conv.

All stories
Active · 33h

Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots

Conversation activity · last 34 hours peak 3/30m

Peak 3 items in one 30m at Aug 10, 1 PM; 20 items over 34 hours Aug 10, 1:09 PM — 3 items · Press 2, Hacker News 1Aug 10, 1:39 PM — no itemsAug 10, 2:09 PM — no itemsAug 10, 2:39 PM — no itemsAug 10, 3:09 PM — no itemsAug 10, 3:39 PM — no itemsAug 10, 4:09 PM — no itemsAug 10, 4:39 PM — 1 item · Hacker News 1Aug 10, 5:09 PM — 1 item · Hacker News 1Aug 10, 5:39 PM — 2 items · Hacker News 2Aug 10, 6:09 PM — no itemsAug 10, 6:39 PM — no itemsAug 10, 7:09 PM — no itemsAug 10, 7:39 PM — 1 item · Mastodon 1Aug 10, 8:09 PM — no itemsAug 10, 8:39 PM — 1 item · Hacker News 1Aug 10, 9:09 PM — no itemsAug 10, 9:39 PM — no itemsAug 10, 10:09 PM — no itemsAug 10, 10:39 PM — no itemsAug 10, 11:09 PM — no itemsAug 10, 11:39 PM — no itemsAug 11, 12:09 AM — no itemsAug 11, 12:39 AM — no itemsAug 11, 1:09 AM — 1 item · Hacker News 1Aug 11, 1:39 AM — no itemsAug 11, 2:09 AM — no itemsAug 11, 2:39 AM — no itemsAug 11, 3:09 AM — 1 item · Hacker News 1Aug 11, 3:39 AM — 1 item · Mastodon 1Aug 11, 4:09 AM — no itemsAug 11, 4:39 AM — no itemsAug 11, 5:09 AM — no itemsAug 11, 5:39 AM — no itemsAug 11, 6:09 AM — no itemsAug 11, 6:39 AM — no itemsAug 11, 7:09 AM — no itemsAug 11, 7:39 AM — no itemsAug 11, 8:09 AM — no itemsAug 11, 8:39 AM — no itemsAug 11, 9:09 AM — no itemsAug 11, 9:39 AM — 1 item · Hacker News 1Aug 11, 10:09 AM — no itemsAug 11, 10:39 AM — 1 item · Hacker News 1Aug 11, 11:09 AM — no itemsAug 11, 11:39 AM — no itemsAug 11, 12:09 PM — no itemsAug 11, 12:39 PM — no itemsAug 11, 1:09 PM — 1 item · Hacker News 1Aug 11, 1:39 PM — no itemsAug 11, 2:09 PM — no itemsAug 11, 2:39 PM — no itemsAug 11, 3:09 PM — no itemsAug 11, 3:39 PM — 1 item · Hacker News 1Aug 11, 4:09 PM — 2 items · Hacker News 2Aug 11, 4:39 PM — 1 item · Hacker News 1Aug 11, 5:09 PM — 1 item · Mastodon 1Aug 11, 5:39 PM — no itemsAug 11, 6:09 PM — no itemsAug 11, 6:39 PM — no itemsAug 11, 7:09 PM — no itemsAug 11, 7:39 PM — no itemsAug 11, 8:09 PM — no itemsAug 11, 8:39 PM — no itemsAug 11, 9:09 PM — no itemsAug 11, 9:39 PM — no itemsAug 11, 10:09 PM — no itemsAug 11, 10:39 PM — no items 3 items · 1:09 PM
4 PMAug 118 AM4 PMnow · 11:09 PM

Clustered from 20 items across 3 sources. Not yet parsed — the coverage below is the raw record.

Press coverage 1

Social posts 2

Voices from the web unedited

  • Specifying units seems to be unreliable; I tried adding a description to the set_thermostat temperature: "temperature": { "type": "number", "description": "degrees Fahrenheit" }, Set the living room temperature to 70 degrees Celsius { "function_calls": [ { "name": "set_thermostat", "arguments": { "room": "living room", "temperature": 70, "mode"…

    wkyHacker News12h agoview on Hacker News ↗
  • This is cool. I definitely think the "micro" sized LLM space is underappreciated, so it's always good to see work like this. I foresee a paradigm in some contexts where you have a hierarchy of LLMs, with more competent models actively training smaller models to solve specific tasks very efficiently, and something like this could be the smallest…

    nater5000Hacker News1d agoview on Hacker News ↗
  • It's definitely cool that you can get any reasoning whatsoever out of such a small model. That said, its reasoning is "interesting":Query: "Make the living room dark" Agent: "User wants lights on in living room. 'dark' implies dim. Room 'living room', action 'on'." (And on every test I did, it just completely ignored the "brightness" parameter)It…

    CarpeNecopinusHacker News19h agoview on Hacker News ↗
  • Can you share more about the architectural/design tradeoffs you considered or decided upon? Particularly for me, why is a model that is intended mainly to just make tool calls and marshal the results back focusing on speed? Speed as an inherent result of small size, I get, but speed as a design focus confuses me because it’s simply not going to be…

    ComputerGuruHacker News13h agoview on Hacker News ↗
  • Funny result from the web demo. I'm well aware that it's an extremely small and, well, stupid, model, but even so:Query: HNResult:{ "function_calls": [ { "name": "lock_door", "arguments": { "door": "front door" } } ], "reasoning": "User wants to lock the door. No specific door mentioned, so use 'front door' as default.", "confidence": 0 }I'd…

    TiberiumHacker News1d agoview on Hacker News ↗
  • When I asked it to call my cable company I got this response. I am wondering how this response comes out… why does it say lights twice and not mention door? LLMs are a mystery to me but I suppose this reasoning is at the heart of the model and it’s only 40% confident in it { "function_calls": [], "reasoning": "No tool for making phone calls or…

    bottlero_cketHacker News9h agoview on Hacker News ↗
  • My first query:> Make it a little warmer in here.The reply:> "name": "set_thermostat", > "arguments": { > "temperature": 65, > "mode": "cool", > ... > "reasoning": "'warmer' implies need for cooling; set_thermostat with temperature 65 (typical warmth) and mode 'cool'.",Maybe I'm doing it wrong?

    dbeardslHacker News1d agoview on Hacker News ↗
  • I'm quite impressed by the results of the web demo, especially given its size and the precision with which it uses the three available tools (tested with German commands). I could imagine that this LLM would fit well into a setup with multiple micro-sized LLMs for different purposes; so 14 MB for precise tool invocation is a reasonable memory…

    silentbob7Hacker News21h agoview on Hacker News ↗
  • I don't quite get what this can be used for.It cannot be used for direct customer interaction, it's simply too limited.Nobody wants to say "lock the front door", they would just do itAnd even quite direct commands like "tighten security" are not understood by the model

    hdhcbdbHacker News6h agoview on Hacker News ↗
  • > Like, what does a 28MB binary get you?For one thing, on beefy-enough recent CPUs, you could keep the weights hot in the L2 cache of a single CPU core. (Which is clearly not the use-case, but might be interesting to those looking for extreme TPS numbers. Or perhaps for efficient training!)

    derefrHacker News6h agoview on Hacker News ↗