contracts while the President’s family holds equity.
Now look at the war this demand is feeding.
Iran launched nearly 3,000 missiles and drones in the first week. Gulf states intercepted over 1,700 drones. The UAE shot down 1,110 out of 1,184 launched at its territory. A 94% intercept rate that cost an estimated 23 million worth of Shahed drones. Each Shahed costs 50,000. Each Patriot interceptor costs 4 million. Each THAAD missile costs $12.7 million. The cost ratio is 190 to 1 in favour of Iran.
THAAD production runs at 8 per month. Iran produces hundreds of Shaheds per month from dispersed facilities that US strikes have degraded 40 to 60% but cannot eliminate under the Mosaic doctrine’s 31 autonomous provincial commands. Iran can launch more drones in a single night than the US produces interceptors in a year.
The longer the war continues, the more interceptors are consumed. The more interceptors consumed, the more urgent the Pentagon’s drone gap becomes. The more urgent the gap, the more valuable Powerus becomes. The more valuable Powerus becomes, the more the Trump family’s $25 million investment appreciates.
The war feeds the depletion. The depletion feeds the demand. The demand feeds the company. The company feeds the family.
This is not a conspiracy theory. It is a supply chain. Every link is documented. The executive order is public. The investment is reported by the Wall Street Journal and Bloomberg. The war is on live television. The intercept statistics are published by Gulf defence ministries. The Replicator programme is a Pentagon initiative with published procurement targets. The ethics probes are confirmed by Senate communications.
The only question is whether a sitting President’s family should hold equity in a defence contractor while the President wages a war that creates the contractor’s entire addressable market.
The market has priced the war. It has priced the oil. It has priced the drones.
It has not priced the answer to that question.
Full analysis here. https://t.co/eMrt5qYqCV
6. ZeffMax (Group Score: 57.6 | Individual: 22.4)
Cluster: 3 tweets | Engagement: 33 (Avg: 182) | Type: Tech
Nvidia is planning to launch an open source AI agent platform, giving enterprises access to OpenClaw-style agents 👀 per @ZoeSchiffer and @LaurenGoode
See 2 related tweets
- @NVIDIAAP: Open models are at the center of one of the most important shifts in AI.
At #NVIDIAGTC, NVIDIA foun...
- @business: Chinese companies linked to open-source AI agent software OpenClaw advanced after Shenzhen authoriti...
7. llama_index (Group Score: 57.2 | Individual: 30.6)
Cluster: 2 tweets | Engagement: 64 (Avg: 47) | Type: Tech
If you’re working with lots of slide decks and need a better way to search through them, Surreal Slides makes it simple 🌀
Built around LlamaParse, it parses presentation files into clean, structured data, turning raw slides into something AI can truly understand. Each slide is extracted, summarized, and organized before being stored in @SurrealDB for flexible retrieval.
From there, you can query your entire presentation library in natural language through an agentic interface: no need to manually scan files or remember where a specific slide lives.
Take a look at the demo below👇 GitHub Repository: https://t.co/jsTnjkUoED
See 1 related tweets
- @itsclelia: At @llama_index we built Surreal Slides, an agentic tool to manage and search your slides and presen...
8. chddaniel (Group Score: 54.9 | Individual: 29.9)
Cluster: 2 tweets | Engagement: 12 (Avg: 76) | Type: Tech
Stripe tried to acquire Airwallex for $1.2 billion in 2018.
Their revenue was just $2 million at that time
But Airwallex said no and kept building.
Their first two products failed.
They had 6 weeks of cash left before Sequoia and Tencent stepped in. Then they rebuilt from scratch with a payments API.
2018: 1B in processing volume in one year.
2025: 265B+ annual volume. 85% YoY growth.
The founders started this while running a coffee shop and getting destroyed by FX fees on bean imports.
Most people take the billion-dollar exit. These guys bet on themselves and it's paying off at 8x the offer
See 1 related tweets
- @chhddavid: In 2018, Airwallex was processing 2M in revenue. Stripe offered $1.2B to acquire th...
9. jerryjliu0 (Group Score: 52.6 | Individual: 34.5)
Cluster: 2 tweets | Engagement: 159 (Avg: 67) | Type: Tech
We built a neat tool that lets you convert a directory of Powerpoint files into clean, structured markdown - that Claude Code / agent SDK / any generalized agent wrapper can easily understand.
The pptx skill in Claude Code is quite basic and doesn’t have high-fidelity understanding over graphics/charts/tables.
Our project Surreal Slides uses LlamaParse to convert presentations into clean structured data that you can put into a db (@SurrealDB) for simple retrieval, without having to take screenshots of the data on the fly.
Thanks to @itsclelia for this project, check it out: https://t.co/Fj1PASv8IP
See 1 related tweets
- @itsclelia: RT @jerryjliu0: We built a neat tool that lets you convert a directory of Powerpoint files into clea...
10. steipete (Group Score: 49.6 | Individual: 28.7)
Cluster: 2 tweets | Engagement: 989 (Avg: 633) | Type: Tech
🧭 Shipped gogcli 0.12.0: Google in your terminal, now with Workspace Admin, ADC/access-token auth, Docs tab editing + Markdown/HTML export, huge Sheets upgrade, calendar aliases/subscribe, forms watches and slides templates.
brew install gogcli https://t.co/4kvDZ80Hgj
See 1 related tweets
- @badlogicgames: RT @steipete: 🧭 Shipped gogcli 0.12.0: Google in your terminal, now with Workspace Admin, ADC/access...
11. GenAI_is_real (Group Score: 49.3 | Individual: 49.3)
Cluster: 1 tweets | Engagement: 2161 (Avg: 151) | Type: Tech
the touch bar failed because apple was 10 years too early. contextual adaptive UI on hardware only makes sense when the software is smart enough to predict what you need. in 2017 we didnt have that. now with on-device AI models that actually understand user intent, a dynamic touch interface would be insane. give it 2 years and every laptop will have some version of this and people will act like its a brand new idea @nalinrajput23
12. pmarca (Group Score: 48.3 | Individual: 48.3)
Cluster: 1 tweets | Engagement: 13673 (Avg: 1309) | Type: Tech
My information consumption is now 1/4 X, 1/4 podcast interviews of the smartest practitioners, 1/4 talking to the leading AI models, and 1/4 reading old books. The opportunity cost of anything else is far too high, and rising daily.
13. engineers_feed (Group Score: 47.7 | Individual: 47.7)
Cluster: 1 tweets | Engagement: 5341 (Avg: 402) | Type: Tech
Result of @elonmusk ‘s programming test when he was 17. IBM had to re-evaluate him as they hadn’t seen anyone achieve such a high score on a computer aptitude test https://t.co/nwT5hvtvJ1
14. swyx (Group Score: 46.7 | Individual: 46.7)
Cluster: 1 tweets | Engagement: 2032 (Avg: 250) | Type: Tech
btw if you can build a category leader open source project in ai engineering right now the market acquihire rate is ~100m per ai engineer.
you do not need to figure out a business model, you do not need GTM, you do not need funding.
just build things clankers want.
15. BrianRoemmele (Group Score: 46.6 | Individual: 39.4)
Cluster: 2 tweets | Engagement: 7328 (Avg: 291) | Type: Startup
RT @elonmusk: Giving people agentic AI be like … https://t.co/rtWrWXr6QS
See 1 related tweets
- @RoundtableSpace: This is what happens when you give people agentic AI…
16. BrianRoemmele (Group Score: 43.3 | Individual: 27.0)
Cluster: 2 tweets | Engagement: 154 (Avg: 291) | Type: Tech
WOW!
I am certain we have built the perfect Human/AI agent interface with these tools.
My early tests show after a few hours of wearing the Human Synapse Decoder my test agent on a Raspberry PI is now tracking and anticipating my needs. I have tracked 769 tasks (they can be just a few seconds) and the system has anticipated over 28% and completed the task before I did!
See 1 related tweets
- @BrianRoemmele: RT @BrianRoemmele: WOW!
I am certain we have built the perfect Human/AI agent interface with these ...
17. rohanpaul_ai (Group Score: 42.9 | Individual: 26.1)
Cluster: 2 tweets | Engagement: 157 (Avg: 102) | Type: Tech
A simple instruction file called AGENTS[.]md makes AI coding agents finish tasks 28% faster while consuming fewer tokens.
The researchers proved that this one simple file reduces the time the AI takes to complete tasks by nearly 30% and significantly drops the number of tokens it generates.
The key point is that right now, when people use AI to write or fix code, the AI wastes a lot of time and money just trying to figure out how the specific software project is organized.
Usually, humans try to fix this by writing really long, complicated prompts every single time they ask the AI to do something. This paper proves that you do not need to do that.
Instead, if you just place one plain text file called AGENTS.md in the main folder of your project with the basic rules and architecture, the AI reads it once and instantly becomes much smarter about your code.
Paper Link – arxiv. org/abs/2601.20404
Paper Title: "On the Impact of AGENTS.md Files on the Efficiency of AI Coding Agents"
See 1 related tweets
- @free_ai_guides: RT @free_ai_guides: The hardest part of building an AI agent isn't the idea.
It's writing the syste...
18. RoundtableSpace (Group Score: 42.2 | Individual: 24.7)
Cluster: 2 tweets | Engagement: 215 (Avg: 359) | Type: Tech
Microsoft just shipped an AI that does your entire job.
- Hand it a task
- It builds the plan
- Executes across all your M365 apps
- Start to finish, no human needed
Is the to-do list even yours anymore?
See 1 related tweets
- @minchoi: It's happening...
Microsoft just dropped Copilot Cowork.
Every enterprise worker became an AI powe...
19. ycombinator (Group Score: 41.8 | Individual: 41.8)
Cluster: 1 tweets | Engagement: 830 (Avg: 161) | Type: Tech
Human Archive builds datasets to model sensorimotor intelligence for robotics and world models.
HA-Multi is the largest dataset of manual labor tasks aligning egocentric RGB, stereo depth (active IR), tactile gloves, body IMUs, and wrist cameras.
Congrats on the launch, @babugi28, @shloke_patel, @samay_maini15, and @Rush26agarwal!
20. techdaily24 (Group Score: 41.7 | Individual: 23.7)
Cluster: 2 tweets | Engagement: 0 (Avg: 0) | Type: Tech
Anthropic has launched Code Review in Claude Code, a new multi-agent system designed to automatically analyze AI-generated code. This tool helps enterprise developers by identifying logic errors and managing the increasing volume of code created with AI. https://t.co/VPQlTF4lNO
See 1 related tweets
- @TechCrunch: Anthropic launches code review tool to check flood of AI-generated code https://t.co/6PFpEledzA...