Ask HN: Are open-weight LLMs the new offline encyclopedias?

1•subhrm•3h ago

There was a time when offline encyclopedias were used heavily as internet was not widely available. I used to install britannica encyclopedia and Microsoft Encarta on my PCs (I had no internet at home at that time) and it had helped me a lot between the years 1999 to 2006.

Now we have open-weight LLMs. There are very good models of size < 15GB.

So I’m wondering:

- Are open-weight LLMs effectively becoming the next evolution of offline encyclopedias?

- If so, what replaces the idea of “citation” and “verifiability” in this paradigm, as they do hallucinate time to time ?

Curious how others here are thinking about this shift, especially folks building offline first educational systems.

Comments

latexr•3h ago

No, no they are not. The last thing I want from an encyclopaedia is information I can’t trust, inconsistent and contradictory depending on when and how I search.

thepasch•3h ago

The point of an encyclopedia is that you can visit a very specific page under a very specific name and receive information that you know has been vetted and properly researched. You get precisely zero of any of this with an LLM, so they just seem like they’re fundamentally the wrong tool to even consider something like this for.

The Mythical Agent-Month

The Dark Forest of Minds

CoreWeave Takes as Much Financial Engineering as It Does Datacenter Design

Show HN: Omega Walls–open-source stateful runtime defense for RAG and AI agents

Game Oriented Assembly Lisp

Why Vim uses hjkl keys as arrow keys

Brainstorming Leaves Traces

Pain and Suffering Are Not the Same Thing

The Compression Paradox in AI: Meaning Breaks Before Models Hallucinate

ShadowChat: Extortion-Proof Signal Fork

Let's talk about Animation Quality

Codacy: Code-level detection of AI model, tools and agent usage

KDE Merges Per-Screen Virtual Desktops After 21 Years

Reshard Button

Show HN: Write better Go integration tests with open source dockertest v4

Show HN: The opensource, reliable, scalable Agentic AI framework under 10MB

Why arent Karpathy's LLM coding principles Claude Code defaults?

Why Everyone Is Obsessed with Claude Code – Enrico Tartarotti [video]

Hollywood stars sign open letter protesting the Paramount-Warner Bros. merger

Tax Day? Git add 1040.pdf

Amazon to acquire Globalstar and expand Amazon Leo satellite network

Show HN: DollhouseMCP 2.0, open-source MCP composable AI building blocks

Show HN: I built a Claude Code plugin that ships production-ready SaaS product

Apple and Amazon Ink Satellite Deal Amid Globalstar Takeover

Show HN: I built an agent memory system for myself and got 90.8% on LongMemEval

Show HN: Answer 9 questions, get a Matter firmware with safe boundaries

Building a BLE Sensor Scanner with Nim and BleuIO

The 'Coin Hangover'

The Canary in the Payment Rail

The NPCs That Test Themselves