frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Illinois Just Passed a Law That Puts Linux on the Hook for Age Verification

https://linuxstans.com/illinois-hb5511-operating-system-age-verification/
175•speckx•1h ago•183 comments

Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model
940•riordan•11h ago•526 comments

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

https://cactuscompute.com/needle
57•HenryNdubuaku•4h ago•29 comments

Sonic Pi v5

https://www.patreon.com/samaaron/posts/sonic-pi-v5-166001392
248•samaaron•3d ago•66 comments

Rust SIMD on the GPU

https://www.vectorware.com/blog/simd-on-gpu/
75•sagacity•3h ago•36 comments

Publishing Schematics Before "Open Source" Was a Word

https://fabscene.medium.com/publishing-schematics-before-open-source-was-a-word-55-years-of-akizu...
22•extralongdivisi•3d ago•2 comments

Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models

https://www.ft.com/content/4e3957f8-ea7c-4c46-a3de-cdce8e526878
254•root-parent•7h ago•313 comments

Launch HN: Stoa Markets (YC S26) – A Marketplace for GPUs and AI Servers

https://www.stoaexchange.com
58•erenberke•5h ago•33 comments

Exploiting System Management Mode with a very long interrupt

https://github.com/xoreaxeaxeax/smiiiiiiiiiiiiiiii
103•WhiteDawn•5h ago•33 comments

The Psychedelic Toad of the Sonoran Desert

https://en.wikipedia.org/wiki/Bufo_Alvarius:_the_Psychedelic_Toad_of_the_Sonoran_Desert
51•simonebrunozzi•6d ago•35 comments

How Claude marks AI-generated content

https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content
5•mfiguiere•19m ago•0 comments

Squeak 6.1

https://squeak.org/release_notes/6.1/
192•fniephaus•9h ago•97 comments

Stop Killing Games: It's time to sue Sony, join us

https://www.massaschadeconsument.nl/collectieve-acties/playstation/
28•EDM115•1h ago•4 comments

Ask HN: In your experience, what are sound conventions for e-ink UI development?

113•BoxOfRain•3d ago•39 comments

Amazon backs power plant that may become top source of US climate pollution

https://arstechnica.com/tech-policy/2026/08/amazon-funds-biggest-gas-power-plant-in-us-despite-cl...
19•pjmlp•30m ago•7 comments

Humanising LLM Outputs Is Dumb

https://kuber.studio/blog/Reflections/Humanising-LLM-Outputs-is-Actually-Dumb
99•kuberwastaken•8h ago•55 comments

Magnitude 7.4 Earthquake – 5 km S of San José del Palmar, Colombia

https://earthquake.usgs.gov/earthquakes/eventpage/us6000tjl2/executive
140•Bender•6h ago•49 comments

Mistral Patent for “Code implemented tool calls”

https://patentsgazette.uspto.gov/week26/OG/html/1547-5/US12670045-20260630.html
192•theanonymousone•8h ago•157 comments

Parametron: 50s Japanese computer that uses neither transistors nor vacuum tubes

https://ethw.org/Milestones:Parametron,_1954
158•xeonmc•11h ago•43 comments

Letter to Governor Abbott on responsible AI infrastructure in Texas

https://openai.com/index/responsible-ai-infrastructure-texas/
72•hackerBanana•7h ago•136 comments

Exploring Claude/GPT Knowledge Cutoffs and Pre-Training Timelines

https://blog.sshh.io/p/exploring-claudegpt-knowledge-cutoffs
76•sshh12•7h ago•11 comments

Show HN: Higher-dimensional lattices unfolded into the 2D plane

https://number-garden.com/?@THLP@
4•unitX•4d ago•1 comments

Extreme 220GHz+Broadband Silicon Capacitor X2SC 0201M 22nF BV11

https://pim.murata.com/asset/pim4/siliconCapacitor/SICAP_X2SC422522_PDF_SILICONCAPACITOR
51•peter_d_sherman•6h ago•19 comments

Tail-call optimization in C is relatively recent (2025)

https://lwn.net/Articles/1034703/
108•prakashqwerty•10h ago•100 comments

Tl;dv: Over 180k meetings left wide open

https://bobdahacker.com/blog/tldv-hack
483•colesantiago•9h ago•165 comments

Back to the Future of Handwriting Recognition (2016)

https://jackschaedler.github.io/handwriting-recognition/
31•at1as•5h ago•8 comments

Mars Bar from 1991 found – and it's 20g bigger than today's

https://www.bbc.com/news/articles/c1j1kjy7gewo
262•RickJWagner•6h ago•401 comments

50k Boat Names

https://www.beautifulpublicdata.com/boat-names/
139•jonathanmkeegan•8h ago•102 comments

Docker Sandboxes – Disposable, isolated sandboxes for AI agents

https://www.docker.com/products/docker-sandboxes/
607•etoxin•15h ago•339 comments

The Tragedy of the Cognitive Commons

https://arxiv.org/abs/2607.29380
79•jmintz•6h ago•62 comments
Open in hackernews

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

https://cactuscompute.com/needle
56•HenryNdubuaku•4h ago
Hey HN,

Henry from Cactus here!

We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2.

The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges 300-700 on sub-$200 phones such as the Samsung A-Series.

On the tool call and mobile device use benchmarks, Needle 2 trades wins with closest small models like LFM2.5 230M and Apple Foundation Model, at 5x to 70x smaller, both at f16 vs Needle 2 at 2bit. Needle is based on Simple Attention Networks from our paper (https://arxiv.org/abs/2607.18363).

Edge AI has lately meant Macs and PCs, but that is just 1.5 billion of over 21 billion connected IoT devices in the world today, and in emerging markets most phones ship under $200, no NPU, cheap GPUs. These include budget phones, Raspberry Pis, microcontrollers, wearables, small robots like Reachy Mini, and connected home devices.

A conventional transformer of Needle's width and depth spends 164 MFLOPs per token, and even one squeezed down to Needle's parameter count spends 87, Needle spends 70. Even on a high-end phone, an always-on assistant lives inside a power budget; every MFLOP is milliwatt-hours, and Needle spends 7x to 85x fewer of them per token than the smallest performant LLMs. More about the architecture in the link.

When we structure intelligence for consumer devices as functions with typed parameters, the only hard part is mapping a messy sentence onto them; which function, with which values. Our research found that when framed that way, the problem needs no world knowledge and no open-ended prose, which is why 45M parameters suffice.

Needle 2 expands to structured extraction where the schema can be passed in-place of tools and the model returns structured output. You can use Needle as a text-classification model with an enum field, as a summarization model by providing a schema that extracts key fields, everything but free-range decode.

Every product has its own tool vocabulary and fine-tuning needle helps it achieve frontier-level performance on custom tasks, so using the python package (https://github.com/cactus-compute/needle), Needle can be fine-tuned Needle on a Mac/PC in minutes to a few hours, with automated data-generation pipeline, just pass a couple samples.

Nonetheless, every response carries a learned confidence score based our Cactus Hybrid technique. If above your threshold, act, below it, escalate to the cloud or bigger model. Combining Needle 2 with a private DeepSeek-v4-Flash deployment works particularly well for enterprise-level tasks at barely any cost, we can help with this setup.

We have put a lot of thoughts into Needle 2 but might still be missing quite a lot, please use the playground in the provided link to test Needle and share your thoughts, always appreciated!

Comments

tolugenius•59m ago
This is really cool, I'm curious how much knowledge can their be in smaller models? It seems the current trade off is you need sizeably larger models for more performance but I'm curious if in your work how far this is true, as edge ai is really what needs to get better before physical ai can take off (my two cents).
msdz•52m ago
I imagine at such a low parameter count, there would be little to no world knowledge whatsoever, and the entire focus is on getting the structure of tool calling etc. right…?

But yeah, in terms of “physical” AI, robotics definitely comes to mind for me as well, where tool calls/structured “device” use in a “realtime”/edge application are highly beneficial (if you wanted to go with LLMs), but beefy hardware can’t be easily used.

Tiberium•49m ago
Funny result from the web demo. I'm well aware that it's an extremely small and, well, stupid, model, but even so:

Query: HN

Result:

{ "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 expect it to at least ignore (call no tools) for the queries that it doesn't understand. And it seems like it does do that, just not consistently.

jszymborski•41m ago
no, this is the appropriate response to hearing the words "HN" :P
Schiendelman•
40m ago
Was that the first message you sent it?
yoavm•32m ago
The website says the model is for "tool calling, device use, and structured extraction". Your example just doesn't seem to be very relevant. FWIW, it did a pretty good job for tool calling when I tried it, and I think it could be pretty nice to have this running on locally and integrate with Home Assistant.
evmaki•21m ago
False positives are definitely relevant and worth measuring - natural language interfaces always have a discoverability problem, i.e., users not knowing what actions the system does and does not support. If the frontend of that system lacks the ability to reject unsupported commands, weird stuff happens.

Nonetheless, this is very cool work! If I can offer a small suggestion to the team at Cactus, it would be to evaluate your releases on some usability criteria (including false positives). Any serious integrator or adopter of these models would want to have that information available.

jdknezek•15m ago
> "confidence": 0

OP and the linked page talk about the confidence score and using it as an action threshold, so it looks like an appropriate total response to me.

evmaki•4m ago
Right, but that's not the same thing as a benchmark across a test set. It doesn't help me determine how well the model does across a decently-large sample size of commands. It doesn't tell me with what reliability the confidence will be below a given threshold when it should be, above that threshold when it should be, etc.
petu•18m ago
"confidence": 0, so I guess you could threshold it
hmokiguess•14m ago
yeah I got the same, almost like its biased heavily towards that as the 0 ranking -- my prompt was just the word 'potato'
arthuqa•29m ago
That's really cool - I was already thinking of compressing `functiongemma-270m-it` down to 1-2 bits so it would work flawlessly in the browser. Your `Fine-tuning` feature is even much more convenient.
HenryNdubuaku•17m ago
Thanks, give it a splin!
dofm•26m ago
Naïve and clumsy question: how would you pair this with speech-text-speech stuff, wake words etc.? Are there good examples of this for a Pi 5?

The demo is super — I'm just having trouble seeing the whole picture for e.g. a screenless device.

HenryNdubuaku•15m ago
Users often stack a transcription model on top to get the voice prompt, then decode to actions. Think of Alexa and Siri.
nater5000•15m ago
The best entrypoint is Home Assistant: https://www.home-assistant.io/

That will get you a lot further than what you're asking, but if you dig a bit through Home Assistant features, resources, etc., you may find the current "best" answers to your questions.

If you want a quick answer: Whisper is a good open-source speech-to-text model which comes in a variety of sizes (https://huggingface.co/openai/whisper-tiny). You can definitely get something like this running on a Pi 5. There are plenty of other STT models out there, some of which are built specifically for this context (again, see the Home Assistant stuff), but Whisper comes up a lot as a good default choice.

So with something like Whisper, you could just have a simple script which is constantly listening to a rolling window of audio and transcribing it. When the transcription includes a key phrase, you can pass the rest of the transcription to Needle2 (or anything else for that matter). From there, you take the results and execute the necessary tool calls.

There's a bit more to all of this to make it work smoothly, but fundamentally this is all there is to it. All this would work very fast on a Pi 5 (although I wouldn't expect the results to be particularly good without some serious hand-crafted logic, fine-tuning, etc.). If you want to mess around this stuff, handing all of this to Claude, Codex, etc., can get you something spun up and functional very quickly.

varispeed•23m ago
What is the difference between this and random sentence generator?
HenryNdubuaku•15m ago
Random sentence is not a function call.
actionfromafar•14m ago
Ask it to lock a door for instance. It seems to convert simple instructions to reasonable tool calls. Check its confidence score.
minimaltom•16m ago
Was really cool to see yous use Engrams to cut down compute!

Given its basically an O(1) lookup with disk space being the main constraint, I was curious if you've tried ablating engram layers and sizes across your setup?

Also, why mHC over attention residuals?

HenryNdubuaku•12m ago
Yes, we ablated Engrams rigorously and found that it returned world knowledge like FFN without without compute expenditure.
nater5000•6m ago
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 layer in that stack.

With that being said, the web demo is not particularly impressive. It really doesn't like anything I throw at it. I'm fine with accepting that fine-tuning is the solution to this, but I wonder if there's anything to gain from a bigger model? I know it's completely counter to the whole point of this, but a 14MB binary using 28MB of RAM seems unnecessarily small and pretty arbitrary.

Like, what does a 28MB binary get you? Or a 140MB binary? Or a 1.4MB binary? I'm guessing the choice of 14MB came from minimizing the size as much as possible while meeting certain requirements/performance expectations, but even a Pi 5 has plenty more room to spare. Curious if there's a good explanation for this (which I may have missed in my skim of the post).

grenli•6m ago
The learned confidence gate is the crucial piece for a 14MB action model. On ambiguous requests such as the HN example, what calibration target decides between abstaining locally and escalating to the cloud?
redrix•5m ago
This is cool!

While most of the industry focuses on the frontier of “intelligence” (function), a release like this represents the frontier of the other end of the spectrum (form).

Both are important if we ever want to see “Opus-level” capability running locally on commodity machines in the future.

HenryNdubuaku•3m ago
thanks!
ianseyler•4m ago
I’d be interested in attempting to run this in a network- enabled 32MiB RAM microVM.
HenryNdubuaku•3m ago
Thanks, how can we help?