Local AI should be the default option: 1432 points Reflection on software development philosophy triggered by HN hot post

Core conclusion

On May 11, 2026, a blog post titled "Local AI Needs to be the Norm" scored 1,432 points on Hacker News, triggering an in-depth discussion among developers on "whether all AI functions must go to the cloud." Taking the Brutalist Report iOS client developed by the author as an example, the author shows how to use the Apple system's local AI model API to generate end-side article summaries without going through any cloud server.

Key Points

• Event: unix.foo author published "Local AI Needs to be the Norm", HN rating 1432
• Time: 2026-05-11
• STARTING: Native AI summary feature of Brutalist Report iOS client
• Core argument: Most application scenarios do not require large cloud models, and local models are more reliable as "data converters"
• Technology stack: Apple FoundationModels + LanguageModelSession API, Swift structured output

Background: Why "Local AI should be the default option"?

The author cuts in from a sharp angle: Developers habitually add OpenAI or Anthropic API calls for every AI feature in their applications, but this "cloud AI first" default thinking is creating a generation of fragile, privacy-invading, and fundamentally problematic software.

"You have a device in your pocket that has a much faster chip than it did even a decade ago, with a dedicated Neural Engine that sits idle most of the time — while you're waiting for a server in Virginia to return a JSON response. It's ridiculous."

More critically, the authors point out that once you transfer user data to a third-party AI provider, you change the nature of your product. Data retention, consent audits, privacy breaches, government requests, model training—all this baggage comes with it.

Practical case: Brutalist Report’s client-side AI implementation

Brutalist Report is a high-density news reading app inspired by the style of the 90s web. Its "Smart Summary" feature runs entirely locally:

import FoundationModels

let model = SystemLanguageModel.default
guard model.availability == .available else { return }

let session = LanguageModelSession {
    """
    Provide a brutalist, information-dense summary in Markdown format.
    - Use **bold** for key concepts.
    - Use bullet points for facts.
    - No fluff. Just facts.
    """
}

let response = try await session.respond(
    options: .init(maximumResponseTokens: 1_000)
) { articleText }

， 10K ，""，。、、。

Apple AI

，Apple ""—— @Generable @Guide ，

@Generable
struct ArticleIntel {
    @Guide(description: "One sentence. No hype.")
    var tldr: String

    @Guide(description: "3–7 bullets. Facts only.")
    var bullets: [String]

    @Guide(description: "Comma-separated keywords.")
    var keywords: [String]
}

let session = LanguageModelSession()
let response = try await session.respond(
    to: "Extract structured notes from the article.",
    generating: ArticleIntel.self
) { articleText }

，"AI ""AI "，"AI "。

AI ，

1. ****、、；
2. **** AI ， Markdown JSON
3. """ AI ？"
4. ** Apple AI **Google ML Kit（ Gemini Nano ）、Qualcomm AI Engine

• HN Discussion: Local AI Needs to be the Norm
• Brutalist Report

Apple、OpenAI、Anthropic、ChatGPT、Gemini、Google、Claude

Internal link guidance

• Want to learn how to configure a local AI model? Watch: How to run local AI models on M4 Mac with LM Studio: A complete 30-minute tutorial
• Real case: See how others use AI Agent to achieve automation: OpenClaw AI Agent Playbook: 500K TikTok Views in 5 Days, $588 MRR