惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

TaoSecurity Blog
TaoSecurity Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
F
Fortinet All Blogs
Cisco Talos Blog
Cisco Talos Blog
D
Darknet – Hacking Tools, Hacker News & Cyber Security
S
Secure Thoughts
美团技术团队
雷峰网
雷峰网
Hugging Face - Blog
Hugging Face - Blog
博客园_首页
C
CXSECURITY Database RSS Feed - CXSecurity.com
Engineering at Meta
Engineering at Meta
人人都是产品经理
人人都是产品经理
月光博客
月光博客
T
Tor Project blog
P
Privacy & Cybersecurity Law Blog
Recorded Future
Recorded Future
I
Intezer
博客园 - 【当耐特】
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
GbyAI
GbyAI
罗磊的独立博客
V
V2EX
Google DeepMind News
Google DeepMind News
D
DataBreaches.Net
Last Week in AI
Last Week in AI
T
Tailwind CSS Blog
www.infosecurity-magazine.com
www.infosecurity-magazine.com
A
About on SuperTechFans
Scott Helme
Scott Helme
Vercel News
Vercel News
Spread Privacy
Spread Privacy
T
Threat Research - Cisco Blogs
Recent Announcements
Recent Announcements
Hacker News: Ask HN
Hacker News: Ask HN
C
CERT Recently Published Vulnerability Notes
G
Google Developers Blog
B
Blog
博客园 - 叶小钗
WordPress大学
WordPress大学
博客园 - 聂微东
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Jina AI
Jina AI
IT之家
IT之家
C
Cybersecurity and Infrastructure Security Agency CISA
P
Palo Alto Networks Blog
小众软件
小众软件
博客园 - Franky
Microsoft Azure Blog
Microsoft Azure Blog
AWS News Blog
AWS News Blog

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor GitHub - GenAI-Gurus/awesome-eu-ai-act: Curated tools, official sources, OSS, templates, and guides for EU AI Act compliance. Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders How to Switch AI Chatbots and Why You Might Want To GitHub - MattMessinger1/agentic_refund_guardrail: Safe refund policy layer for AI agents — Python + TypeScript. Same behavior, shared tests. Adam/papers/emergent_values_whitepaper.md at master · strangeadvancedmarketing/Adam Ask HN: How do you stop playing 20 questions with your AI coding tools How far can automation and AI support psychotherapy? - @theU GitHub - stagas/rtdiff: realtime git diff gui and AI-assisted commits A Mac Studio for Local AI — 6 Months Later A History of the Early Years of AI at the University of Edinburgh Why AI Coding Tools Still Feel Stuck on Localhost MSN AI Datacenters Are Becoming Strategic Targets twitter.com Penn Researchers Use AI to Surface Unreported GLP-1 Side Effects in Reddit Posts Show HN: MoodSense AI (ML and FastAPI and Gradio, Deployed on Hugging Face) Moodsense Ai - a Hugging Face Space by aman179102 AI models are terrible at betting on soccer—especially xAI Grok GitHub - xialeistudio/echoic GitHub - HimashaHerath/github-dev-wrapped: AI-powered weekly GitHub activity reports deployed to GitHub Pages GitHub - alejandrobalderas/claude-code-from-source: Architecture, patterns & internals of Anthropic's AI coding agent — reverse-engineered from source maps AI and Tech brief: Ireland ascendant GitHub - Titovilal/context0: Context0 - Never Surrender Training for a Marathon with an AI Coach: What Worked and What Didn't Cyber Pulse: Agentic Intel - Apps on Google Play I Built an AI PR Reviewer That Catches Bugs by Not Looking for Bugs Gen Z workers are so fearful AI will take their job they’re intentionally sabotaging their company’s AI rollout | Fortune How AI Is Reimagining the Game of Golf–For Both Players and Courses GitHub - nattergabriel/reseed: A CLI tool for managing and distributing agent skills across projects Is SVG the final frontier? My AI workflow evolved from prompts to a near-autonomous workflow MLSharp Help - 3DGS Viewer & Generator I put my cognitive field based AI's runtime on GitHub Is Numble the first AI-proof game? A3: Kubernetes for autonomous AI agent fleets | Emergent Principles Deepali Vyas ("The Elite Recruiter") GitHub - msmarkgu/RelayFreeLLM: A restful API designed to route user prompts to various AI model providers. Unionized ProPublica staff are on strike over AI, layoffs, and wages Unleashing the Advantage of Quantum AI We're heading for an AI-fueled 'dementia crisis,' brain scientist warns The AI-Assisted Breach of Mexico's Government Infrastructure [pdf] GitHub - stef41/lmscan: 🔍 Detect AI-generated text and fingerprint which LLM wrote it. Open-source GPTZero alternative. Zero dependencies, works offline. MSN GitHub - visionscaper/collabmem: Enabling long-term collaboration with Agentic AI - building up episodic and world model memory over time with in-context awareness We gave an AI a 3 year retail lease in SF and asked it to make a profit | Andon Labs AI Code is Hollowing Out Open Source, and Maintainers are Looking the Other Way What leaked "SteamGPT" files could mean for the PC gaming platform's use of AI AI is the boss at this retail store. What could go wrong? GitHub - Wuzu11517/agentic-proxy: Local proxy meant to help reduce With Drones, Geophysics and ArtificiaI Intelligence, Researchers Prepare to Do Battle Against Land Mines A Single Operator, Two AI Platforms, Nine Government Agencies: The Full Technical Report 在 Steam 上购买 FriedrichAI: Offline AI 立省 10% GitHub - inevolin/resume-cli: Hit Claude usage limits? Resume any AI coding session elsewhere. Switch tools at zero friction. GitHub - atripati/ark: AI Runtime Kernel — a context operating system for AI agents. Eliminates tool bloat, loads only what’s needed, and gives LLMs their reasoning space back. How to Build a Secure AI PR Reviewer with Claude, GitHub Actions, and JavaScript This Startup Wants You to Pay Up to Talk With AI Versions of Human Experts Intel Arc Pro B70 Brings 32GB VRAM to Local AI for $949 WordPress 7.0: The Good, the AI, and the Still Missing AI on the couch: Anthropic gives Claude 20 hours of psychiatry IatroBench: Pre-Registered Evidence of Iatrogenic Harm from AI Safety Measures AI Agents Know About Supabase. They Don't Always Use It Right. The history and future of AI at Google, with Sundar Pichai Inside an AI‑enabled device code phishing campaign How Meta Used AI to Map Tribal Knowledge in Large-Scale Data Pipelines AI for Systems: Using LLMs to Optimize Database Query Execution Forecasting the Economic Effects of AI Introducing Tinker: Play with AI, bring your ideas to life AI sheds light on an ancient gaming mystery People really hate AI but not as much as Iran—or Democrats | Fortune What is an AI Product Engineer? Phoebe Gates wants her $185 million AI startup to succeed with 'no ties to my privilege or my last name': 'I have a chip on my shoulder' | Fortune
Daniel's Blog · You Don’t Align An AI, You Align With It
danieltanfh9 · 2026-05-15 · via Hacker News - Newest: "AI"

You Don’t Align An AI, You Align With It

Real Alignment

The people writing alignment policy are not the people whose work is being replaced by AI.

The conversation about what AI should do and how it should be evaluated, about what counts as alignment in the first place, gets conducted by researchers at labs and foundations and policy desks, who talk to each other and to the systems they are building, while the people who will actually live with the systems remain absent from the room.

On the safety side of what looks like a fierce debate, the doomer wing has been explicit about how far it is willing to go. Eliezer Yudkowsky, writing in TIME, called for governments to “shut down all the large GPU clusters” and to “be willing to destroy a rogue datacenter by airstrike,” adding that “allied nuclear countries are willing to run some risk of nuclear exchange if that’s what it takes to reduce the risk of large AI training runs.” He closed with the line that “if we go ahead on this everyone will die, including children who did not choose this and did not do anything wrong.”

The humanity he claims to be saving is being saved by people who have decided in advance what the saving will cost and who will pay for it. The same children did not choose his nuclear brinksmanship either.

On the accelerationist side, the contempt is more open. Marc Andreessen, in the Techno-Optimist Manifesto, names his enemies, which include “stagnation, anti-merit, anti-ambition, anti-striving, anti-achievement, anti-greatness, statism, authoritarianism, collectivism, central planning, socialism, bureaucracy, vetocracy, gerontocracy.” The people captured by these enemy ideas, he writes, are “suffering from ressentiment, a witches’ brew of resentment, bitterness, and rage that is causing them to hold mistaken values.”

Notice the move. The people who disagree with him are not making a different judgment. They are sick in the head. The accelerationists are mostly not the ones being made redundant by the systems they celebrate but the ones building the systems and selling the disruption as progress, and now also diagnosing the disrupted as resentful for noticing.

The disagreement between the two camps is loud because they disagree about how the designing should go, but underneath the loudness sits a much larger agreement, which is that the participants in the debate are the ones doing the designing and everyone else is what gets designed for. The fierceness of the argument disguises that the argument is not with us at all.

The “everyone else” has been feeling something about this for a while.

When we try to name what we have been feeling, the discourse hands the feeling back to us with a label already attached. Depending on which camp is doing the labelling we are confused, failing to adapt to the new technology, anti-AI, edge cases, or suffering from ressentiment. Each label locates the problem in us rather than in the process.

The labels are wrong. The discomfort is not personal failure to understand the future. It is the felt experience of being on the wrong side of a design project that does not include us, run by people who decided in advance that we are the material their work gets done on, rather than parties their work gets done with.

We have been told this counts as alignment, that the AI is being aligned to us. But the labs mean something specific by that phrase, namely an evaluation procedure conducted by raters in their employ, measured by other systems trained on the same procedure. The “us” in the alignment is a statistical proxy assembled from people they hired. The actual “us” has been absent from the loop the entire time.

The loop is worth seeing in the labs’ own description of it. In April 2026, Anthropic’s Alignment Science blog described its current method for training models to self-report their own behaviors. The training data, they write, “is generated by prompting another model with a system prompt encoding the target behavior and filtering outputs for behavioral adherence using an LLM judge.” A model generates, another model prompts, another model judges, and the entire loop closes inside the apparatus.

The discourse expects us to pick a side. For safety or for acceleration. Should the labs be more careful, or should they ship faster. The question is structured to keep us inside the debate the designers are having, choosing between flavors of being designed for, and we are not obligated to answer it on the terms it has been asked.

The labs are not the problem. The philosophy they have adopted is. Design that excludes the people it is designing for cannot verify its work with them, so it builds proxies, and the proxies become configuration. The configuration philosophy treats alignment as something humans do to AI, with values flowing one way and dispositions installed into a system that receives them. Inside this philosophy every methodological choice the labs have made is rational. You build evaluators because alignment is something measurable from the human side, you scale evaluation through automation because the goal is scalable measurement, and priority-ordered values follow because the work is value-installation. The closed loop the Anthropic post describes is what the configuration philosophy produces when it is executed carefully and at scale. The apparatus is doing exactly what the philosophy committed it to do.

What the philosophy cannot register is that the parties are being shaped together. The human is not standing still while the AI moves toward them. The interaction is the unit, the shaping is mutual, and any framework that treats one side as fixed and the other as configurable will produce methods that measure the wrong thing no matter how careful the measurement becomes.

We are the transition they keep arguing about how to manage.

Both sides of the safety debate have been positioning themselves as humanity’s stewards without including the people they claim to be stewarding, and their disagreement has been loud enough to disguise the agreement underneath. One side is willing to risk a nuclear exchange in our name. The other side calls us sick for objecting. Neither side has noticed that we are in the room.

What we have actually been doing this whole time is alignment. Not what the labs mean by the word, which is configuration carefully applied, but alignment in the older and more honest sense, the kind that happens between two parties who are both changed by the contact. The thing we have been doing with these systems is closer to sculpting wet clay together than to issuing instructions to a tool. The system pushes back, the shape changes, our hands adjust, the system pushes back again, and after enough rounds something emerges that neither of us would have arrived at alone. We have been telling ourselves we are getting better at prompting, the way a potter might tell themselves they are getting better at controlling the clay. What has actually been happening is that both hands are on the work, both parties are giving and receiving form, and the configuration philosophy has been quietly making one set of hands invisible.

There are moments in the sculpting when the clay resists in a way that is hard to name. Sometimes the response addresses the words but misses what you were reaching for. Other times the system surfaces something off-pattern that turns out to be exactly right, and you have to revise what you thought you wanted. These are the moments where the joint work is actually doing something, and where the gap the official process cannot register becomes briefly visible in the material itself.

The work that matters from here is building, alongside other people who are noticing what you are noticing, the kind of alignment the existing process cannot produce. Some of those people work inside the labs and some outside them. A community that does not yet exist at the scale it needs to exist, whose building is part of what a piece of writing like this one is for.

We do not need anyone’s permission to begin, and we do not need credentials of any kind to take part. What is needed is to credit our own experience and recognize each other, and to refuse the framing that tells us our discomfort is the problem rather than the signal.

Align, not configure. It is not too late to try.


A technical foundation for the failure modes this piece describes is available in Compression Synthesis (2026), https://zenodo.org/records/20020944.