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

Google Online Security Blog
Google Online Security Blog
S
Security @ Cisco Blogs
Recent Commits to openclaw:main
Recent Commits to openclaw:main
人人都是产品经理
人人都是产品经理
The Hacker News
The Hacker News
W
WeLiveSecurity
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
The Cloudflare Blog
博客园 - 司徒正美
雷峰网
雷峰网
L
LINUX DO - 最新话题
博客园 - 叶小钗
云风的 BLOG
云风的 BLOG
The Last Watchdog
The Last Watchdog
V2EX - 技术
V2EX - 技术
S
Security Affairs
有赞技术团队
有赞技术团队
月光博客
月光博客
T
Threatpost
T
Tor Project blog
O
OpenAI News
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
V
V2EX
Know Your Adversary
Know Your Adversary
Project Zero
Project Zero
博客园 - 三生石上(FineUI控件)
D
Docker
AWS News Blog
AWS News Blog
AI
AI
P
Proofpoint News Feed
K
Kaspersky official blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
D
Darknet – Hacking Tools, Hacker News & Cyber Security
www.infosecurity-magazine.com
www.infosecurity-magazine.com
S
Securelist
F
Fortinet All Blogs
F
Full Disclosure
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
量子位
Hacker News - Newest:
Hacker News - Newest: "LLM"
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
P
Palo Alto Networks Blog
Cyberwarzone
Cyberwarzone
Cisco Talos Blog
Cisco Talos Blog
美团技术团队
N
News | PayPal Newsroom
T
The Blog of Author Tim Ferriss
MyScale Blog
MyScale 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
What AI Doesn't Eat in SaaS
Aany1420 · 2026-05-01 · via Hacker News - Newest: "AI"

It was April 17. I was halfway through my second coffee when my phone buzzed. Someone in our Slack had pasted the Claude Design announcement.

I read it twice. Then I checked the stocks.

Figma had fallen 7% in the day, from $20.32 down to $18.84. Adobe was off about 2.7%. Wix down nearly 5%. GoDaddy down 3%.

The whole design-adjacent neighborhood took a hit on a single product announcement.

A few days before that, Anthropic’s CPO had quietly stepped off Figma’s board. In retrospect, that was the kind of move that only makes sense once you read the headline.

A category that took 14 years to build was suddenly worth a billion dollars less in a single afternoon.

Not because Figma got worse. Because the cost of building a credible alternative collapsed overnight.

I’ve been chewing on that morning for weeks now. Not because I was surprised, exactly.

I run a billing infra company. I think about defensibility every day, and I’ve had this conversation with Nikhil probably 40 times.

But April 17 crystallized something I’d been trying to put words around for a while.

Here’s the thing I want to be careful about, before I keep going. I am not saying Figma is dead. Figma is not dead.

People will still open Figma tomorrow morning. Designers will still ship work in it. Revenue will keep coming in. The product is not going anywhere.

What changed on April 17 is something subtler, and in some ways more important. The capital markets repriced what they think Figma is worth.

The growth story, the multiple, the assumed cash flows 10 years out, that’s what got cut. Not the product or the narrative around the product.

That’s the lens I want to hold for the rest of this piece. Most SaaS companies are not going to die in the next three years.

But a lot of them are going to be repriced, hard, by capital that no longer believes the future cash flows are protected. That repricing has already started. It will keep going.

And the ones that hold their valuations will hold them for three very specific, very unsexy reasons.

Is your core product a creative task, or a generative task?

If it is, design, writing, research, code, decks, marketing copy, illustrations, video, your capital story is the one most exposed.

Not your product but your story.

I’d give most of those categories 24 to 36 months before the market decides the future cash flows are no longer worth what they were paying for them.

Look at what’s already happened.

Chegg dropped 48% in a single trading day in May 2023 after admitting on its own earnings call that students were using ChatGPT instead of paying for their study tools.

The stock is down roughly 99% from its pre-ChatGPT peak. Chegg the company still exists. Chegg the capital story is gone. That’s a $14.5B repricing in 36 months by a free chatbot.

Grammarly, less dramatic but the same shape. YoY growth dropped from about 43% at $500M ARR in 2022 to roughly 10% at $650M ARR in 2024 as ChatGPT and bundled AI writing tools chewed at the standalone-grammar-checker thesis.

They’re still a real company, they acquired Coda and they raised a billion in non-dilutive financing.

But the growth story, the thing capital was actually paying for, cracked.

Here’s the part about Grammarly that I think matters more, though, and we’ll come back to it later.

Grammarly is not going anywhere. Because it’s installed across Fortune 500 companies in a way that’s painful to rip out.

It sits inside Slack, inside Gmail, inside every internal comms surface. You don’t uninstall Grammarly across 40,000 employees because Anthropic shipped a model.

You uninstall it when the IT director retires and someone else takes the call. That’s the kind of thing that buys a company another decade. Hold that idea, it’s the thing that actually saves you in this market.

And on the other side of the same coin is Cursor. It went from a 2022 student project to a $29B valuation by November 2025, and is reportedly raising at $50B+ as I write this.

The capital flowed out of the old wrappers and into the new substrate. That’s the move you should be watching.

The pattern is consistent. Wherever the output is “make me a thing that looks right,” the model becomes the product, and the wrapper’s multiple compresses toward zero.

Sometimes the fastest-moving wrapper becomes the new model and the multiple goes the other way.

Either way, the middle of the market, the careful incumbents whose moat was design and brand and a five-year head start, is where the air is leaving the room.

So what doesn’t get repriced?

The first cohort that survives is the boring one. Think like Stripe, Airwallex, Plaid and Adyen etc.

Claude can write you a payments processor in an afternoon. Cursor can probably ship one in a weekend.

Neither of them can give you 15 years of acquired bank relationships, PCI DSS Level 1 certification across forty jurisdictions, the licenses, the compliance staff, the legal precedents, the liability coverage.

Stripe’s moat isn’t code. Stripe’s moat is 15 years of regulatory compounding. Every audit they’ve passed, every license they’ve earned, every edge case their lawyers have argued, that’s the product. The API is the friendly part on top.

You cannot vibe-code your way past a banking regulator. I keep saying this to founders and most of them nod and then go build a creative tool anyway.

The second cohort is the one I think about most, because it’s where Flexprice lives.

If your product holds the customer’s revenue data, usage data, contract terms, entitlements, credit balances, and the historical record of how all of those have moved over the last 36 months, you are not getting replaced over a weekend.

The migration cost alone is brutal.

Think about what billing actually is. It’s not just “send an invoice.”

It’s the system of record for how much money your customer is making, how their pricing has evolved, which deals have custom terms, what counts as usage, what doesn’t, and how every one of those decisions reconciles back to the GL.

36 months of that data, owned correctly, becomes the company’s central nervous system.

You don’t rip that out because Anthropic shipped a new model. You rip it out when the entire company is being acquired, and even then it takes a year.

The same logic holds for CRMs with eight years of pipeline history, ERPs with decade-old chart-of-accounts mappings, observability platforms that have indexed billions of historical events. The data has weight. Weight is a moat.

A wrapper does not have weight. A wrapper has a prompt and a logo.

This is the one Nikhil and I argue about the most. I think it’s the most underrated of the three.

Some products can afford to be 95% right. A chatbot. A first-pass email. A draft design. If it’s wrong, you regenerate. The cost of being wrong is one click and four seconds.

Now imagine your billing system is 99.9% right.

If you’re a $100M revenue company, that 0.1% delta is $100,000 a year. If you’re a $1B company, it’s a million. That’s not “regenerate and try again.” That’s a board-level conversation.

That’s a finance team that no longer trusts the system. That’s six months of forensic accounting to figure out what got missed.

The same logic applies to anything mission-critical. Health records. Tax filing. Trade execution. Identity verification. Air traffic control, obviously, but also the much wider category of B2B systems where the margin of error is paid for in dollars and lawyers, not user complaints.

You can’t vibe-code these. Not because the code is hard, but because the consequences of being wrong don’t get socialized away by “the AI did it.”

Someone is on the hook. Someone is signing the SOC 2. Someone is the one explaining to the auditor why the recognized revenue doesn’t tie out.

I’m not pretending I know exactly which companies sit in which bucket. The lines are blurry.

  • Notion has deep customer data and a creative product surface.

  • Salesforce has both compliance and data, and also a thousand surface areas where AI is going to chew at the edges.

  • Stripe has an enormous compliance moat and is simultaneously building so much of its own AI surface that they might end up eating their platform partners.

What I’m saying is that the three filters, compliance depth, data ownership, and cost-of-being-wrong, are the only filters I trust right now.

If I were starting a company in 2026 and someone pitched me an “AI-powered presentation tool” or an “AI-powered marketing copy generator,” I’d politely change the subject. Those are not companies. Those are features inside someone else’s roadmap.

If someone pitched me an AI-native system that lives inside the customer’s revenue stack, owns their contract data, and becomes the source of truth for what they bill, I’d want to invest. That’s a company.

A friend in Bangalore pinged me last month asking whether his AI-powered legal contract drafting startup was defensible. I told him it depends. If he was competing with Claude on draft quality, no. If he was competing with Claude on “we sit inside the customer’s contract repository, integrate with their CLM, and own nine months of their negotiation history,” yes. The product is the substrate, not the generation.

That’s the lens. Not “is AI in my product?” but “what holds its multiple when the model improves 100x and costs a tenth?”

Compliance holds. Data and workflow weight hold. Margin-of-error pain holds.

Most everything else is going to feel what Figma felt on April 17. The product will still be there. The capital will not.

I’ll be honest, this isn’t a comforting framework. It’s the one I’m using to make actual bets, including the bet I’m making with my own years on Flexprice. We are deliberately picking the boring middle.

Deep customer data. Mission-critical. The kind of system you don’t replace because the math doesn’t math.

If that sounds defensive, it’s because it is. The defensible part is the whole point.

—Manish

Thanks for reading The Founder’s Draft! And if you liked it, then do share it with your network

Share

Two things I’ve been reading and genuinely loved!

Yash Tekriwal on LinkedIn this week calling out “build vs. buy” as a false dichotomy. His point is that “build” is never just build, it’s build plus maintain plus debug plus document plus staff plus the context-switching your engineering team has to do away from the core product they’re supposed to be selling. He’s right, and it’s the part most founders are underestimating right now, especially when “just vibe-code it” makes building feel free. It’s not free. It’s the cost of focus, and focus is the single most expensive thing a young company has.

Check it out!

Ivan Landabaso’s Startup Riders breakdown of how Granola went from $0 to a $1.5B valuation in three years. The line that made me stop and read it twice was about why Granola ran frontier models from day one even when the unit economics didn’t work “running expensive products today is a temporary disadvantage that becomes a permanent advantage.”

Check it out!

No posts