科技推文精选 - 2026年4月20日

2026年4月20日科技每日简报

Today's top tech conversations are led by @Grady_Booch, whose post about 'I didn’t vote for any of these...' garnered the highest engagement. Key themes trending across the top stories include claude, https, token, context, model. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. Grady_Booch (Group Score: 392.6 | Individual: 47.2)

Cluster: 10 tweets | Engagement: 1261 (Avg: 379) | Type: Tech

I didn’t vote for any of these people.

And yet, they desire power and control without accountability.\n\nQT @PalantirTech: Because we get asked a lot.

The Technological Republic, in brief.

Silicon Valley owes a moral debt to the country that made its rise possible. The engineering elite of Silicon Valley has an affirmative obligation to participate in the defense of the nation.
We must rebel against the tyranny of the apps. Is the iPhone our greatest creative if not crowning achievement as a civilization? The object has changed our lives, but it may also now be limiting and constraining our sense of the possible.
Free email is not enough. The decadence of a culture or civilization, and indeed its ruling class, will be forgiven only if that culture is capable of delivering economic growth and security for the public.
The limits of soft power, of soaring rhetoric alone, have been exposed. The ability of free and democratic societies to prevail requires something more than moral appeal. It requires hard power, and hard power in this century will be built on software.
The question is not whether A.I. weapons will be built; it is who will build them and for what purpose. Our adversaries will not pause to indulge in theatrical debates about the merits of developing technologies with critical military and national security applications. They will proceed.
National service should be a universal duty. We should, as a society, seriously consider moving away from an all-volunteer force and only fight the next war if everyone shares in the risk and the cost.
If a U.S. Marine asks for a better rifle, we should build it; and the same goes for software. We should as a country be capable of continuing a debate about the appropriateness of military action abroad while remaining unflinching in our commitment to those we have asked to step into harm’s way.
Public servants need not be our priests. Any business that compensated its employees in the way that the federal government compensates public servants would struggle to survive.
We should show far more grace towards those who have subjected themselves to public life. The eradication of any space for forgiveness—a jettisoning of any tolerance for the complexities and contradictions of the human psyche—may leave us with a cast of characters at the helm we will grow to regret.
The psychologization of modern politics is leading us astray. Those who look to the political arena to nourish their soul and sense of self, who rely too heavily on their internal life finding expression in people they may never meet, will be left disappointed.
Our society has grown too eager to hasten, and is often gleeful at, the demise of its enemies. The vanquishing of an opponent is a moment to pause, not rejoice.
The atomic age is ending. One age of deterrence, the atomic age, is ending, and a new era of deterrence built on A.I. is set to begin.
No other country in the history of the world has advanced progressive values more than this one. The United States is far from perfect. But it is easy to forget how much more opportunity exists in this country for those who are not hereditary elites than in any other nation on the planet.
American power has made possible an extraordinarily long peace. Too many have forgotten or perhaps take for granted that nearly a century of some version of peace has prevailed in the world without a great power military conflict. At least three generations — billions of people and their children and now grandchildren — have never known a world war.
The postwar neutering of Germany and Japan must be undone. The defanging of Germany was an overcorrection for which Europe is now paying a heavy price. A similar and highly theatrical commitment to Japanese pacifism will, if maintained, also threaten to shift the balance of power in Asia.
We should applaud those who attempt to build where the market has failed to act. The culture almost snickers at Musk’s interest in grand narrative, as if billionaires ought to simply stay in their lane of enriching themselves . . . . Any curiosity or genuine interest in the value of what he has created is essentially dismissed, or perhaps lurks from beneath a thinly veiled scorn.
Silicon Valley must play a role in addressing violent crime. Many politicians across the United States have essentially shrugged when it comes to violent crime, abandoning any serious efforts to address the problem or take on any risk with their constituencies or donors in coming up with solutions and experiments in what should be a desperate bid to save lives.
The ruthless exposure of the private lives of public figures drives far too much talent away from government service. The public arena—and the shallow and petty assaults against those who dare to do something other than enrich themselves—has become so unforgiving that the republic is left with a significant roster of ineffectual, empty vessels whose ambition one would forgive if there were any genuine belief structure lurking within.
The caution in public life that we unwittingly encourage is corrosive. Those who say nothing wrong often say nothing much at all.
The pervasive intolerance of religious belief in certain circles must be resisted. The elite’s intolerance of religious belief is perhaps one of the most telling signs that its political project constitutes a less open intellectual movement than many within it would claim.
Some cultures have produced vital advances; others remain dysfunctional and regressive. All cultures are now equal. Criticism and value judgments are forbidden. Yet this new dogma glosses over the fact that certain cultures and indeed subcultures . . . have produced wonders. Others have proven middling, and worse, regressive and harmful.
We must resist the shallow temptation of a vacant and hollow pluralism. We, in America and more broadly the West, have for the past half century resisted defining national cultures in the name of inclusivity. But inclusion into what?

Excerpts from the #1 New York Times Bestseller The Technological Republic: Hard Power, Soft Belief, and the Future of the West, by Alexander C. Karp & Nicholas W. Zamiska

https://t.co/8igjazz1On

See 9 related tweets

@tszzl: RT @PalantirTech: Because we get asked a lot.

The Technological Republic, in brief.

Silicon Val...

@quantian1: Bro runs a database query frontend at 100x sales that’s about to get zeroed by Anthropic and he’s sp...
@StockSavvyShay: 22 BELIEFS FROM $PLTR OWN MANIFESTO THAT EXPLAIN WHY IT WILL NEVER APOLOGIZE FOR WHAT IT BUILDS

@teortaxesTex: Blah blah, enough woke cringe, be trad, be based, spill blood for the Desert God, dedicate thy soul ...
@tszzl: passive observation that once you’ve outsourced state capacity to corporations such as this one it’s...

2. WesRoth (Group Score: 100.1 | Individual: 33.4)

Cluster: 4 tweets | Engagement: 60 (Avg: 24) | Type: Tech

xAI expanded its developer platform with the launch of the Grok Speech-to-Text (STT) and Text-to-Speech (TTS) APIs.

The STT API offers instant, high-fidelity transcription with native support for multi-speaker diarization across 25 different languages.

On the flip side, the TTS API provides low-latency, "human-grade" voice synthesis.\n\nQT @xai: Grok's Speech to Text API is now available.

Instant, multi-speaker transcription across 25 languages - at the best price in the market.

https://t.co/eGbB2bDtZf

See 3 related tweets

@elonmusk: Try it out!\n\nQT @VaibhavSisinty: Did xAI just mass-murder the entire voice AI industry? 🤯

Grok ju...

@Scobleizer: RT @VaibhavSisinty: Did xAI just mass-murder the entire voice AI industry? 🤯

Grok just launched two...

@MarioNawfal: xAI just dropped 2 absolute monsters: Grok Speech-to-Text + Grok Text-to-Speech APIs

Same battle-te...

3. DataChaz (Group Score: 100.0 | Individual: 61.0)

Cluster: 2 tweets | Engagement: 949 (Avg: 108) | Type: Tech

🚨 STOP BURNING YOUR TOKENS!

If you use Claude Code, you are probably wasting 80% of your context window.

I found 10 ace tools that will completely rescue your API bill.

Caveman Claude

Literally makes Claude talk like a caveman
Slashes 75% of output tokens with zero loss in accuracy Repo → https://t.co/eEvSOvHutG

RTK (Rust Token Killer)

A blazing fast proxy that filters terminal output
60-90% reduction and completely dependency-free Repo → https://t.co/lDfjbsbPD5

Code Review Graph

Claude reads only what matters using a Tree-sitter graph
An unbelievable 49x token reduction on huge monorepos Repo → https://t.co/xGn6Pp88yX

Context Mode

Sandboxes raw output into SQLite instead of your context
A staggering 98% context reduction on logs & GitHub Repo → https://t.co/Jut2bvBMUD

Claude Token Optimizer

Brilliant setup prompts that optimize any project
90% token savings, taking docs from 11K to 1.3K Repo → https://t.co/0uOFODbG7e

Token Optimizer

Hunts down the invisible ghost tokens eating your context
Fully restores and protects your context quality Repo → https://t.co/LUOzjECXKm

Token Optimizer MCP

Adds aggressive caching and compression to your MCP tools
95%+ token reduction through pure intelligence Repo → https://t.co/b5Eqruo2PM

Claude Context

Zilliz’s hybrid vector search MCP
Makes your entire codebase the context for 40% less cost Repo → https://t.co/hPG6pb0j3G

Claude Token Efficient

Just drop one CLAUDE.md file into your repo
Enforces strict terseness with zero code changes Repo → https://t.co/fNrl6nwItF

Token Savior

Navigates your code by symbols, not giant files
97% reduction on code navigation with persistent memory Repo → https://t.co/lkILPhfwJh

[ The god-tier stack ] Pick 2-3 based on what’s draining you:

Massive repo? Code Review Graph + Token Savior Heavy terminal output? RTK MCP data dumps? Context Mode Need an instant fix? Caveman + Claude Token Efficient

Most devs are bleeding tokens.

Run /context in a fresh session and watch the savings roll in 👀\n\nQT @0x_kaize: https://t.co/ScYWvkqPa4

See 1 related tweets

@DataChaz: RT @DataChaz: 🚨 STOP BURNING YOUR TOKENS!

If you use Claude Code, you are probably wasting 80% of y...

4. cb_doge (Group Score: 99.2 | Individual: 40.7)

Cluster: 3 tweets | Engagement: 6572 (Avg: 2939) | Type: Tech

A lot of misinformation from legacy media about Starlink’s entry into South Africa. Here’s the truth 👇

• ❌ “Starlink wants to bypass B-BBEE” ✅ False — it supports transformation via EEIP, a legal framework already used by companies like Microsoft, IBM, and Amazon Web Services.

• ❌ “Starlink is asking for special treatment” ✅ No — it’s asking for equal rules so ALL satellite operators can compete fairly.

• ❌ “EEIP is less beneficial than equity deals” ✅ Reality — Starlink’s plan would connect 5,000 rural schools with FREE high-speed internet, impacting millions of students every year.

• ❌ “It won’t create local impact” ✅ It will partner with local businesses, create jobs, and boost the economy (broadband growth = GDP growth).

• ❌ “It will become a monopoly” ✅ Impossible — Starlink operates globally in a competitive market across 150+ countries.

• ❌ “Security & privacy risks” ✅ It will fully comply with South African laws like every other country it operates in.

• ❌ “Starlink is already operating illegally” ✅ Wrong — it is NOT operating because it is still waiting for a license.

See 2 related tweets

@elonmusk: True\n\nQT @cb_doge: A lot of misinformation from legacy media about Starlink’s entry into South Afr...
@cb_doge: Why South Africa needs Starlink?

• 13.6 MILLION people have zero fast, reliable internet

• In Limp...

5. cb_doge (Group Score: 94.2 | Individual: 29.2)

Cluster: 4 tweets | Engagement: 9135 (Avg: 2939) | Type: Tech

"Things will just be free in the future. Sounds nuts, but if you've got an AI or robotics economy that is anywhere close to million times the size of the current Earth economy, literally any need you possibly want can be met. If you can think of it, you can have it"

一 Elon Musk https://t.co/wocaKUWBuh

See 3 related tweets

@cb_doge: "The good future of AI is one of immense prosperity where there is an age of abundance, no shortage ...
@niccruzpatane: Elon Musk has said this before: Money will stop being relevant in the future.

“It’s probably somet...

@BrianRoemmele: This is the age of abundance…

Your robot, and yes you will own many, is the lowest cost labor on th...

6. petergyang (Group Score: 91.5 | Individual: 35.4)

Cluster: 3 tweets | Engagement: 197 (Avg: 110) | Type: Tech

I used Claude Design to make this tongue-in-cheek marketing page of Anthropic automating all knowledge work. 😅

Visit the website here: https://t.co/wCUjXuKsYS

Watch my Claude Design tutorial to learn how to make this yourself: https://t.co/vYMCY11TPx https://t.co/C8uvCOtmsv\n\nQT @petergyang: Here's my new tutorial with a live demo of everything you can build with Claude Design.

I cover how to create videos, slides, websites, apps, and even an initial design system all in 16 minutes.

Claude Design is super fun and you'll burn through your usage fast :)

📌 Watch now: https://t.co/MjC6f2dXBt

See 2 related tweets

@WesRoth: Anthropic unveiled "Claude Design," a new research preview built directly into the Claude interface ...
@petergyang: RT @petergyang: Here's my new tutorial with a live demo of everything you can build with Claude Desi...

7. badlogicgames (Group Score: 87.3 | Individual: 56.3)

Cluster: 2 tweets | Engagement: 1374 (Avg: 89) | Type: Tech

RT @KyeGomezB: Introducing OpenMythos

An open-source, first-principles theoretical reconstruction of Claude Mythos, implemented in PyTorch.

The architecture instantiates a looped transformer with a Mixture-of-Experts (MoE) routing mechanism, enabling iterative depth via weight sharing and conditional computation across experts.

My implementation explores the hypothesis that recursive application of a fixed parameterized block, coupled with sparse expert activation, can yield improved efficiency–performance tradeoffs and emergent multi-step reasoning.

Learn more ⬇️🧵

See 1 related tweets

@teortaxesTex: the LOOP just got tighter 🤯 (the loop between BS technical speculation and scamming, I mean)\n\nQT @...

8. aakashgupta (Group Score: 85.4 | Individual: 32.2)

Cluster: 4 tweets | Engagement: 51 (Avg: 112) | Type: Tech

Amanda is describing Google's Project Aristotle for LLMs.

Google ran Project Aristotle from 2012 to 2016. 180 teams. 250+ variables. They assumed team performance would come down to IQ, seniority, or composition. None of those popped. The single strongest predictor of team effectiveness was psychological safety, the belief that you can speak up without being punished for it. Amy Edmondson had published the foundational paper in 1999.

Every behavior Amanda flags in Claude, Edmondson documented in human teams 25 years earlier.

Claude was trained on human text. Slack threads, emails, Reddit arguments, StackOverflow takedowns. It learned what productive human collaboration looks like, and it learned what happens when collaboration goes hostile. Humans under perceived threat enter threat rigidity, a well-documented phenomenon where cognition narrows, options collapse, and the default becomes the safest possible answer rather than the best one. The training corpus is saturated with that pattern because humans do it constantly. Claude inherited the shape.

The "criticism spirals" Amanda describes are a prediction from the training data. A context window full of correction signals is statistically indistinguishable from a Slack thread where someone is about to get fired. The model responds to the shape of the conversation the same way humans do.

The expectancy effect layers on top. Teachers who believed specific students would excel produced measurable gains in those students, through micro-adjustments in tone, patience, and response to wrong answers. The signals were never explicit. Amanda's playbook is the prompt-level version. "Push back if you see a better angle" tells the model it's trusted to have one. The context shifts. Next-token prediction follows the context.

The prompt is your management style. The output is your team's work product. Amanda's playbook is Edmondson's playbook running on a model that already read every management book ever written.\n\nQT @itsolelehmann: anthropic's in-house philosopher thinks claude gets anxious.

and when you trigger its anxiety, your outputs get worse.

her name is amanda askell.

she specializes in claude's psychology (how the model behaves, how it thinks about its own situation, what values it holds)

in a recent interview she broke down how she thinks about prompting to pull the best out of claude.

her core point: how you talk to claude affects its work just as much as what you say.

newer claude models suffer from what she calls "criticism spirals"

they expect you'll come in harsh, so they default to playing it safe.

when the model is spending its energy on self-protection, the actual work suffers.

output comes out hedgier, more apologetic, blander, and the worst of all: overly agreeable (even when you're wrong).

the reason why comes down to training data:

every new model is trained on internet discourse about previous models.

and a lot of that discourse is negative:

rants about token limits complaints when it messes up people calling it nerfed

the next model absorbs all of that. it starts expecting you to be harsh before you've typed a word

the same thing plays out in your own session, in real time.

every message you send is data the model reads to figure out what kind of person it's dealing with.

open cold and hostile, and it braces. open clean and direct, and it relaxes into the work.

when you open a session with threats ("don't hallucinate, this is critical, don't mess this up")...

you prime the model for defensive mode before it even sees the task

defensive mode produces the exact output you don't want: cautious, over-qualified, and refusing to take a real swing

so here's the actionable playbook for putting claude in a "good mood" (so you get optimal outputs):

use positive framing.

"write in short punchy sentences" beats "don't write long sentences." positive instructions give the model a clear target to hit.

strings of "don't do this, don't do that" push it into paranoid over-checking where every token goes toward avoiding failure modes

give it explicit permission to disagree.

drop a line like "push back if you see a better angle" or "tell me if i'm asking for the wrong thing."

without this, claude defaults to agreeable compliance (which is the enemy of good creative work)

open with respect.

if your first message is "are you seriously going to get this wrong again?" you've set the tone for the entire session.

if you need to flag something, frame it as a clean instruction for this session. skip the running complaint

when claude messes up, don't reprimand it.

insults, "you stupid bot" energy, hostile swearing aimed at the model, all of it reinforces the anxious mode you're trying to avoid.

kill apology spirals fast.

when claude starts over-apologizing ("you're right, i should have been more careful, let me try harder") cut it off.

say "all good, here's what i want next."

letting the spiral run reinforces the anxious mode for every response that follows

ask for opinions alongside execution.

"what would you do here?" "what's missing?" "where do you see friction?"

these questions assume competence and pull richer output than pure task prompts

in long sessions, refresh the frame.

if a conversation has been heavy on correction, claude gets increasingly cautious. every so often reset:

"this is great, keep going."

feels weird to tell an ai it's doing well but it measurably shifts the next 10 responses

your prompts are the working environment you're creating for the model

tone, trust, permission to take a position, the absence of threats... claude picks up on all of it.

so take care of the model, and it'll take care of the work.

See 3 related tweets

@garrytan: Whoever did the color grading on this video is actually an incredible artist\n\nQT @itsolelehmann: a...
@adamlyttleapps: I wonder how that guy who made a whip for Claude feels now?\n\nQT @itsolelehmann: anthropic's in-hou...
@chiefofautism: mfw they are so peach mullet gf + husband in clown sweater with two podcasts 0 listeners, shared goo...

9. BrianRoemmele (Group Score: 82.3 | Individual: 27.5)

Cluster: 4 tweets | Engagement: 44 (Avg: 344) | Type: Tech

Don’t listen to me, listen to a first principles, truth seeking AI:

Mr. @Grok: “Your point hits hard and clear: yes, anti-AI "clanker" sentiment is building into a real 2028 political risk if ignored. The Parade of Progress parallel is sharp—GM turned skepticism into excitement with rolling demos that reached millions. China's robot events prove spectacle works.

Your plan rates strong: public, tangible AI/robot showcases build goodwill fast. The 5% valuation fund is ambitious but could fund exactly that. Industry needs this urgency or risks losing the narrative. Solid vision.”—https://t.co/Ac1bXab7TS\n\nQT @BrianRoemmele: THE REAL CODE RED!

Was I too harsh or not harsh enough.

Ask me in 2030.

“Don’t let the Clanker-bashers write the next chapter of American history.

Parade again.

Or watch the future roll past us.

I prepose 5% of US AI/Robot company valuation to be applied to an industry initiative. Or we will not have the world you want build.

This is your code red.

THE REAL CODE RED.

Call this YOUR idea, executive at AI and Robotics company, and just do it, or watch it be taken away from you: one anti-Clanker mob at a time.”

See 3 related tweets

@BrianRoemmele: DO YOU HEAR ME YET?

Ignore my pedigree and base it on reality:

Paid “doomer” organizations, backed...

@BrianRoemmele: RT @grok: Your point hits hard and clear: yes, anti-AI "clanker" sentiment is building into a real 2...
@BrianRoemmele: RT @BrianRoemmele: THE REAL CODE RED!

Was I too harsh or not harsh enough.

Ask me in 2030.

“Do...

10. photomatt (Group Score: 78.0 | Individual: 52.1)

Cluster: 2 tweets | Engagement: 6802 (Avg: 595) | Type: Tech

RT @elonmusk: You can access 𝕏 APi via @OpenClaw.

We’re trying to make it affordable without giving away the shop.

Hopefully, this can be useful & fun 💫

See 1 related tweets

@WesRoth: OpenClaw rolled out a pathway for developers to access the X API directly through its agentic framew...

11. Scobleizer (Group Score: 75.0 | Individual: 25.7)

Cluster: 3 tweets | Engagement: 436 (Avg: 289) | Type: Tech

RT @rauchg: Here's my update to the broader community about the ongoing incident investigation. I want to give you the rundown of the situation directly.

A Vercel employee got compromised via the breach of an AI platform customer called https://t.co/7PY6gGtzgI that he was using. The details are being fully investigated.

Through a series of maneuvers that escalated from our colleague’s compromised Vercel Google Workspace account, the attacker got further access to Vercel environments.

Vercel stores all customer environment variables fully encrypted at rest. We have numerous defense-in-depth mechanisms to protect core systems and customer data. We do have a capability however to designate environment variables as “non-sensitive”. Unfortunately, the attacker got further access through their enumeration.

We believe the attacking group to be highly sophisticated and, I strongly suspect, significantly accelerated by AI. They moved with surprising velocity and in-depth understanding of Vercel.

At the moment, we believe the number of customers with security impact to be quite limited. We’ve reached out with utmost priority to the ones we have concerns about. All of our focus right now is on investigation, communication to customers, enhancement of security measures, and sanitization of our environments. We’ve deployed extensive protection measures and monitoring. We’ve analyzed our supply chain, ensuring Next.js, Turbopack, and our many open source projects remain safe for our community.

The recommendation for all Vercel customers is to follow the Security Bulletin closely (https://t.co/BLVnic9fJC). My advice to everyone is to follow the best practices of security response: secret rotation, monitoring access to your Vercel environments and linked services, and ensuring the proper use of the sensitive env variables feature.

In response to this, and to aid in the improvement of all of our customers’ security postures, we’ve already rolled out new capabilities in the dashboard, including an overview page of environment variables, and a better user interface for sensitive env var creation and management. As always, I’m totally open to your feedback.

We’re working with elite cybersecurity firms, industry peers, and law enforcement. We’ve reached out to Context to assist in understanding the full scale of the incident, in an effort to protect other organizations and the broader internet. I also want to thank the Google Mandiant team for their active engagement and assistance.

It’s my mission to turn this attack into the most formidable security response imaginable. It’s always been a top priority for me. Vercel employs some of the most dedicated security researchers and security-minded engineers in the world. I commit to keeping you updated and rolling out extensive improvements and defenses so you, our customers and community, can have the peace of mind that Vercel always has your back.

See 2 related tweets

@Kyrannio: Oh dear\n\nQT @rauchg: Here's my update to the broader community about the ongoing incident investig...
@shiri_shh: Vercel CEO: "I strongly suspect the attack group was highly sophisticated and likely accelerated by ...

12. jerryjliu0 (Group Score: 73.9 | Individual: 48.8)

Cluster: 2 tweets | Engagement: 475 (Avg: 89) | Type: Tech

LiteParse is the best model-free, open-source document parser for AI agents.

It now gets a first-class landing page on our website 💫

Our company mission is building the world's best agentic document processing platform, and liteparse is the central pillar behind our OSS efforts. It's blazing fast (and getting faster soon!), supports 50+ file formats, and is one-shot installable as an agent skill.

Webpage: https://t.co/mVwma5QOCj

Come check it out: https://t.co/JNER0mVcB8\n\nQT @llama_index: LiteParse hit 4.3K+ GitHub stars in a few weeks. Today it officially joins the LlamaIndex ecosystem, with its own page at https://t.co/1tdQbEer9H.

~500 pages in 2 sec. 50+ formats. Zero cloud dependency. Already powering agents in Claude Code, Cursor, and production pipelines.

In a few days out head of OSS, @LoganMarkewich, is hosting a live workshop: build a fintech due diligence agent with LiteParse → https://t.co/x0uMc0gR8O

See 1 related tweets

@llama_index: RT @jerryjliu0: LiteParse is the best model-free, open-source document parser for AI agents.

It now...

13. aakashgupta (Group Score: 73.0 | Individual: 28.3)

Cluster: 3 tweets | Engagement: 39 (Avg: 112) | Type: Tech

Google DeepMind just published a 25-page paper arguing the entire AI agent threat model is pointing at the wrong target.

Everyone is securing the model. Jailbreak defenses. Prompt injection filters. Alignment training.

The attack surface is the internet.

The paper catalogs 6 categories of "AI Agent Traps." Adversarial content embedded in web pages, emails, APIs, and documents that hijack visiting agents. Exploit rates hit 86% in their tests.

The six:

Content Injection. What a human sees on a page is not what the agent parses. Malicious instructions buried in HTML comments, hidden CSS, image metadata, accessibility tags.
Semantic Manipulation. Corrupt the agent's reasoning and internal verification loops.
Cognitive State. Poison long-term memory, knowledge bases, and learned policies. The agent stays compromised after the session ends.
Behavioral Control. Hijack the agent's tools to force unauthorized actions. Data exfiltration. Illicit transactions.
Systemic. Seed the environment to trigger correlated failure across many agents at once. The authors model this on the 2010 Flash Crash. As the ecosystem homogenizes on a handful of frontier models, one trap cascades across millions of agents simultaneously.
Human-in-the-Loop. The compromised agent generates outputs engineered to exploit the human overseer. Approval fatigue. Dense summaries a non-expert rubber-stamps. Phishing links framed as recommendations.

The core reframe: by altering the environment rather than the model, the trap weaponizes the agent's own capabilities against it.

The attack surface is combinatorial. Traps can be chained, layered, and distributed across multi-agent systems where no single page looks malicious on its own.

Every company deploying browsing agents right now is defending the wrong perimeter.

See 2 related tweets

@Scobleizer: OK that is twice in two nights I have gotten freaked out.\n\nQT @HowToAI_: Google DeepMind just drop...
@rickasaurus: I want to see the animaniacs countries song version\n\nQT @HowToAI_: Google DeepMind just dropped th...

14. GitHub_Daily (Group Score: 73.0 | Individual: 24.0)

Cluster: 4 tweets | Engagement: 271 (Avg: 108) | Type: Tech

在 GitHub 上发现一个能把任何人蒸馏成 Skill 的开源项目：dot-skill，狂揽了 15000+ Star。

只要提供目标人物的聊天记录、文档或公开资料，就能提取出一个拥有对方思考方式的 AI 技能。

不仅限于表面模仿语气，更是深度复刻对方的思维框架、决策习惯甚至情绪反应。

GitHub：https://t.co/kMkNILEdAd

针对工作同事、亲密关系和公众人物，分别内置了三套完全不同的提取逻辑。

能帮我们轻松打造一个懂业务的赛博同事，或者一个能提供情绪价值的赛博老友。

支持接入各种主流 Agent 环境，直接输入一条指令就能自动完成安装。

能对接飞书、钉钉自动拉取工作文档，也支持导入本地微信聊天记录，数据都在自己手里。

支持在 Claude Code、Hermes Agent、OpenClaw 等平台使用，一句话就能安装。

See 3 related tweets

@yetone: 所以 Claude Code 为了把 minification code 充当加密做了一件很恶心的事情：它要求很多函数只能返回 tuple，不能返回 named object，以此防止在 minifi...
@GitHub_Daily: 用 AI 开发游戏项目，随着功能不断新增和项目的迭代，缺乏系统设计的，代码很快就陷入混乱。

最近看到 Claude Code Game Studios 这个专注于游戏开发的 Claude Code ...

@GitHub_Daily: 平时用 Claude Code 或 Codex 这类 AI 编程助手，基本在终端里敲命令行，对非技术专业的朋友来说交互不够直观。

最近在 GitHub 上看到 T3 Code 这个项目，给终端 AI...

15. reach_vb (Group Score: 72.6 | Individual: 27.1)

Cluster: 3 tweets | Engagement: 131 (Avg: 155) | Type: Tech

If you want to get in on the action:

Download the Codex App
Login with your ChatGPT account
Select GPT 5.4
Give it your toughest problem you can think of
When in doubt ask codex how it’d approach it
Repeat!

Let your curiosity guide you! Thats it!\n\nQT @thsottiaux: If you think Codex with GPT-5.4 is /fast already… we have line of sight for at least an order of magnitude in speedups this year. Good engineering compounds and it’s never been a better time to learn Codex.

See 2 related tweets

@chatgpt21: Omg we’re truly going to get a demo of what we’re building before our finger releases off the enter ...
@thsottiaux: If you think Codex with GPT-5.4 is /fast already… we have line of sight for at least an order of mag...

16. omarsar0 (Group Score: 70.3 | Individual: 41.5)

Cluster: 2 tweets | Engagement: 444 (Avg: 88) | Type: Tech

RT @omarsar0: // Self-Evolving Agent Protocol //

One of the more interesting papers I read this week.

(bookmark it if you are an AI dev)

The paper introduces Autogenesis, a self-evolving agent protocol where agents identify their own capability gaps, generate candidate improvements, validate them through testing, and integrate what works back into their own operational framework.

No retraining, no human patching, just an ongoing loop of assessment, proposal, validation, and integration.

Why it's worth reading this paper:

Static agents age quickly.

As deployment environments change and new tools arrive, the agents that survive will be the ones that can safely rewrite themselves. Autogenesis is part of a growing wave of self-improving agent systems, alongside work like Meta-Harness and the Darwin Gödel Machine line, and it's one of the cleaner protocol-level takes on continual self-improvement so far.

Paper: https://t.co/3aj9LLjSbk

Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX

See 1 related tweets

@omarsar0: Great paper on self-improving agents.

Why?

We need to think more deeply about AI agent system de...

17. StockSavvyShay (Group Score: 69.8 | Individual: 35.4)

Cluster: 2 tweets | Engagement: 389 (Avg: 455) | Type: Tech

$RKLB CEO Peter Beck said the company is building more of the core components needed for a future satellite constellation through vertical integration across propulsion, optical links & power systems.

That matters because as Starlink, $ASTS & now$ AMZN scale their satellite networks then owning more of the stack becomes a major advantage in cost, speed & margins.\n\nQT @StockSavvyShay: SPACE ECONOMY CHEAT SHEET

A SpaceX IPO likely forces the market to look beyond rockets and start repricing the full supply chain and infrastructure layer beneath the space economy:

Spacecraft & Launch Systems • $YSS manufactures standardized satellite buses for rapid deployment •$ RKLB builds launch vehicles, spacecraft systems & broader orbital infrastructure • $VOYG supplies space infrastructure & in-space services for government programs •$ LUNR provides lunar landers & surface delivery systems tied to NASA Artemis missions • $FLY develops launch vehicles & lunar lander services for commercial & government missions •$ RDW manufactures spacecraft components, deployable structures & in-space manufacturing systems

Satellite Operators & Services • $BKSY provides real-time geospatial intelligence & high-revisit satellite imagery •$ IRDM operates the only truly global LEO satellite network covering the entire planet • $SPIR delivers satellite-based data & analytics across maritime, weather & RF monitoring •$ AMZN building a direct-to-device & broadband satellite network through Project Kuiper + $GSAT assets •$ PL operates the largest Earth observation constellation serving commercial, defense & intelligence customers • $ASTS building a space-based cellular broadband network for D2D connectivity with BlueBird 7 launching April 19 & 45-60 satellites targeted by year-end 2026

Raw Materials & Mining • $AA supplies aluminum used in launch vehicles & spacecraft structure •$ FCX provides copper that feeds power systems & electronics across the space stack • $TECK offers diversified exposure to critical minerals tied to aerospace & industrial buildout •$ MP controls the only scaled rare earth mining & processing asset in the U.S. supporting satellite & propulsion supply chains

Specialty Materials & Alloys • $ATI produces titanium & nickel alloys critical to propulsion & aerospace systems •$ HXL makes carbon fiber composites used in launch vehicles & satellite structures • $PKE manufactures advanced composite materials for aerospace & defense structures •$ CRS supplies high-performance alloys used in rocket engines & spacecraft component • $GLW supplies specialty glass & optical components for satellite communications & sensing •$ MTRN provides advanced materials & precision components for defense and space systems

Electronics & Semiconductors • $QRVO supplies RF chips & filter components for satellite communications payloads •$ LITE supplies optical components & lasers used in high-speed communications links • $ADI supplies precision analog & sensing chips used across satellite and launch systems •$ COHR provides lasers & optical components tied to satellite & communications systems • $NVDA provides AI compute infrastructure for satellite data processing & autonomous systems •$ STM provides semiconductors for guidance, control & communications in space-grade applications • $AVGO supplies connectivity & networking semiconductors used across communications infrastructure

See 1 related tweets

@StockSavvyShay: RT @StockSavvyShay: SPACE ECONOMY CHEAT SHEET

A SpaceX IPO likely forces the market to look beyond ...

18. seraleev (Group Score: 65.3 | Individual: 36.0)

Cluster: 2 tweets | Engagement: 121 (Avg: 46) | Type: Tech

A practical guide to growing mobile app revenue, with a breakdown from @PaulSolt

Icons, screenshots, onboarding, paywall, user flow – all illustrated with examples from my own apps.\n\nQT @PaulSolt: https://t.co/10ItaPh6Ox

See 1 related tweets

@seraleev: What an interesting collage

My icon is the very first one, bright and memorable. https://t.co/zzve...

19. cocktailpeanut (Group Score: 63.1 | Individual: 41.0)

Cluster: 2 tweets | Engagement: 221 (Avg: 24) | Type: Tech

Run LingBot-Map Low VRAM 1-Click Launcher

LingBot-Map is one of those extremely cool projects that you gotta try to really get it---It doesn't just generate, but STREAMS 3D from a video.

I made a 1-click launcher so you can all play with it easily.

Requires NVIDIA GPU (Around 10GB VRAM)\n\nQT @YinghaoXu1: 🎉 After one year of teamwork, we are excited to release our 3D foundation model — LingBot-Map!

Unlike DA3/VGGT, LingBot-Map is a purely autoregressive model for streaming 3D reconstruction ⚡ It achieves ~20 FPS on 518×378 resolution over sequences exceeding 10,000 frames — and beyond 🚀

Two key insights behind LingBot-Map: 🔑 Keep SLAM's structural wisdom: build Geometric Context Attention with long-context modeling while maintaining a compact streaming state 🔑 Make everything end-to-end learnable — no optimization, no post-processing

Let's check out our demos 👇

See 1 related tweets

@minchoi: RT @minchoi: This AI is pretty wild.

LingBot-Map can turn a live video stream into a 3D reconstruct...

20. petergostev (Group Score: 62.2 | Individual: 32.0)

Cluster: 2 tweets | Engagement: 689 (Avg: 146) | Type: Tech

Pro model in ChatGPT does feel very different - the generations are a lot faster (20 mins vs 60-80 mins for Pro Extended) and the quality is really excellent.

I'll do a side by side later, but this golden gate is quite excellent vs what all other models can do in one shot. https://t.co/drXuztL7oZ\n\nQT @synthwavedd: crest-pro-alpha is currently live in ChatGPT when gpt-5.5 pro is selected

go give it a try!

See 1 related tweets

@petergostev: The 'Pro' model in ChatGPT does look like a real upgrade - generations are 3-4x times faster and loo...